A Literature Review of Recent Advances on Innovative Computational Tools for Waste Management in Smart Cities

Abstract

This article reviews the literature surrounding innovative computational tools for waste management within smart cities. With the rise of urbanization and the increasing challenges of waste management, innovative technologies play a pivotal role in optimizing waste collection, sorting, recycling, and disposal processes. Leveraging computational tools such as artificial intelligence, Internet of Things, and big data analytics, smart waste management systems enable real-time monitoring, predictive modeling, and optimization of waste-related operations. These tools empower authorities to enhance resource efficiency, minimize environmental impact, and improve the overall quality of urban living. Through a comprehensive review of recent research and practical implementations, this article highlights the key features, benefits, and challenges associated with the development of cutting-edge computational tools for waste management. Emerging trends and opportunities for research and development in this rapidly evolving field are identified, emphasizing the importance of integrating technological innovations for building sustainable and resilient waste management in smart cities.

1. Introduction

The rapid pace of urbanization has presented new challenges for waste management in cities worldwide [1]. The increase in generated waste poses challenges for traditional waste management systems (WMS). To address these challenges, smart cities use technological innovations to improve various aspects of urban life, including waste management. The expansion of urban areas and changes in consumption habits are crucial factors in the development of cities worldwide. Population growth and socio-economical factors are closely linked to waste production. Waste management is a widespread issue in urban areas. It is estimated that more than two billion tons of waste are generated per year globally, and at least 33 percent of that is not collected in an environmentally safe manner and is deposited in open dumps [2]. However, the collection rate is highly dependent on the region, socioeconomic level, and the type of urban area. Waste management poses a significant challenge due to the complex nature of the collection and disposal process, which is influenced by legal, social, political, environmental, and economic factors.

This article is motivated by the critical need to explore and assess recent advancements in computational tools specifically tailored for waste management within the context of smart cities. With the advent of cutting-edge technologies such as artificial intelligence (AI), Internet of Things (IoT), and big data analytics, a unique opportunity exists to revolutionize waste management practices [3]. Harnessing the power of these computational tools allows municipalities and waste management decision-makers to optimize waste collection, sorting, recycling, and disposal processes, thereby improving resource efficiency, minimizing environmental impact, and enhancing the overall quality of urban living.

The main contribution of this article lies in its comprehensive review of recent research and practical implementations in the field of smart waste management, particularly in the application of computational intelligent tools to waste management. By synthesizing and analyzing a diverse range of studies and case studies, this article aims to elucidate the key features, benefits, and challenges associated with innovative computational tools for waste management in smart cities. Furthermore, it seeks to identify emerging trends and opportunities for further research and development in this rapidly evolving field. Through this exploration, the critical importance of integrating technological innovations into the fabric of modern WMS is delved into, laying the groundwork for the creation of sustainable and resilient smart waste managements in cities of the future.

The article is organized as follows. Section 2 presents an overview of waste management in modern smart cities. Section 3 describes the research methodology for the systematic search to identify relevant articles contributing to the field. A review of existing reviews on the topic is presented in Section 4. Section 5 presents the developed review, focusing on routing and optimization in WMS; computational intelligence, data-driven approaches, and IoT systems; and social aspects of waste management. Section 6 discusses the main findings, including emerging trends and opportunities for further research and development in this rapidly evolving field, emphasizing the importance of integration of technological innovations for building sustainable and resilient WMS in smart cities. Finally, Section 7 presents the conclusions and interesting lines for future work.

2. Waste Management in Modern Smart Cities

Waste management is the process of collecting, transporting, processing, recycling, and disposing of waste materials in an environmentally responsible and economically efficient manner [4]. It has been a fundamental aspect of urban life since ancient times, given the substantial volume and diversity of materials generated by human activities [5]. Improper waste disposal usually leads to air, water, and soil pollution, posing significant health hazards to humans and wildlife alike.

WMS are divided into several stages [6]. The first stage is waste generation. Waste is typically generated in dwellings, small businesses, and institutions. The next stage is waste collection, which is performed in different ways depending on the collection type that is implemented. In the door-to-door collection type, waste is collected through a door-to-door service by collection vehicles visiting every dwelling regularly. In the community bins type, waste is collected at designated collection points distributed throughout the city. Although this method requires citizens to carry their waste to these points, it is preferable in densely populated urban areas, as it reduces the number of stops for collection vehicles. After the collection, waste is carried to the treatment stage, which varies in complexity depending on the level of technological advancement of WMS. Some cities may only compact waste at transfer plants to reduce logistics expenses, while others employ more advanced processes such as sorting waste into different fractions and employing chemical and biological treatments. Finally, waste is sent for final disposal in landfills or to recycling plants to recover useful materials. The main stages of the system are shown in Figure 1.

Overall, effective waste management requires combining strategies aimed at reducing waste generation, promoting recycling and reuse, and ensuring safe and environmentally sound disposal practices. It is a complex and multifaceted process that requires cooperation and coordination among governments, businesses, communities, and individuals to achieve a sustainable and efficient outcome. In this regard, modern smart technologies contribute to enhancing the efficiency and sustainability of several stages of waste management.

The various stages of waste management have prompted the development of innovative computational tools. In the waste generation stage, occurring at households, researchers have focused on predicting waste generation patterns [7], identifying valuable resources within waste to reduce waste output [8], implementing smart charging systems that assign costs based on household waste production [9], and employing smart source classification to facilitate subsequent recycling stages [10], most frequently using computational intelligence [11]. For the accumulation stage, at waste accumulation points, studies have explored bin locations [12], smart bins equipped to monitor waste accumulation levels [13], and smart technologies for optimizing capacity design [4]. In waste collection, researchers have tackled the waste collection routing problem using computational methods, both in static [14] and dynamic routing scenarios [15,16], usually applying heuristics or metaheuristic methods [17]. Finally, in the last stage of WMS encompassing transfer plants, treatment plants, and/or final waste disposal, computational intelligence tools have been utilized to introduce smart industry concepts into recycling or waste transformation processes [18], develop advanced separation mechanisms for waste classification [19], and explore Waste-to-Energy technologies [20]. Waste management in modern smart cities is a multifaceted endeavor that demands innovative solutions and technological integration. By harnessing the power of computational tools, waste management authorities pave the way towards sustainable and resilient urban environments. Figure 1 also summarizes the applications of computationally intelligent tools in WMS.

In this line of work, this review article focuses on providing a review of recent proposals applying innovative computational tools for waste management. The article pays special attention to the most pressing issues in waste management, including smart ways of dealing with increasing waste generation, the limitations of existing waste infrastructure, the risks of handling special and hazardous waste, innovative ways to promote recycling and circular economy, and citizens’ awareness and behavioral changes.

Computational intelligence plays an important role in mitigating the environmental impact of waste management by providing innovative solutions to reduce waste generation, promoting recycling, offering advanced waste treatment and waste-to-energy conversion alternatives, enabling sensor-based monitoring of waste collection systems, and improving traceability and transparency in waste management systems. In turn, new technologies also help to mitigate health hazards associated with waste management, by proposing improved techniques for waste handling and containment, safe protocols for handling toxic waste, technological devices for pollution control and emissions reduction, automated systems for remote monitoring and early warning systems, tools for promoting public awareness and education, advanced data analytics to assess health risks, and disease surveillance and outbreak management using sensor networks and AI.

Finally, new technologies have the potential to reduce social inequalities in waste management by implementing improved infrastructure and decentralized waste management solutions focusing on less wealthy neighborhoods, developing accessible solutions based on mobile applications to disseminate information and engage communities in waste management practices, promoting inclusive recycling programs, applying low-cost waste-to-energy solutions for less favored communities, developing advanced decision support systems to consider specific waste generation patterns, identifying areas with inadequate waste infrastructure, and predicting future needs, especially in underserved communities, to properly distribute the available resources in an equitable way to address disparities between rich and poor regions within a city. New technologies must be applied jointly with inclusive policies, adequate funding, and community engagement to ensure equitable access and benefits for all, taking into account social, cultural, and economic contexts to address the specific needs and challenges of different communities.

3. Research Methodology

This section presents the methodology for the systematic literature review, applied to retrieve the most relevant articles of recent advances on innovative computational tools for waste management in smart cities and a bibliometric analysis of related works.

3.1. Systematic Literature Review

The primary objective of a systematic literature review is to provide readers with a comprehensive understanding of the current state of the art pertaining to a specific topic or research question, elucidating how past developments can be applied in present or future scenarios [21]. It entails a meticulous depiction, synthesis, critique, and impartial evaluation of existing knowledge through a robust and rigorous methodology [22]. Additionally, classifications and conceptual categorizations play a pivotal role in aiding readers’ comprehension of the key similarities and distinctions among the reviewed proposals [23].

Regarding the applied methodology, the presented study falls within the ambit of systematic search and review [24]. The study encompasses two primary components: (i) a thorough mapping, categorization, and quantitative/qualitative analysis of extant literature, and (ii) a critical appraisal elucidating the reported methodologies and outcomes. The initial step in this process involves delineating a pertinent research topic or question to guide the scope and focus of each review.

The central subject of this review is recent advances in innovative computational tools for waste management in smart cities, known as smart WMS, a cutting-edge and highly relevant research area. Existing studies from the past five years are reviewed to analyze recent innovative computing methods for waste management. After a careful review of the related literature, we classified the relevant articles into three main topics: (i) computational intelligence and data-driven approaches, including routing and other optimization problems in waste management; (ii) computational intelligence and IoT in WMS, including blockchain, waste identification and classification, waste generation characterization and forecasting, smart bins, and IoT systems; and (iii) social aspects of waste management in smart cities. These social aspects include circular economy, efficiency analysis, field studies, and other social aspects of waste management.

Subsequently, a systematic and exhaustive search was conducted to identify pertinent articles, reports, and theses contributing substantially to the field. Ensuring the quality of a review relies heavily on the reliability and accessibility of data sources. In this review, the search primarily relied on two major scientific databases, namely, Scopus and Web of Science (WoS), to encompass a wide array of peer-reviewed journals pertinent to the subject matter. The period to retrieve the articles was set to 2019–2023, to capture the most recent advances in the subject. Employing the TITLE-ABS-KEY schema with specified keywords such as ‘waste management’, ‘smart cities’, ‘computational intelligence’, and related terms minimized the risk of bias in the search process. A total of 267 articles were retrieved from Scopus, with an additional 270 from WoS, resulting in 161 overlapping articles between both databases. Consequently, 376 unique articles were subjected to analysis.

The initial scrutiny focused on evaluating titles, keywords, abstracts, and sources. Subsequently, 81 articles were deemed unsuitable because they did not apply computational intelligence tools, dealt with other types of waste (not solid waste), were conference papers incorrectly classified as journal articles in the consulted databases, or were of poor quality with weak research methodologies. These unsuitable articles were discarded. The remaining 295 articles underwent another phase of thorough filtering through content analysis. This phase involved the application of predetermined eligibility criteria and expert judgment to filter out relevant articles. Each article underwent meticulous scrutiny, with particular attention paid to its indexing information to ascertain its relevance. Articles unrelated to computational intelligence tools in WMS of smart cities, or those presenting outdated, non-formalized, or unrealistic studies, were excluded from the review. This meticulous filtering process culminated in the identification of 180 articles to be included in the review. Among these, 23 were identified as review articles and the remaining 157 constituted original research articles. The final stage entailed organizing the selected articles into categories aligned with existing research topics and the objectives of the review. Three primary categories were defined, focusing on specific features of the problems and variants, methodological approaches, and research goals.

3.2. Bibliometrics

Bibliometric analysis has become a powerful tool for evaluating scholarly literature. By quantitatively assessing publication trends, citation patterns, and collaboration networks, bibliometric studies provide insights into the growth, impact, and dynamics of research. With the exponential increase in scientific output, this approach systematically identifies key contributors, landmark studies, and emerging trends. This study aims to deepen the understanding of waste management in smart cities through bibliometric indicators and inform future research directions and strategic decision-making.

Figure 2 presents a schematic analysis of the main keywords found in the set of analyzed articles conducted using VOSviewer. Only keywords with at least five repetitions were plotted. Prior to the analysis, a meticulous revision was conducted to merge very similar words, e.g., “smart city” and “smart cities” or “waste management” and “solid waste management”, to ensure representative results. Then, the most frequently occurring keywords were “smart city” (48 occurrences), “internet of things” (43), “artificial intelligence” (40), “machine learning” (33), “waste management” (30), “sustainability” (21), “circular economy” (17), “systematic review” (15), “industry 4.0” (14), “big data” (12), “optimization” and “multi-criteria decision” (both 10), and “deep learning” (9).

The main journals publishing articles on this topic are Sustainability (MDPI) and Journal of Cleaner Production (Elsevier), both with 14 articles; Sensors (MDPI), with 9 articles; IEEE Access, with 8 articles; Applied Sciences (MDPI) and Sustainable Cities and Society (Elsevier), with 6 articles; and Smart Cities (MDPI), with 5 articles. Thus, these journals have typically published cutting-edge advances in smart WMS.

Figure 3 presents the number of articles published per year. A clear increasing trend is shown, with the number of articles rising steadily from 21 in 2019, 25 in 2020, 40 in 2021, 45 in 2022, and reaching a peak of 46 in 2023. The seven articles in 2024 include those first published online in 2023 and subsequently included in the 2024 volumes of the journals.

Figure 4 presents the number of articles published per main topic.

From the total number of 180 articles reviewed, 13% were reviews. Among the application articles, the majority were focused on route optimization for WMS (15%), reviews (13%), and IoT systems (11%). Other relevant topics included waste identification and classification (9%), smart bins (9%), waste characterization and forecasting (7%), smart WMS and circular economy (7%), and data analytics and other optimization problems (7%).

Based on this classification, three main topics for discussion were proposed: (i) Computational optimization methods in waste management (41 articles, 23%), including route optimization for WMS and data analytics and other optimization problems; (ii) Computational intelligence and IoT in WMS (81 articles, 45%), including blockchain applied to smart waste management, waste identification and classification, waste characterization and forecasting, smart bins, IoT systems, and computational intelligence for other relevant problems in waste management; and (iii) Social aspects of WMS in smart cities (36 articles, 20%), including smart WMS and circular economy, field studies of WMS, efficiency analysis of WMS, and other social-related aspects of WMS. The category of computational intelligence and IoT in WMS had the largest number of articles.

4. Existing Reviews

This section discusses previous reviews focusing on the application of smart waste management technologies.

Nižetić et al. [25] presented a compendium of articles on smart technologies addressing modern city issues, including smart waste management. The compendium was not an exhaustive review, but a collection of conference articles related to sustainability and smart resource usage. Several topics were discussed, such as the relationship of WMS with recycling, circular economy, and plastic pollution. Pardini et al. [26] reviewed the application of IoT architecture and protocols in WMS. Based on the review of relevant articles on the subject, the main finding was the lack of IoT-based solutions that enhanced the citizens participation in waste management. The involvement of citizens contributes to cutting the expenses of WMS mainly by reducing collection times. In a brief study, Rasool et al. [27] analyzed 18 articles about IoT technology in WMS in small cities. The analysis was limited to a description of the reviewed articles in terms of the infrastructure and software used for processing without identifying research gaps in the literature.

Sodiq et al. [28] reviewed food waste management in smart cities. Certain practices, such as using food waste disposers, were found to be controversial. Whereas these disposers help treat food waste (50–55% of solid waste in developing countries) and reduce landfill waste, they can also damage sewage systems. Sahu et al. [29] presented a brief discussion of 15 technical reports and scientific articles that studied the application of advanced industrial technologies to increase waste classification in Indian WMS. The review was restricted to studies developed in Indian cities. Ahmad et al. [30] reviewed the application of blockchain technology in WMS. The study highlighted the capacity of blockchain to enhance WMS processes such as real-time tracing, tracking, and reliable channelization. The review also discussed the integration of blockchain-based smart contracts into waste management practices, emphasizing their role in ensuring compliance with waste handling regulations.

Akram et al. [31] reviewed articles applying Information and Communication Technologies (ICT) in waste management. The study emphasized the importance of selecting the appropriate wireless communication protocol in the implementation of IoT-based smart bins to optimize the energy usage of batteries, thereby maximizing their operational efficiency. Blockchain was identified as a cornerstone technology in establishing interconnected networks involving dwellings, authorities, waste collectors, and managers. Such an integrated system ensures that each entity fulfills its responsibility in the waste management process, ultimately contributing to more effective and sustainable WMS. Shash et al. [13] reviewed 16 articles related to smart waste level monitoring systems published in 2015–2020. Most solutions used SMS text messages for smart bin communications. There was a greater variation regarding the types of sensors and micro-controllers used to operate the system software. In a general review of IoT technologies in smart cities, Ramirez et al. [32] dedicated a brief section to review 22 articles dedicated to discussing IoT technologies for smart bins.

Cheng et al. [33] analyzed 13 waste management articles in Malaysia and discussed strategies to promote smart waste management in three Malaysian municipalities. Concari et al. [34] presented a bibliometric analysis to study the recycling behavior of citizens after revising 2061 articles from the literature with text mining tools. Although the main topic of the review was not directly related to smart cities, the authors concluded that the implementation of smart city policies contributes to enhancing the recycling behavior of citizens. D´Amico et al. [35] reviewed articles that aim to digitalize the cycle of resources in a modern society from products to be used by citizens to waste (the “urban metabolism circularity” approach). The application of ICT to digitalize the handling of waste in modern cities that encourage circular economy was discussed, and some case studies were presented.

Mousavi et al. [36] reviewed the application of IoT technologies in waste management, focusing on data acquisition, transmission, and processing. The review included spatial technologies for WMS: Geographical Information Systems (GIS), the Global Positioning System (GPS), and Remote Sensing. Namoun et al. [37] reviewed the application of machine learning in the flow organization of solid waste management in smart cities. The study examined 23 articles, emphasizing the use of artificial neural networks (ANN) for predicting waste generation and waste disposal behavior. Specific challenges were identified, such as the scarcity of up-to-date waste datasets and benchmark case studies for comparison.

Sosunova and Porras [38] conducted a comprehensive review of 173 publications spanning from 2014 to 2022, focused on the integration of IoT technologies in waste management within smart cities. The literature was categorized into two main groups, addressing city-wide issues of WMS (such as route planning, system optimization, and environmental impact analysis) and the technologies for smart waste bins, focusing on data gathering and transmission. The main contributions outlined different sensor and actuator technologies and their applications in various WMS, identified the direct and indirect stakeholders that benefit from the gathered information, and delineated the types of data shared between WMS and stakeholders. The review also acknowledged the potential to further leverage IoT-generated data to enhance decision-making processes in waste collection, reduce operational costs, improve service quality, and mitigate environmental impacts.

Our previous article [4] presented a extensive review of strategies for waste bin locations, which plays a crucial role in enhancing the overall efficiency of WMS. The primary objective of the review was to provide decision-makers in this field with a valuable point of reference for strategies to apply in modern smart cities. The main results of the analysis indicated that numerous optimization criteria and resolution approaches have been utilized. However, only a limited number of proposals have simultaneously optimized both the location of waste bins and waste collection. Additionally, few studies have taken into account the uncertainty associated with model parameters or integrated multiple approaches.

Vishnu et al. [39] presented a thorough discussion on the specific technologies used for sensors in waste management, including Radio Frequency Identification (RFID), Wireless Sensor Networks (WSNs), and IoT-enabled sensors, from articles published between 2007 and 2021. The main conclusion was that IoT-based sensors were superior to other technologies for waste management applications, due to their interoperability and dynamic adaptation of nodes. Moreover, LoRaWAN was identified as the preferred communication protocol.

Fang et al. [40] reviewed the application of computational intelligence in WMS for smart cities. The article provides an overview of waste types, production, and associated issues in waste management, divided into nine sections. Practical AI applications were highlighted as impacting waste management, e.g., in smart bin systems, waste-sorting robots, and predictive waste-tracking models. Also, AI could help manage hazardous waste, reduce illegal dumping, and recover valuable resources from the waste stream. Furthermore, the article discussed the impact of AI on waste logistics and transportation, including reducing distance and cost and improving collection efficiency. Finally, the article concludes that AI will change the interaction between people and waste through a sustainable future, including economic, ecological, and intelligent WMS.

Salman and Hasar [41] analyzed 25 articles related to smart WMS. The challenge of integrating smart bins and real-time information sources for routing optimization into practical, real-world implementations was noted, as most studies were limited to pilot applications. Santibañez et al. [42] analyzed 163 articles related to the applicability of the concept of circular economy in smart cities, published between 2013 and 2022. The review mainly offered a bibliometric analysis without specific discussion about the analyzed articles. Specifically in the context of waste management, the article emphasized how circular economy policies significantly contribute to waste reduction by promoting the reuse of products and the recovery of valuable materials from end-of-life products.

Soo et al. [43] discussed the application of machine learning techniques in enhancing the composting of organic waste. The application of these techniques aims to improve logistics management, monitor waste movement and separation, reduce labor intensities, and predict emission levels from waste decomposition. However, the implementation of machine learning techniques faces challenges due to limited data availability. To overcome this hurdle, the review emphasized the need for robust machine learning models to effectively train despite data gaps, employing edge and fog computing approaches.

Szpilko et al. [44] conducted a bibliometric analysis of current and future research related to waste management issues in smart cities. Several potential directions for future research were identified, including technological advancements in Industry 4.0, specific waste challenges for e-waste, digitization and circular economy, energy recovery and sustainable solutions, transportation in waste management, community engagement and environmental awareness, policy development and standardization, security and privacy considerations, novel frameworks and business models, waste prevention, economic and environmental impact assessment, and global implications of WMS.

Kannan et al. [45] presented a review of articles applying technologies and strategies considered as part of the Fourth Industrial Revolution (sensors, IoT, cybersecurity or blockchain, data analytics, and computational intelligence) in waste management. This study pioneered the establishment of a connection between the concepts of Industry 4.0 and waste management. The paradigm of smart WMS 4.0 was introduced, which comprises four subtopics: smart people, smart cities, smart enterprises or businesses, and smart factories. The review revealed that most articles focused on the subtopic of smart cities, indicating a need for further research in the other three subtopics. The results of the review were also validated through expert opinions, further strengthening the findings.

Figure 5 summarizes the findings on studied reviews on smart waste management technologies. Reviews have highlighted the promising potential of blockchain technologies in improving real-time tracing, tracking, and compliance in waste management through the use of smart contracts. Additionally, machine learning techniques, particularly ANN, have been applied to predict waste generation and disposal behaviors. However, these methods have faced challenges such as the availability of real data, outdated existing datasets, and a lack of benchmark studies to evaluate their accuracy and effectiveness. See Appendix A.

The integration of IoT in waste management has been categorized into two categories: city-wide WMS challenges and smart bin technologies. IoT data holds significant potential for enhancing decision-making, reducing costs, and improving service quality. However, selecting the appropriate wireless communication protocols is crucial for optimizing energy consumption in IoT-based smart bins.

AI applications have shown potential in waste management, particularly in smart bins, waste-sorting robots, and predictive models. Additionally, the optimization of waste bin locations is critical, but often fails to consider both collection and uncertainty factors simultaneously. The implementation of local strategies and the influence of smart city policies on recycling behaviors have yielded positive results. Overall, although advanced technologies (blockchain, IoT, AI) offer revolutionary potential, they also pose challenges such as data limitations and the need for integrated approaches. To fully leverage the capabilities of smart WMS technologies, future research should focus on addressing these gaps and developing scalable, comprehensive frameworks.

5. Review of Innovative Computing Methods for Waste Management in Smart Cities

This section reviews recent advancements in computational tools for waste management in smart cities, covering optimization, computational intelligence, data-driven, and IoT-related methods.

5.1. Topic: Computational Optimization Methods in Waste Management

Computational optimization methods have been applied for efficient waste management, including optimizing waste collection and transportation routes, implementing IoT-based waste collection and disposal systems, and establishing efficient protocols and typologies to prioritize waste information. Smart strategies aim to improve economic efficiency and reduce environmental impact in WMS. Other relevant optimization problems include maximizing revenue from recycling, allocating resources from separation facilities to recovery plants, conducting spatial analysis to identify patterns in informal recycling systems, implementing smart e-waste reverse systems, locating and monitoring waste bins, and optimizing waste sorting and storage, among other applications. See Appendix A.

5.1.1. Route Optimization Methods Applied to Smart Waste Management

Monishan et al. [46] proposed an IoT-based system to minimize waste collection and disposal costs, integrating household bins and automated mobile waste collectors. A routing algorithm was applied to identify the shortest path between the household bins and mobile waste collectors. A simple experimental evaluation on a synthetic scenario combining two kinds of bins under a test-bed-controlled environment was conducted. Results showed that the approach was well-organized for IoT-enabled waste collection and disposal, but no validation or deployment over real scenarios was reported.

Idwan et al. [47] proposed a smart IoT-based WMS to optimize the schedule and routes of waste collection trucks via waste-level sensors in bins, connected to the municipal wireless network. A two-step metaheuristic based on GA computed the most efficient collection routes using the minimum number of trucks. The evaluation was performed over two scenarios: a traditional scenario without IoT, simulated with 115 uniformly distributed bins, and a smart IoT scenario with 64 bins selected from different sectors, employing real-world geographic locations in Islamabad, Pakistan. The proposed approach achieved better results than conventional routes by minimizing the truck time collection and distance traveled in both evaluated scenarios, but no comparison with other methods was reported. A real validation was reported comparing the smart bins against conventional routes.

Ahmad et al. [48] proposed a recommendation system to minimize collection route distance and fuel consumption, and maximize the collected waste. A constrained mixed integer linear optimization model was applied to limit the total allowable distance and time considering bins, solved using Particle Swarm Optimization (PSO), the Genetic Algorithm (GA), and the Bat Algorithm. The evaluation was performed using monthly, weekly, and daily waste data from 2017–2018 from a local area in Jeju Island, South Korea. The PSO computed the best solution regarding the total cost of vehicle fuel consumption and the number of human resources required for waste collection. No real implementation was described.

Hannan et al. [49] proposed a system using real-time bin status to improve collection efficiency, reduce collection costs, and reduce emissions. Fixed and variable routing optimization models were applied. The system was evaluated on a very small toy scenario, considering only one depot, two vehicles, and 20 bins randomly located on an Euclidean plane. The system achieved the targeted goals and demonstrated the feasibility of an optimization model for the waste management sector. No real evaluation was presented.

Nidhya et al. [50] proposed an efficient routing technique for a smart WMS considering quality of service (QoS) to minimize data communication delays. An Enhanced Route Selection algorithm was applied for routing messages to notify a remote server when smart bins reached over 90% capacity. Then, the system notified the pickup truck driver with a map showing the shortest path to the dump yard, visiting all filled bins. The system was evaluated via the ns-3 simulator, demonstrating good performance compared to existing methodologies, considering various QoS parameters. No real validation was reported.

Nowakowski et al. [51] proposed a metaheuristic approach for routing e-waste vehicles, modeled as the Vehicle Routing Problem (VRP). A Harmony Search algorithm was applied to optimize routes from on-demand collection points within specified time windows, from requests of citizens. The evaluation was performed over randomly generated problem instances representing real-life conditions of requesting e-waste pick up from a household. Results showed that the proposed approach demonstrated improvements over greedy, Tabu Search (TS), and Simulated Annealing (SA) methods. No validation over real scenarios was reported.

Wu et al. [52] optimized wet waste collection and transportation, modeled as a constrained capacitated VRP, considering carbon emission costs, smart bins, and waste fill and credibility levels. A hybrid combining PSO and SA was applied to minimize the total cost, and evaluated on three scenarios, including ten datasets from the literature. Waste transportation improved under different credibility levels. No real validation was presented.

Abdullah et al. [53] proposed a metaheuristic approach to optimize path routing for wireless sensor networks, mobile ad-hoc networks, and IoT devices in WMS. A hybrid metaheuristic combining GA and a Dijkstra algorithm to connect broken paths was applied to identify an efficient path satisfying the service requirements for various network infrastructures. The evaluation was performed over three scenarios of four network topologies. Accurate results showed that the proposed approach selected alternative paths in addition to the most efficient path, also exhibiting a shorter distance than a Dynamic Source Routing baseline algorithm comparison. No real validation or implementation was reported.

Lu et al. [54] proposed an ICT-based smart waste classification and collection system to optimize total cost and workload balance for multiple waste disposal centers, transfer stations, and waste bins. A hybrid bi-objective metaheuristic combining whale optimization and GA with improved convergence and non-dominated sorting was applied. The evaluation was performed over a real-world scenario in Pudong, Shanghai, China. The proposed system computed better solutions in terms of multiobjective optimization metrics than Non-dominated Sorting Genetic Algorithm II (NSGA-II) and Multiobjective Evolutionary Algorithm Based on Decomposition. No real implementation was reported.

Kshirsagar et al. [55] proposed a shortest path routing method using real time information gathered by IoT sensors. Two key parameters were considered in the study: the filling level of bins and the distance between nearby bins. Metrics of the IoT-based monitoring system (end-to-end delay, packet size) were included in the analysis to define revenue function for the waste collection process. The shortest path routing was based on iteratively checking the satisfaction of constraints, over graphs successively modified according to the IoT information. A simple evaluation was reported via simulations to analyze relevant parameters and the revenue values. No real validation or implementation was reported.

Ali et al. [56] proposed a framework for optimizing the economic performance and environmental impact of waste management. Data analytics and machine learning models (linear regression, multilayer perceptron, and sequential minimal optimization regression) were used. The system was evaluated over a synthetic dataset, including data on the total amount of raw materials (food, plastics, paper, and agricultural waste), the types of waste conversion technologies, and the expected end products, for a case study in a Malayan local zone. The multilayer perceptron computed the best results in testing data, with a correlation coefficient of 0.7169. The generalization of the framework was not studied.

Akbarpour et al. [57] proposed a metaheuristic approach to solve two sub-models to minimize the transportation cost and maximize recycled revenue for WMS. The first sub-model used the VRP to find the waste collection routes, and the second sub-model allocated resources from separation facilities to a set of recovery plants or landfills. Four metaheuristic algorithms were employed: SA, GA, hybrid GA-SA, and hybrid GA-PSO. Problem instances were built over a baseline industrial example to define ten real-world test scenarios, including small, medium, and large sizes. The results showed that the GA-PSO method found better results than the other methods implemented for the evaluated scenarios. Validation of critical parameters was reported through sensitivity analysis.

Bin et al. [58] applied metaheuristics to optimize the vehicle schedule and routes for collecting bin waste. A modified Ant Colony Optimization (ACO) using a non-dominated sorting mechanism was applied for the split delivery VRP, based on multiple cleaning services for the same collection point according to the real-time waste volume. The system was evaluated in a case study in Tianjin Wudadao, China, with 23 streets, 32 collection points, four types of vehicles, and four classified waste disposal stations. The proposed approach improved over a traditional ACO and an artificial bee colony algorithm. A real validation was reported with the process of classified waste disposal in the studied area.

Manoharam et al. [59] proposed a GIS-based approach to optimize the distance and transportation cost for solid waste management. TS and Dijkstra algorithms through ArcMap/ESRI Network Analyst software were applied. The evaluation was performed over a real instance of 24 companies in an industrial park in Malaysia. The approach computed better waste collection routes, positively impacting the operation cost, carbon emission, and fuel consumption. Although real validation using GIS software was reported, no comparison against state-of-the-art methods was performed.

Mekamcha et al. [60] proposed a metaheuristic approach to optimize truck routes for waste collection. TS and SA were applied to find the best paths for each vehicle in short periods with low costs. The evaluation was performed over real instances of 21 waste collection circuits, considering information from Tlemcen City, Algeria. Results showed that the SA method improved the total distance traveled in the 21 circuits over TS. No real implementation was described.

Shevchenko et al. [61] proposed a smart e-waste reverse system using intelligent local delivery services. Interactive online maps were used to request the collection of e-waste, turning potential mobile collection points into active locations for vehicles to optimize the transport routes. The system was evaluated over a dataset of collection points according to the local requests and specialized transports in Sumy, Ukraine. A cost–benefit analysis was reported to provide value to all owners of electronic equipment at the end of its useful life by optimizing the transportation routes of collection vehicles. The proposed system selected the most efficient collection route for local e-waste reverse operators of specialized enterprises. However, no specific details of real-life implementation were reported.

Saeidi et al. [62] presented a multi-objective model to maximize the total profit of hazardous waste management and minimize both CO₂ emissions and visual pollution. The problem was modeled as a capacitated VRP, and the NSGA-III method was applied and tuned through the Taguchi method. The experimental evaluation was performed using GAMs software over a case study to determine the transportation cost for 22 districts of Tehran, Iran, for 20 randomly generated instances with different sizes. Results showed that the proposed approach reduced the total costs and increased the profitability of hazardous waste management systems. No real validation or implementation were described.

Salehi et al. [63] studied a two-stage multi-objective model using IoT for waste collection routing. The first stage decided which bins should be emptied at a given time, using real-time IoT information about the filling level of bins and the location of collecting vehicles. Wet and dry waste was considered, to be transported to a separation facility. Sensors were used to determine the waste generated in each area. To optimize the route schedule, minimize route distance, and reduce the number of required vehicles, the green capacitated VRP was solved. The optimization process took into account the number of non-visited bins, to design efficient waste collection routes that reduce travel time, load/unload stops, air and noise pollution, and traffic congestion. The second stage segregated waste into six categories (paper, plastic, metal, glass, electronic, and wet) using a multi-objective green split pick-up VRP to maximize value recovery, and minimize distance and air and visual pollution and guarantee a proper load balancing during transportation. Several metaheuristic algorithms were applied for the first sub-model, and the $ϵ$-constraint method was applied for the second sub-model using the CPLEX solver. The experimental evaluation was performed over synthetic scenarios, without using real IoT data. The studied algorithms achieved accurate results, but the analysis did not perform a comparison against state-of-the-art optimization methods. In turn, no validation over real scenarios was reported.

Burduk et al. [64] proposed an approach for waste collection routing using real data. The problem took into account the travel time of vehicles between waste bins and different filling rates. A TS and a GA were proposed to solve the problem, but the article lacked a formal description of the optimization methods. The evaluation was performed using simulations and undisclosed company data. A prototype design of a smart bin using IoT sensors was introduced, but no specific details of the implementation were presented. Several variants of the developed algorithms were studied, but no comparison with other methods was reported. No real validation or implementation were described.

Ghahramani et al. [65] proposed an IoT-based model for smart bin waste collection including intelligent vehicle routing strategies coupled with spatial constraints. The problem was modeled as discrete and continuous optimization problems. The problem-solving approach used GA and ANN to compute the associated cost in each iteration of the GA. The evaluation was performed over real-world instances of bins in Docklands, Dublin. The discrete model was superior to the continuous one in spatial optimization and computational time, but the analysis did not compare against state-of-the-art continuous methods. A visual validation of a case study was reported.

Abdullah et al. [66] proposed an IoT-based system to optimize organic and medical waste collection and transportation. The system received messages when a smart bin exceeded the threshold value of the residual free space rate. The problem was formulated as a constraint satisfaction model considering all smart bins along the route to the destination. A GA was proposed to solve the problem. The evaluation was performed over a case study involving a single truck handling one waste type across 50 locations over a 1000 m² area in the Khalidiyah zone, Saudi Arabia. The problem instance included waste information such as smart bin status and type, focusing on source waste locations and public service areas. Results showed that the proposed approach enhanced operational efficiency, reducing the number of locations and smart bins that needed to be visited by 46% for both waste types. A real validation was reported contrasting conventional and existing systems, which required daily visits to all locations, but no comparison with other methods was reported.

Hu et al. [67] proposed a framework to identify sustainable waste collection routes. A hybrid multi-objective metaheuristic combining PSO and GA was applied for different waste collection modes under classification. The evaluation was performed over real instances from Zhangjiagang City, Jiangsu Province, in China. The proposed framework achieved over NSGA-II in hypervolume, error rate, and execution time, indicating better solution efficiency and quality. A validation of the case study of Zhangjiagang City, a pioneer in waste classification in China, was reported.

Barth et al. [68] proposed a trade-off WMS model between cost savings and service quality using waste bin sensor modules. An agent-based simulation model with discrete event simulation elements was applied to reduce costs by adjusting the waste collection frequency and the number of waste bins included in the collection route, based on their fill levels. The evaluation was performed via simulations using the Anylogic software over a dataset from a Swiss municipality involving 98 waste bins. The proposed model achieved significant cost savings even with values set below 50% of the maximum bin fill level without compromising service quality. No validation was reported due to confidentiality.

Saha and Chaki [69] proposed a framework for IoT-based smart WMS to collect waste from COVID-affected households. The framework works in three steps: a smart bin deployment, a hybrid network architecture for cost-effective waste-level monitoring based on QoS, and shortest path computation between the smart bins and the waste collectors. The evaluation used data from public and non-biodegradable waste bins on COVID-affected residents in two public regions and two residential areas. The framework ensured an efficient solution for handling waste generated in urban areas, offering dynamic waste collection scheduling and route optimization. A validation of the hybrid network architecture strategy for efficiently managing garbage bins in public and residential areas was reported. The generalization of the proposed framework was not studied.

Bouleft and Elhilali [70] proposed a metaheuristic approach to minimize total transportation costs by including penalty costs for exceeding bin capacity in waste collection. A hybrid GA was applied for the dynamic multi-compartment VRP in separate waste collection for each smart bin equipped with a sensor at each compartment. The evaluation was performed using CPLEX over real-world instances built with real-time data of waste quantities of different waste types at each station corresponding to a bin from Valorsul Company. The proposed approach selected the most efficient route for the multi-compartment problem, ensuring that at least one container had a fill level greater than or equal to 70% capacity. No comparison against state-of-the-art optimization methods was described.

Boudanga et al. [71] proposed a novel medical waste management approach, integrating collaboration and technological advancements to optimize waste sorting, storage, and treatment operations. The proposed optimization approach for vehicle routing collection and distribution of medical waste used sensor-based systems, GPS-enabled vehicles, and explainable AI. The experimental evaluation was performed on real-world instances with ten treatment centers and four waste transport vehicles with a capacity of 250 units, based on a case study in Casablanca, Morocco. The implementation used decision trees and Random Forest (RF) models from the scikit-Learn framework for predicting the filling level of medical waste, providing transparent and reliable insights for waste management. Results showed that the proposed approach enhanced operational efficiency, reduced costs, and promoted better resource utilization. A validation over a synthetic dataset and a real-world dataset was performed, but no comparison with other methods was reported.

Rahmanifar et al. [16] proposed a metaheuristic approach for optimizing waste collection management. Real-time data from IoT devices embedded in bins were used to monitor waste levels. The optimization process involved a two-step VRP to collect waste, transport it to separation centers, and transfer it to recycling centers. The first step minimized transportation costs and CO₂ emissions. The second step minimized the transportation cost of high capacitated vehicles. The experimental evaluation was performed on over fifteen synthetic instances of different sizes (number of bins and low-capacitated vehicles). Nine heuristics were used to find the initial solution and the first bin of each tour. Other nodes of the routes were determined based on a nearest neighbor strategy. The metaheuristic optimization applied SA and Social Engineering Optimizer. Experimental results indicated that SEOH5 and SAH4 computed the most cost-effective routes. No real validation or implementation was reported.

Pollak et al. [72] proposed an approach for efficiently managing the autonomous emptying process and predicting bin filling levels. Autonomous robots were used for collecting waste, assuming different filling rates across route planning; thereby, minimizing energy consumption was crucial. The problem was solved using a greedy algorithm for route planning, and XGBoost binary classifier for predicting the filling levels of bins. The experimental evaluation was performed over synthetic datasets built with 49 bins from a real-world operational service area in James-Simon and Monbijoupark, Berlin. Results showed that the XGBoost binary classifier achieved an 82% accuracy in predicting filling levels, compared to a baseline of 59%. Incorporating the predicted filling level data in the route planning reduced operational time by 26% and energy consumption by 31%. No route optimization methods were studied to provide efficient energy and time for the robots.

5.1.2. Data Analytics and Other Optimization Problems in Waste Management

Ali et al. [73] proposed a smart WMS based on IoT for monitoring waste bins and solid waste collection in urban areas. The proposed system utilized IoT devices to wirelessly monitor and control electric waste bins, and a statistical model was used to determine the optimal distribution of bins based on size and distance. The system was evaluated over a dataset of collected data to predict filling levels and estimate the monthly and yearly waste volume from Najran City, Saudi Arabia. A comparison with well-known existing systems was reported. Using IoT devices allowed the proposed system to achieve improved efficiency in waste collection compared to non-IoT methods.

Lee et al. [74] proposed a GIS spatial analysis approach to identify patterns within the informal recycling system, specifically regarding population demographics, waste levels, and urban planning, and focusing on the prevalence of junk shops. A linear regression model was applied to predict the number of junk shops in specific areas in Seoul and Daejeon, South Korea. The evaluation was performed over a dataset combining information about the location of junk shops gathered from the Korean web portal and interviews with junk shop owners, urban planning researchers, and government officials. Results showed that despite policy changes, informal waste-gathering continued to exist in nearly every populated area and that the spatial layout of these junk shops was related to population demographics, spatial characteristics of the studied area, and dominant industries. The spatial analysis presented difficulties in isolating spatial patterns of formal and informal sectors.

Chen et al. [75] proposed an optimization approach based on IoT to minimize waste generation and transshipment costs. A multi-objective model was applied based on NSGA-II. The Ward clustering method was applied to group waste generation points with transfer stations. The evaluation was performed over an instance involving information from three administrative districts of Huangshi, China. The proposed approach achieved accurate results and suggested the construction of thirteen waste transfer stations in Huangshi, with five in the Huangshigang district, four in the Xisaishan district, and four in the Xialu district. However, the analysis did not perform a comparison against state-of-the-art optimization methods. A validation with a practical application in China was reported.

Kaya et al. [76] proposed a waste-to-energy framework to predict the energy obtained from municipal solid waste by thermal conversion. A Gradient-boosting machine learning model was applied to estimate the energy generated in a transfer station using predicted daily waste data and waste characterization information. The framework was evaluated over a real dataset from the Center of Istanbul Waste Management and City Cleaning, considering meteorological and socio-economic features. Accuracy results showed that the proposed framework improved from 2% to 19% compared to the currently used machine learning models in the literature. No real waste-to-energy validation was reported.

Ahire et al. [77] proposed a GIS-based multi-criteria decision analysis (MCDA) to identify suitable potential landfill sites. The study combined spatial analysis using ArcGIS software and ERDAS Imagine 2018 software for satellite image mosaic and sharpening. The analysis was evaluated via multispectral images from a primary satellite and the Digital Elevation Model with a 30 m spatial resolution from the Nashik and Environs municipalities in Maharashtra, India. The proposed analysis achieved precision of the consistency ratio, indicating 15 potential landfill sites located in the wasteland near the municipal corporation and away from populated areas. No real implementation was described.

Shahab and Anjum [78] proposed an approach applying a multipath Convolutional Neural Network (CNN) to identify and locate waste dumps on streets and roadsides in images. The approach performed two steps: a binary classification to categorize waste and non-waste, and a classification to locate waste regions generating a mask in the images classified as waste. The evaluation was performed over a dataset of 12,000 images, including waste and non-waste types, compiled from various sources on the internet and captured through a camera. The proposed approach achieved a classification accuracy of 98.33%, demonstrating suitability for locating waste regions. No real implementation was reported.

Roshan and Rishi [79] optimized a path selection routing protocol for continuous waste monitoring and disposal based on IoT in a smart WMS. The approach, named Seline Trustworthy-based TSP algorithm, operated on an IoT network to enable faster data communication about the level of waste in the bins, ensuring timely disposal. The approach was simulated using MATLAB for problem instances with 50, 100, 150, and 200 nodes. Results showed that the proposed approach improved over traditional methods, such as Artificial Bee Colony, ACO, and PSO, based on measures like delay, throughput, energy, and packet delivery ratio. Accurate results showed that the proposed approach achieved a maximal packet delivery ratio of 99%, a minimum delay of 0.11 s, and a maximal throughput of 38,400 kbps. No real validation or implementation was reported.

Ijemaru et al. [80] used Internet of Vehicles (IoV) for waste management in smart cities. Vehicles were used as opportunistic mobile data collectors to optimize energy efficiency. IoV and big data infrastructure were used to exploit embedded intelligence, cyber-physical systems, and data communication systems for autonomous operation. The network topology comprised static source nodes (smart waste bins embedded with sensors), mobile nodes (collection vehicles), and sink nodes (big data collection centers). Data collected from the smart bins were transmitted to the big data centers. Four known communications routing protocols based on shorter route selection were evaluated via simulations. Results showed that the approach was energy-efficient, but it depended on application scalability and flexibility, latency constraints, security authentication, complexity, incompatibility among devices, reliability in critical applications, and mobility challenges. Later, the authors proposed a swarm intelligence approach for opportunistic IoV data collection in WMS [81]. Four swarm intelligence-based routing protocols were studied to minimize latency times and optimize energy consumption for data collection. The evaluation was performed via ten simulations over each routing protocol using different metrics, i.e., packet delivery ratio, throughput, network lifetime, energy consumption, energy efficiency, and latency. Results showed that the Improved Energy-Efficient Ant-based approach prevailed over the other tested methods. No comparison against state-of-the-art optimization methods was performed. No real validation or implementation was reported.

Demircan et al. [82] proposed a multi-criteria decision-making (MCDM) approach for sustainable WMS in smart cities. The approach combined fuzzy analytic hierarchy process and fuzzy technique for order preference by similarity to obtain the ideal solution using fuzzy TOPSIS methods. The evaluation was performed over a synthetic dataset from interviews with ten experts in academia and the private and public sectors. Results revealed that authorities should consider operational feasibility and initial investment costs before implementing sustainable waste management strategies. No real validation or implementation was reported. Pellatt and Palfreman [83] proposed a smart technology solution to enhance solid waste management service performance. A descriptive statistics method was applied to analyze the data and observe the impact on the service provider and the local authority across various parameters influenced by the implementation of the proposed system. The evaluation was performed over a dataset of 421 citizen surveys from Dar-es-Salaam, Tanzania. The proposed system increased total revenue collection post-implementation, validating the proposed system. However, there was a disparity between the increase in refuse collection charges rates and transaction coverage.

Aloui et al. [84] proposed a WMS based on a generic and comprehensive context ontology and smart containers. The proposed WMS used an intelligent and adaptive IoT system, considering various waste contexts, objectives, and activities. The system was evaluated over a synthetic dataset created through daily and monthly waste collection reports, focusing on waste generation, collection, transformation, and segregation from the Municipality of Jeddah, Saudi Arabia. The smart sensors and IoT technologies gathered and tracked real-time data, allowing administrators to dispose of, reduce, reuse, and prevent waste collection peaks. The proposed system generated optimal paths and improved collection points through user interaction. No real validation was reported.

5.2. Topic: Computational Intelligence and IoT in WMS

This subsection reviews articles applying computational intelligence methods and IoT technologies for waste management in smart cities. See Appendix A.

5.2.1. Blockchain Applied to Smart Waste Management

Sen Gupta et al. [85] addressed the challenges linked to conventional WMS, and proposed an approach applying concepts from blockchain and Business Process Model and Notation for smart planning using smart waste bins. Different stages and algorithms were proposed to express the functionality of the entities involved in the process: waste dumping, sensing and updating the status of bins, finding nearby vehicles, sending notifications, updating the state of vehicles and sending responses, generating collection notifications, collecting waste, setting segregation status, transferring incentive to the user, and finding the best route for disposal. The interoperability of the entities was also described. A case study was presented, comparing the proposed smart WMS with a traditional strategy, highlighting the benefits of using blockchain technology for addressing limitations of traditional WMS.

Kuok and Promentilla [86] studied the adoption of blockchain technologies for waste management in smart cities. T-Spherical Neutrosophic Fuzzy Decision Making Trial and Evaluation Laboratory were applied to develop an analytic network process to analyze the relationships between the identified barriers for blockchain adoption, according to a literature review and the opinion of experts. A case study was presented in Phnom Penh, Cambodia. The numerical results were analyzed quantitatively, applying chord diagrams for the identified barriers, and a causal map for the main problems. Results showed the importance of policy intervention to stimulate investments to develop the physical infrastructure for blockchain solutions and enforce a robust legal data privacy framework.

Liu et al. [87] proposed a framework for managing construction waste information applying blockchain. The main goal was applying blockchain to optimize the recycling process of construction waste via a credit system for stakeholders. Smart contracts were applied to determine the responsibility and ownership of the relevant stakeholders and compute their value-contribution. Information exchange was studied at different levels (user, application, service, and infrastructure). The discussion studied the transparency and traceability of information and value contribution of stakeholders, the challenges of applying blockchain technologies in the construction sector, and the theoretical and practical contributions of the developed approach. No real evaluation was reported.

Santhuja and Anbarasu [88] applied blockchain, smart contracts, circular economy, IoT, and LoRa for e-waste management. A monitoring infrastructure was proposed using sensors to transmit relevant data to a central server for defining collection scheduling policies. Smart contracts were applied for e-waste collection and transportation processes and a blockchain-based approach was applied for sustainability. The computation of relevant indicators was described, but no evaluation of the proposed system was performed.

Hui et al. [89] provided a short discussion on how blockchain technology can enhance waste management by increasing system transparency and fostering trust among stakeholders. However, the authors emphasized the necessity of an established protocol for introducing data into the system to ensure the validity of the process. The article briefly discussed the application of AI and deep learning in waste management to detect fraud, forecast waste generation, and optimize waste collection. Only descriptive analyses were offered. No real validation or implementation was discussed.

5.2.2. Waste Identification and Classification

Mittal and Mittal [90] proposed a generic approach for waste segregation using cloud computing. A portable moisture sensor was proposed to be used to analyze waste. A cloud-based processing method applying Map-Reduce using Hadoop was proposed for data analysis over a Raspberry Pi, using a ESP8266 system-on-a-chip. No specific indicators or metrics were described. No experimental validation was reported, and no results were reported.

Kumar et al. [91] applied deep learning for identifying and segregating different types of waste, using YOLOv3 ANN within the darknet-53 framework. Training and validation experiments considered the classification of waste into six classes: organic, glass, paper, metal, plastic, and cardboard. The proposed deep learning model computed accurate results, with an average precision over 85% for all classes, and over 90% for all classes except paper. Results significantly improved over YOLOv3-tiny, which failed to correctly detect waste in six out of nine test images. In turn, the best Intersection over Union value of the YOLOv3 model was 67%, suggesting that there is room to improve regarding this metric. The proposed approach contributed to recycling and disposal in the context of smart cities. However, the experiments were performed over an ad hoc own dataset and no comparisons against state-of-the-art classification methods were reported.

Alqahtani et al. [92] proposed a system based on IoT for human activity detection and waste management classification. Feature extraction was accomplished by scale-invariant transformation. A recurrent ANN was applied for waste identification, to learn long-term dependencies between features. Cuckoo search was applied for optimizing the weights. The evaluation was performed over a synthetic dataset from the TrashNet images database, including six classes: plastic, glass, cardboard, paper, trash, and metal. The proposed system achieved high accuracy results, demonstrating improvement over other computational intelligence methods for classification. However, no state-of-the-art waste classifiers were included in the comparison. No real validation or implementation was reported.

Gondal et al. [93] proposed a real-time system for waste classification combining two machine learning models. A multilayer perceptron was applied for binary classification and a CNN was applied for classification. The system was conceived to operate in real time in an assembly line to categorize waste. Images were taken by an integrated camera and sensors were used to control and monitor the workflow. Accurate results were reported for an ad hoc set of images, but the comparative analysis only included one previous related work. From the description, it is not clear why a multilayer perceptron is needed for binary classification. No real implementation or case study was presented.

Adeleke et al. [94] proposed an ANN to predict the winter and summer variability of organic, paper, plastics, and textile waste. The evaluation was performed over a synthetic dataset of waste characterization from the summer of 2015 to the winter of 2016 in Johannesburg, South Africa, comprising organic, paper, plastic, and textile waste streams. Accurate results were reported for ten training algorithms, four activate functions, and different ANN architectures. The Levenberg–Marquardt algorithm, scaled conjugate gradient backpropagation, and resilient backpropagation computed the best results.

Al Duhayyim et al. [95] proposed a system applying deep learning for waste detection and classification. The proposed approach operated in two phases: a DenseNet-based Mask Regional CNN was applied for detection, and a deep Q-learning ANN using reinforcement was applied for classification. The setting of hyperparameters was performed using a dragonfly optimization algorithm. The system was evaluated via simulations over a benchmark classification dataset from the Kaggle repository (classes cardboard, glass, plastic, paper, metal, and generic trash). Accuracy results showed that the proposed system improved over other ANN-based methods. No real evaluation was presented.

Mohammed et al. [96] proposed sorting and classifying waste based on ANN and features fusion techniques. Feature extraction was performed using fusion scores. No details were provided about the applied ANN. Validation experiments were performed over a manually acquired dataset from Google Images and Flickr, spanning six waste categories. A comparison with other methods was reported, but no state-of-the-art deep learning model was considered. No real implementation was described.

Zhang et al. [97] designed an hybrid deep learning model for waste classification by combining data reprocessing, feature extraction using the AlexNet CNN architecture, waste generation forecasting using Deep Belief Network, and automatic hyperparameter optimization using the Optuna framework. The model was evaluated over a dataset of images from ImageNet, available in Kaggle, for classification in only two categories (recyclable and others). The evaluation confirmed that the proposed approach computed competitive results in comparison with five other state-of-the-art learning methods. No real evaluation of implementation was proposed.

As a simple application of automated problems in smart cities, Malik et al. [98] presented a CNN for waste classification. Multiple categories of bio-degradable and non-biodegradable solid waste were considered and a transfer learning approach was applied, using EfficientNet-B0 [99], trained over ImageNet from Google. Evaluated over a dataset of images from Kaggle, the transfer learning approach using EfficientNet-B0 achieved high accuracy for classifying images of domestic items, taken using a low-resolution camera. The article offered a simple example of the deployment of intelligent waste classification using limited resources, but no formal analysis or real application was described.

Udayakumar et al. [100] proposed a deep learning approach for waste classification. The classifier applied Single Shot Detection (SSD) for object detection and a MobileNetV2 CNN for feature extraction. A hybrid metaheuristic combining GA and PSO was applied for hyperparameter tuning. Finally, a multi-class SVM was applied for categorization. The system was evaluated over the TrashNet dataset, considering six categories. Accurate results were reported in the comparison with other deep learning models for waste classification. No real validation or implementation was reported.

Vijayalakshmi et al. [101] combined Elitist Barnacles Mating Optimizer and deep learning for classification in IoT-based WMS. The feature extraction process used the MobileNetv2 CNN and the metaheuristic was applied for hyperparameter tunning. Waste classification was performed using CNN and LSTM. The evaluation was performed over the standard benchmark dataset from Kaggle, considering six categories (cardboard, glass, metal, paper, plastic, and thrash). The proposed method computed high accuracy values in classification. No real implementation or validation was presented. The article offered a very limited description about the IoT infrastructure and mechanisms used for data acquisition.

Rajalakshmi et al. [102] developed a smart waste detection and classification approach. The detection module was based in RetinaNet and the Adagrad optimizer for hyperparameters setting. The classifier applied a stacked autoencoder, combining supervised fine-tune learning and unsupervised learning for training. Earthworm optimization was used for parameter tuning. A standard benchmark dataset from Kaggle was used in the evaluation, to categorize samples into six categories (cardboard, glass, metal, paper, plastic, and trash). The approach was compared with other techniques, including an intelligent deep reinforcement learning model and a CNN. An overall accuracy of 99% was reported for the experiments performed. No real implementation or case study were reported.

Liu et al. [103] addressed the issue of illegal waste disposal in smart cities, based on an automatic visual identification and warning system. Cognitive computing technologies were applied to analyze videos captured by cameras installed in urban areas, enabling the system to identify and alert the municipality to the presence of large-scale waste in inappropriate locations. The MultiBox SSD method was employed for detecting bulky waste and a CNN was applied for classification. An alert was issued to activate prompt removal of the dumped waste. The approach was implemented using the SSD MobileNet v2 architecture with a featured pyramid network, and validated through simulations. It outperformed other rather simple ANNs (R-CNN, WasteNet, AlexNet), obtaining high accuracy, precision, and recall. However, the validation was performed with a minimal dataset with 13 waste types. No real implementation of the approach was presented.

Alsubaei et al. [104] proposed a deep learning method for detecting and classifying small objects for waste management, using RefineDet (VGG16 backbone) for object detection and Functional Link ANN for classification. The system was evaluated over six classes of waste (glass, cardboard, paper, plastic, metal, and generic), employing a benchmark waste dataset from Kaggle. Accurate results were reported in the comparison with other ANN-based methods (AlexNet, VGG16, ResNet50, and multi-layer hybrid CNN). However, some of the baseline methods in the comparison obtained poor results, suggesting that they were not properly optimized for a fair comparison. No real evaluation was reported.

Cheema et al. [105] proposed a real-time WMS for waste classification using IoT, deep learning, and edge/cloud computing. An edge node performed image capture, object detection, and grid segmentation, all tasks needed for classification. Classification was performed using the VGG16 CNN architecture. A synthetic validation was performed over the TrashNet dataset for classification in six classes. Several metrics were reported, including classification accuracy, latency overhead, and resource/energy utilization. A comparison was reported with other two ANN-based classifiers (MobileNetV2 and Fast R-CNN). No real validation or implementation was reported.

Sankar et al. [106] proposed a waste classification system using IoT and deep learning. A CNN-based approach was applied for waste classification at collection sites, using a Rotation-Invariant Attention Network to isolate spectral bands and reduce redundancy. The system was evaluated on a standard dataset from the Kaggle repository with six categories. Several classification metrics were reported, but the comparison did not include state-of-the-art methods. No real validation or implementation was described.

Hamza et al. [107] presented a hybrid approach, combining Deep Consensus Network, Whale Optimization, and Naive Bayes, for object detection and classifying waste in smart cities into recyclable and non-recyclable categories. The approach applied the RetinaNet architecture and a Feature Pyramid Network for feature extraction. An additional upsampling layer was included to increase image resolution. A Centroid Proposal Network was applied to forward the resulting feature map at all scales, and a Region Proposal Network was used for predicting the bounding boxes of objects. The Whale Optimization algorithm was applied for tuning hyperparameters. The system was evaluated over the benchmark waste dataset from Kaggle repository, with six classes (glass, cardboard, paper, plastic, metal, and generic). Accurate result were reported, but the comparison with existing methods only considered simple approaches. No real evaluation was reported.

5.2.3. Characterization and Forecasting of Waste Generation

Existing studies focusing on WMS have considered various factors and variables for the characterization and forecasting of waste generation. Among the most frequently used variables, population size and socioeconomic status, local job market dynamics, and presence of businesses within the studied region or location have been the most relevant factors. In the context of general community scale (neighborhood, municipality, city), several attributes have been commonly chosen, including population size, income levels, expenditure patterns, demographic characteristics, types of waste collection practices, and the implementation of waste management programs. These variables were frequently selected due to their ease of availability and strong correlation with waste generation.

Abbasi et al. [108] proposed a model for forecasting monthly and seasonal waste generation based on a Radial Basis Function (RBF) ANN. A specific city-scale case study was analyzed with data from Teheran, Iran, covering the period 1991–2013, to predict monthly and seasonal waste generation. Statistical analysis was employed to investigate the impact of a combination of meteorological, socioeconomic, and demographic factors on waste generation. Furthermore, the gender of educated individuals was also taken into consideration as a significant feature in the analysis. Nine attributes were considered, including population, household size, number of educated women and educated men (demographic indicators), GDP, unemployment rate, and income (socioeconomic indicators), and maximum temperature and rain (meteorological indicators). The applied statistical tools included linear and nonlinear correlation tests. Distance correlation was applied to evaluate the nonlinear dependence between the input variables and waste generation, and Pearson correlation was used to evaluate the linear dependence between the input variables with waste generation. The feature correlation analysis indicated a stronger correlation on a seasonal scale compared to a monthly scale. This result is attributed to the fact that seasonal patterns play a crucial role in estimating the volume of waste generation within a city. The presence of educated women was found to have a strong association with waste, whereas the number of educated men was not identified as a significant factor in this regard. In a long-term analysis, more robust associations were observed between socioeconomic, demographic, and meteorological variables and waste generation, compared to a short-term analysis. However, certain factors such as the unemployment rate did not exhibit a discernible trend. Notably, variables such as GDP, income, household size, and the presence of educated women displayed significant correlations with both monthly and seasonal waste generation, with their respective strengths varying based on the time scale considered. Regarding waste generation forecasting, the RBF model was evaluated using an eight-fold cross validation technique, outperforming an ad hoc adaptive neuro-fuzzy inference system (ANFIS) and ANN models. However, the values of the R² metric were low (0.678 for RBF, 0.53 for ANFIS, and 0.43 for ANN) for monthly prediction. Better results were obtained for seasonal prediction (0.85, 0.62, and 0.56, respectively).

Soni et al. [109] studied several models for waste generation, including ANN, ANFIS, a discrete wavelet theory ANN, a discrete wavelet theory–adaptive neuro-fuzzy inference system, a GA–ANN, and a GA–adaptive neuro-fuzzy inference system. The approaches were evaluated for a case study in New Delhi, India, using time-series data of yearly waste generated in the city from a previous study. The hybrid model combining GA and ANN computed the best results (i.e., lowest RMSE and highest IA and R² values.

Wu et al. [110] studied the scalability of models based on ANN for forecasting waste generation. The main goal of the study was to determine if the successful accuracy values obtained in specific cities could be replicated for larger (regional or country) scales, in large countries. ANN models were developed for mainland China, aimed at large-scale waste generation forecasting for many cities, grouped in regions. Seven relevant socio-economic attributes were considered as input variables in the predictive model: population and living standards (permanent resident population and per capita disposable income), production and economics (gross domestic product and shares of primary, secondary and tertiary industries), and the condition of public infrastructure for waste disposal and management (general public budget expenditure). The model was evaluated for 25 cities in China, geographically grouped into three regions, using data from the National Bureau of Statistics. A data screening procedure was performed to ensure the quality of the data, applying interquartile range filtering and Min-Max normalization techniques. Additionally, a correlation analysis was conducted to assess the influence of each attribute on waste generation and to eliminate attributes that exhibited high correlation with each other or showed no significant relationship with waste generation. The Pearson correlation coefficient was applied for the statistical analysis. The dependence of the model on each predictor was studied to analyze the regional differences, and comparative analyses were performed. Results showed the differences between the studied regions in China. The accuracy of the proposed ANN forecasting models was high, and it improved significantly when taking into account the regional differences and the influence of different attributes. However, no comparison with other forecasting models was presented. The evaluation helped to understand the varying impacts of different predictors on waste prediction across different regions. The developed large-scale models are useful for estimating waste generation in cities, regions, and countries where few historical data are available.

Sengupta and Chinnasamy [111] employed predictive analytics to process cloud data collected from radio wave sensors deployed on smart waste bins. An active RFID tag and an IoT-enabled receiver was used for efficient data transmission. K-means clustering was applied for the identification of patterns and trends in WMS. The information was used to plan the routes of collection vehicles. The evaluation was performed on a synthetic scenario using randomly generated data, and few relevant metrics were reported. No real evaluation or deployment was reported.

Tsai et al. [112] developed a hierarchical framework for waste management, applying exploratory factor analysis to examine the potential direct or indirect connections between attributes and enabling a comprehensive exploration of their interrelationships. Then, a combination of fuzzy decision-making trial and evaluation laboratory and technical integration using linguistic references was employed for identifying suitable strategies, considering opinions from 36 academic and institutional experts with more than 8 years of experience in the field. The approach took into account the interdependencies among different attributes and aimed to determine the most appropriate course of action. A total of 14 attributes from the related literature were chosen for the analysis: community concerns, legislation and policies, involvement of stakeholders, institutional and organizational administration framework, occupational safety and health, efficiency of resources, pollution standardization, environmental mitigation, financial and operational requirements, city parameters, economic benefits, facilities and infrastructure requirements, innovative capacity of waste treatment, and technology functionality and appropriateness. These attributes encompassed four key aspects, namely, sociality, environment, economy and technique, to assess a causal waste management model specifically tailored for Vietnam. Studies were developed in Hanoi, Danang, and Ho Chi Minh, three cities that lack adequate and effective strategies to achieve the main objectives of WMS in terms of public health, environmental enhancement, and societal contentment. The findings revealed that technical integration and social acceptability play pivotal roles in waste management. Treatment innovations, safety and health considerations, economic benefits, and the functionality and appropriateness of technology were identified as the key criteria influencing these aspects. Moreover, notable distinctions between the three studied cities were identified in terms of their waste management approaches and outcomes.

Cubillos [113] applied a multi-site Long Short-Term Memory (LSTM) ANN for forecasting weekly household waste generation. The task posed significant challenges, particularly when dealing with household-level data characterized by rapid short-term fluctuations and highly nonlinear dynamics. The forecasting approach was applied in Herning, Denmark, from 2011 to 2018. The main focus of the study was exploring the correlations and the impact of climate variables. The considered input data included eight predictor variables: air temperature (minimum, average, and maximum), wind (low, middle, and high), accumulated precipitation, humidity, atmospheric pressure, and sun time. All data was obtained from the Danish Meteorological Institute. A standard approach using a sliding window technique was applied to use past observations as input variables for the LSTM training. The window size used as historical information for the LSTM was eight weeks. The model was compared against a traditional Autoregressive Integrated Moving Average Model (ARIMA) and a simple multilayer perceptron ANN. The key findings of the study revealed that the applied LSTM approach was able to outperform the accuracy of the ARIMA model by an impressive margin of 85% and the accuracy of the multilayer perceptron by 78%. Moreover, the results demonstrated that adopting a multi-site approach, as opposed to developing individual models for each household, enhanced the forecasting accuracy of the LSTM model by an average of 28%.

Nguyen et al. [114] applied a data-driven approach and computational intelligence methods for forecasting the generation rate of solid waste in residential areas. The approach was validated for a case study in Vietnam, comprising selected residential areas. The considered input dataset comprised eight variables covering different socio-economic aspects, demography, consumption habits, and waste generation characteristics specific to the studied area. All data were obtained from the Vietnam National Bureau of Statistics. The data analysis stage applied correlation analysis to the independent variables to avoid overfitting, and feature ranking, applying a generalized linear model and a multiple linear model. The Breusch–Pagan test was applied to analyze heteroscedasticity and its influence on error prediction. The analysis was performed on two datasets: a comprehensive dataset and a condensed dataset. The comprehensive dataset comprised ten variables: province details, urban population, total retail sales of consumer goods, average monthly income per person, average size of home per person, population density, average monthly consumption expenditure per capita, total hospital beds, total residential land per province, and total solid waste collected per day. In contrast, the condensed dataset included only total solid waste collected per day and urban population. The comprehensive dataset encompassed 63 entries, representing the provincial administrative units of Vietnam in the year 2016, while the condensed dataset contained 189 entries, using historical data from 2015 to 2017. Through statistical models and simulations, urban population, average monthly consumption expenditure, and total retail sales were found to exert the most significant influence on waste generation. Regarding the machine learning models studied, RF and k-nearest neighbor (KNN) showed strong predictive capabilities when trained on the training data (80% of the dataset), yielding an $R^2$ value exceeding 0.96 and a mean absolute error ranging from 121.5 to 125.0 for the testing data (20% of the dataset). The developed models offered dependable forecasting of waste generation data, thereby facilitating the planning, design, and implementation of an integrated solid waste management action plan for Vietnam. Some acknowledged limitations of the study include the heterogeneity of the dataset, particularly the absence of data from lower administrative units in the country.

Namoun et al. [115] proposed an ensemble learning approach for predicting domestic waste generated in urban environments. The first stage applied the Optuna framework for machine learning to determine suitable values for a large number of hyperparameters of the seven machine learning methods studied: SVR, XGboost, LightGBM, RF, ANN, KNN, and Extreme Trees. In the second stage, the applied methods were combined in a mix linear meta regressor trained using the results from the previous models. The proposed approach was validated on a case study considering real data of average weekly household waste generation in 2011–2021 in the southeastern region of the United Kingdom. The meta regressor outperformed state-of-the-art methods and the average ensemble results of the models studied in the first stage. The proposed approach was reported as useful for enhancing the predictive quality of existing models, even for datasets with few features. The improvement was consistent across several precision metrics studied.

Lakhouit et al. [116] applied machine learning to predict domestic waste generation. Linear regression, regression trees, Gaussian process regression, support vector machine, and autoregressive integrated moving average methods were compared for time-series analysis over two case studies: (i) global data from Saudi Arabia (2010 to 2021) and Bahrain (1997 to 2021) authorities, and (ii) a single large family (eleven people) for one month. Accurate results were reported for the studied forecasting methods. Applications regarding the economical, social, and environmental benefits of recycling were discussed.

Chen [117] proposed an automatic waste recycling framework for separating complex waste materials in a mixed recycling scenario. A set of sensors were described (temperature, weight, metal sensors), but no description of their integration into a smart bin was presented. Logistic regression was applied to forecast the state of each waste bin. The monitoring procedure was not clearly presented. A module in charge of computing the fastest route from collection vehicles to waste bins was also described. The model was validated using simulations and undisclosed data. No real validation or implementation was presented.

Mohammed et al. [96] presented a smart bin using IoT and machine learning for waste segregation into biodegradable and nonbiodegradable categories. The smart bin included a camera to capture images of the disposed waste, load and ultrasonic sensors to monitor the filling level, a Raspberry Pi to execute the classification algorithm, a console, two gates, and a servomotor for rotating a collecting plate to drop the waste into the corresponding chamber. A CNN SqueezeNet model was used for waste classification. The validation was performed on synthetic images from well-known waste repositories, considering seven categories (cardboard, e-waste, glass, metal, paper, plastic, and trash). A Generative Adversarial Network was applied for data augmentation. The system was able to identify and separate waste, operate in real time, and high-quality results were reported in the comparison with other pre-trained models for waste separation. A prototype was shown for the smart bin, but no real implementation or validation in real scenarios were reported.

AlDuhayyim [118] proposed an approach for waste categorization applying computational intelligence. The approach integrated a single-shot model for object detection using the VGG16 CNN architecture, a MixNet CNN for feature extraction, and SVM for classification. The hyperparameters of the feature extraction model were set applying the modified cuttlefish swarm optimization method. The approach was validated via simulations over the TrashNet standard dataset, considering six categories (cardboard, glass, metal, paper, plastic, and trash). Accurate results were reported in the comparison with state-of-the-art methods. No real validation or implementation were described.

Oğuz and Ertuğrul [119] proposed a deep learning approach to determine the fullness status of waste bins using information from camera images. The main goal of the problem was to classify waste bins into full (dirty) and non-full (clean). Four deep ANN models (DenseNet-169, EfficientNet-B3, MobileNetV3-Large, and VGG19-Bn) were evaluated in a comparative analysis performed over the standard CDCM dataset from the Kaggle repository. Accurate results were reported, but no comparison with state-of-the-art methods was performed, just a mention to the results of the creators of the CDCM dataset. No real case study or development/integration of the proposed system was described.

Lingaraju et al. [120] proposed a smart WMS using sensors and IoT for waste segregation, GPS-based location tracking, and real-time monitoring of weight, filling level, and air quality around the bins. An ultrasonic sensor was used to check the filling level of smart bins and issue alerts to a open-source microcontroller unit build using the ESP8266 system-on-a-chip. An inductive proximity sensor was used to detect and segregate metallic waste, and a raindrop sensor was used to detect and segregate wet waste. An Infrared Radiation sensor was used to detect when a user was near the bin, and a camera captured an image of the person throwing the waste, to be used as evidence in case of any illegal activities. The GPS module was used to track the location of bins, so that the waste was disposed of quickly when the bin reached maximum capacity. An MQ135 air quality sensor was used to monitor the air quality near the waste bins to protect the health of citizens. A load cell sensor measured the weight of the waste in each bin, to determine the ideal time for waste disposal. A servo motor rotated a circular disk containing several bins. The control software was developed using Blynk IoT. The ThingSpeak IoT platform service was used to analyze air quality and weight live data streams. Machine learning was applied for classification of air quality levels and waste weights. Notifications were sent when bins are full or near full, along with their exact location, to facilitate quick disposal. Despite being a comprehensive and integrated solution for efficient waste management, the proposed system lacked proper evaluation. Only screenshots of reports and graphics from the ThingSpeak IoT platform were included in the article. No real evaluation or implementation was reported.

5.2.4. Smart Bins

Fataniya et al. [121] presented a prototype of a smart bin including separate shelves for wet and dry waste, a moisture sensor, servo motors, an ultrasonic sensor, a gas sensor, and a GPS module. The moisture sensor was applied for waste separation (dry or waste), by sensing the volumetric content of water in the waste. The ultrasonic sensor was used to monitor the filling level. All collected data were sent to the Adafruit cloud platform at regular intervals, and notifications were sent to administrators and collection vehicles. The system was evaluated in a case study in Ahmedabad, in western India. Cost analysis and traveled distances by collecting vehicles were reported on a simple graph with just six nodes, five of them set as pickup points. No realistic evaluation was reported.

Kang et al. [122] proposed a smart bin for electronic waste collection to be used on households. The smart bin included sensors to determine the filling level and communicated to a backend server integrated with a notification system to notify collecting vehicles. The study provided a basic validation for the proof-of-concept of the smart e-waste collection system and discussed potential implementation on a larger scale in Sunway City, Malaysia. However, no actual implementation or validation was presented.

Vishnu et al. [123] presented a system for solid WMS using IoT and two types of waste bins, for public and domestic use. The monitoring of public waste bins is performed via a LoRaWAN networking architecture, whereas for domestic waste bins, Wi-Fi-based communication is used. The filling level is monitored by an ultrasonic sensor. A GPS module is used for geolocation of smart bins. Other system components include a LoRa/WiFi module, an embedded antenna, and a host microcontroller. Limited validation experiments were performed in a laboratory, studying the functionalities, energy consumption, and life expectancy. No real validation or implementation was reported.

Anagnostopoulos et al. [124] proposed a WMS involving dynamically allocating collection and transfer points, followed by the transportation of waste to processing facilities. Several models were presented, to support a shift from the conventional method of dumping waste into larger containers to a system that involves replacing full waste bins with empty ones. Another relevant concept in the proposal was mobile depots, used as intermediate collection and transfer points. The system was evaluated via quantitative (collection and transportation time, load, distance, fuel) and qualitative metrics (satisfaction) over three case studies with data from St. Petersburg, Russia. The WMS workflow was also studied. Results showed different efficiency ans satisfaction levels for the studied cases.

Cruz et al. [125] reported the experience of Lisbon City on applying low power WAN technology for waste management, instead of existing GSM communications. LoRa and LoRaWAN technologies were evaluated for several case studies defined in the city. Other relevant features studied included the types of recycling waste bins, their placement (surface or underground), and sensor types. Underground waste containers were more challenging, due to the container itself causing a significant attenuation on the LoRa link.

Sallang et al. [126] proposed a smart waste bin enhanced with software to detect and categorize waste, to be stored into different compartments. The proposed system combined deep learning and IoT to improve the collection efficiency and reduce costs. The SSD MobileNet Quantized ANN was applied for waste detection and classification into five categories (cardboard, paper, metal, glass, and plastic). The model was implemented in TensorFlow Lite over a Raspberry Pi 4 single-board computer. The IoT module used the camera for waste detection and a servo motor to segregate waste into different compartments. Other components included an ultrasonic sensor for monitoring the remaining empty capacity of the bin, a GPS module for geolocation, and a LoRa module for communication, protected with an RFID locker. The system was evaluated on a bin prototype, showing that the waste classification process was accurate and efficient, the ultrasonic sensors properly monitored the fill of the bin, the location was correctly determined, and all information was sent to a receiver at up to 5 km using LoRa. However, few data were considered in the learning process for classification and the object detection module required a GPU. The lifetime of batteries to power the system was another point of concern.

Ashwin et al. [127] proposed a smart bin for automatic waste disposal, including a solar panel, an ultrasonic sensor for human detection, a servo motor for opening and closing the lid automatically, and a separator for dry and wet waste. All components were controlled using an Arduino Uno microcontroller that issued an alert message when the smart bin was near full. A selection algorithm was presented for routing data packets with information for a group of installed bins. The system was evaluated via simulations over a configuration of smart bins placed in random locations. The analysis was simplistic and straightforward results were computed. No real deployment or evaluation was presented.

John et al. [128] proposed a smart bin using IoT for monitoring and waste disposal prediction. Several sensors were used, including infrared, ultraviolet, and weight sensors, and a GPS module. The system was controlled by an Arduino microcontroller and Wi-Fi was used for data transmission. A LSTM ANN was used to learn patterns of waste generation/disposal. Hardware algorithms were included for automatically opening and closing the lid when a person stands nearby, and for sealing the bin when it is full. Firebase Cloud messages were issued to notify the filling level of smart bins to operators. The validation of the system was performed using 160 data samples, collected for three days. Four LSTM variants were applied for predicting the waste level in the next 15 min, using the information from the current level. A stacked LSTM computed the best forecasting results, but the evaluation considered only a small number of cases and data points. No comparison with other methods was reported. No real implementation was reported.

Baldo et al. [129] introduced a smart waste management system considering several LoRaWAN classes: (i) smart bins to collect waste disposal data, (ii) smart drop-off containers with a RFID reader and a GPS module, and (iii) video surveillance units applying machine learning for fire detection in nearby bins or drop-off containers, applying image recognition using YOLO. The proposed system was implemented applying the edge computing paradigm. The system was evaluated in a laboratory and in a real city location. The LoRaWAN approach implemented over edge computing and using ANN for image recognition reduced the costs and the required management infrastructure. The modules of the system were able to properly locate positions, evaluate the filling level of bins and containers, check the temperature and detect vandalism using fire, and perform video surveillance. Limitations of the system were related to slight inaccuracies and fluctuations on filling measurements and the need of a specialized computer for machine learning and image processing, since low-cost computers proved incapable of producing accurate results on those tasks.

Sidhu et al. [130] proposed a simple design of smart bins for domestic use, incorporating a weight scale and a web application used for forecasting waste generation. A collection route was computed considering the forecasted data, the location of the collector, and the transportation mean used for collection. The related VRP with time windows was solved applying constrained programming using the ILOG solver. The proposed system was evaluated on a synthetic (simulated) scenario, built using 176 experiments considering different user behaviors to collect data from Quezon, Philippines. Results showed that more than 80% of the waste bins were collected in the simulations performed. However, only a very small problem scenario was considered, accounting for only eight waste bins and two collectors, and no comparison with other planning algorithms was presented.

Jim et al. [131] proposed a smart WMS based on smart bins accommodating up to four rotating containers, positioned 90° apart from each other, and installed below ground level. Only one container appears above ground at a time, making improved use of the available space on the street. After filling, containers rotate, and once all the containers are full, the bin sends a message to a central server that communicates with an autonomous car to collect the waste, following a predefined route. All communications are based in the IoT paradigm. A very simple evaluation was developed for a small prototype smart bin with four containers (dimension: 15 × 15 × 20 cm³). No real design or validation was reported.

Majidi [132] proposed a smart bin for waste collection and separation, composed of a smart landfill structure to be installed underground, with waste collection tanks in various materials, designs, and shapes. An intelligent control chamber was designed to oversee all waste mechanisms using a remote system that enables communication between the system and human operators for issuing commands. A mobile frame was incorporated to transport waste bins to the ground, located on a lift. A weighing system was also included to gather and transmit filling information to the control chamber, alerting the human operator when the tanks need to be emptied. The upper part was located on the ground in such a way that the desired dry waste was received in each tank. An intelligent air conditioning system was included to automatically activate when the percentage of waste gases exceeds the standard level. Finally, the system was equipped with a fire extinguishing system to handle any potential fires. No real evaluation of the proposed system was reported.

Bourougaa et al. [133] proposed a low-budget smart bin consisting of six separate cans to store different types of waste. Other physical components included an Arduino UNO microcontroller, a bluetooth module, a micro servo motor, and temperature, humidity, and ultrasonic sensors. Waste was classified using a a CNN, applying three models: VGG16, Dense201, and Resnet50, executed on a Raspberry Pi. The evaluation was performed over the standard TrashNet dataset, considering six waste types (glass, paper, cardboard, plastic, metal, and generic). DenseNet201 computed the best accuracy results, and a comparison with other classifiers from the literature was reported. A mobile application was also proposed to manage the smart bin. No real evaluation was reported.

An et al. [134] conducted a case study highlighting a practical demonstration of an intelligent WMS that uses smart bins connected through IoT, along with an integrated dashboard system for decision-makers, for Wyndham City, Australia. The research focused on determining whether there existed a correlation between the filling levels of the smart bins and specific dates or months. The proposed smart bin used solar-powered compacting technology and was connected to a software platform to store historical data. Notifications were used for filling or near filling levels. Few results were reported and the impact of the proposed system was not clearly evidenced. No real validation was reported.

Ahmed et al. [135] proposed a smart WMS based on IoT, clustering smart bins to reduce energy consumption to extend the operational life, dealing with missing data from smart bins, and defining waste collection routes to reduce time and maximize fuel efficiency. Clustering was performed using LEACH [136] and an artificial humming algorithm. k-nearest neighbors was applied for data imputation and multiobjective artificial humming was applied for routing. The system was evaluated on undisclosed synthetic scenarios. The scope of the results was limited. No real evaluation or implementation was reported.

Thamarai et al. [137] proposed a self-powered WMS for smart cities integrating a separation module, a power generator unit, a smart bin using IoT sensors, and an electronic control unit. A CNN was applied for waste separation into organic and inorganic. Organic waste was used for power generation and residues were recycled as fertilizer. The power generator unit produced biogas by combustion, for electric power generation. Components were monitored and controlled by a control unit on a Raspberry Pi that sent alerts to collection vehicles when a smart bin is full. The system was evaluated over a synthetic dataset from Kaggle with eight categories and compared with other ANNs (VGG16, InceptionV3, ResNet50, and Inception ResNet). No real implementation was presented.

5.2.5. IoT Systems

Zeb et al. [138] proposed an IoT-based communication architecture for managing smart waste bins. A specific technique was applied for minimizing end-to-end delay for the deployed wireless sensor network, and for providing on-time waste collection to avoid overfilling and collecting non-filled bins. The proposed communication architecture was evaluated through simulations and results showed that it improved over other communication approaches. No real evaluation or prototype development was reported.

Srikanth et al. [139] proposed an IoT-based system to issue notifications about filling levels. Several IoT devices were used, including smoke and flame sensors, strain gauge drivers, a Bluetooth module, a MCU node for information exchange, all coordinated by an Arduino Uno microcontroller. No validation or results were reported.

Sharma et al. [140] presented a solution to address the issues of waste overflow and ineffective waste disposal. A smart bin using IoT was proposed to separate waste into metallic and non-metallic categories. An ultrasonic sensor was used to measure the filling level. Other IoT components included a GPS module, a RFID logging system, temperature and humidity sensors, and a LinkIt ONE kit with an antenna for data communications (Wi-Fi/Bluetooth/GSM (2G)/GPS/GLONASS). Although the system was described as aimed to optimize waste collection routes, no routing experiments were reported. No results were reported for separation. No real implementation was presented.

Bano et al. [141] proposed a framework for real-time monitoring of smart bins applying IoT and AI. A standard system design was proposed, including smart trash bins, a collecting vehicle, and a centralized database for storage. Real-time information was used to avoid overloading of the bins, as smart bins sent a request for collection after the waste level reached a predefined threshold. Fuzzy logic was applied for selecting the locations to install bins. The proposed framework was implemented in a multiagent modeling environment and evaluated in an undisclosed real environment, but few relevant results were reported.

Catarinucci et al. [142] presented a WMS using sensors and RFID technology to provide services, including citizen recognition and registration, waste weighting, and timestamping and location mapping of disposal and collection. The system included smart bins equipped with an RFID tag with weight-sensing capabilities and an antenna, coordinated via cloud software with several features: storing and processing data from wearable RFID readers and a web app, along with integrated business logic to handle data streams (requests from and responses to third-party APIs). The web app provided users with relevant information. The evaluation was performed on an experimental setup to study the performance of the RFID components and a functional proof-of-concept validation via simulations in laboratory for a week. All components showed correct functionality. No evaluation on real scenarios was reported.

Pardini et al. [143] proposed an inclusive WMS using IoT, taking into account citizens in the process. Sensors were used to monitor the filling levels of smart bins. Real-time data processing was used to optimize collection routes and statistical analysis. The generated information was available to citizens via web or mobile applications. A prototype of the system was evaluated in a real-scale experiment in a undisclosed Brazilian city. No results were reported. Overall, the system showed reasonable potential to bring about a transformation in how citizens approach waste disposal. Additionally, the system demonstrated its ability to optimize economic and material resources, further enhancing its value.

Aktay and Yalçın [144] proposed a WMS using IoT focused on sustainability. Several sensors were used in the proposal, including a long-range low-power transceiver, an open-source single channel LoRa gateway, an Arduino Uno microcontroller, an ultrasonic distance measurement module, an adapter for peripherals, and other components (antenna, battery, etc.). The system issued an alert when the bin was full, and the information was dynamically used to re-schedule the collection service. Validation experiments were reported in laboratory and real scenarios in Istanbul, Turkey. No real deployment was reported.

Jhaveri et al. [145] proposed a smart WMS using low-cost components, including a micro controller, ultrasonic sensors, and a GSM module for wireless communication, to be integrated with waste bins. Two software components were also included: a cloud computing back-end storage system for saving data and notifications, and a web application for operation by authorities. The database included a flow prediction module using linear regression and a Naive Bayes classifier to determine the probability of a given bin to be full on the next day. The web application included several tools for data analysis, visualization, and report generation. Statistics and reports were showcased for a case study in Bharuch, western India. The system was presented as a contribution towards sustainable development, as an innovative, energy-efficient and easy-to-deploy solution.

Senthilkumar [146] assessed an existing waste management scenario for a semi urban location in Annamalai Nagar Panchayat (a settlement in transition from rural to urban) in the Cuddalore District of Tamil Nadu, India. The existing waste management practices were evaluated, highlighting their drawbacks. Furthermore, the study proposed an IoT-based WMS and compared the potential benefits of implementing this technology-enhanced approach with the current waste management practices in real-time scenarios. No results were reported, other than the cost of the proposed WMS.

Marques et al. [147] presented a case study of waste management as an example of application of a multilevel architecture for management in smart cities, applying IoT. The architecture combined Cyber Physical Systems, embedded intelligence, and communications, identifying four layers for physical objects (IoT sensors), communications (using RFID, Bluetooth, or Zigbee protocols), a cloud platform for data storage and processing, and services for developing smart city applications. The system was evaluated in an indoor environment with just two types of waste bins (for organic and recyclable waste) and in a public park with four types of waste bins (for organic waste, paper, glass, and plastic). The evaluation compared three communication protocols (CoAP, HTTPS, and MQTT) regarding energy consumption and communication-aware metrics (latency, jitter, and throughput). The scalability analysis showed low response times when handling up to 3902 bins.

Ramson et al. [148] proposed an IoT system for monitoring waste bins from a central station. Ultrasonic sensors were installed on bins for monitoring the fill level and information was transmitted to a wireless access point that concentrated information from several nearby bins. Data were uploaded to a central server, for operators and authorities to conduct analysis and visualization. Data communications were performed using Transport Layer Security and Secure Socket Layer protocols to guarantee security. The system was evaluated in a laboratory to determine the correct monitoring of the filling level, the life expectancy of the monitoring hardware, the maximum distance for communications between smart bins and the concentrator, and the whole cost. No evaluation on real locations was reported.

Zhang et al. [149] proposed a smart waste removal system integrating smart bins, a waste collection vehicle, a Route Operation System robot, and a planning algorithm based on Rapid Exploration of Random Trees. An ultrasonic sensor was used for waste overflow detection on smart bins, and signals were sent to collecting vehicles after reaching a threshold value. Several tasks were included in the software control of the collecting vehicle, including location, planning/decision making, obstacle detection and avoidance, path following, and motion control. All communications used LoRa low power WAN wireless IoT technology. The system was validated on a small case study in a closed park in Guangzhou, China, using the Gazebo simulator. Accurate results were reported: the proposed system was able to reduce the labor cost by 20% over a traditional WMS.

Yuvaraj et al. [150] proposed an integrated system for smart cities, including the management of electric waste, considering vehicles, traffic, and blockchain technology for information security. The information about bins was collected using IoT devices. Deep ANNs were used for traffic prediction, considering several features (waste payload, weather conditions, bin and vehicles location, etc). All communication data between smart bins and collecting vehicles were secured by applying blockchain, for protection against cyberattacks. The system was evaluated via simulations considering a small geographic scenario. Several metrics were studied regarding decision-making, electric vehicles routing, energy efficiency, and security. The scalability was also studied, but just by incrementing the number of vehicles and waste bins. No evaluation on real scenarios was reported.

Azyze et al. [151] proposed a communal WMS designed to provide administrators with real-time status updates on the filling level of smart bins, and other relevant indicators such as temperature, humidity, and air quality within the bin. IoT sensors were coordinated by an Arduino controller and transmitted via WiFi to a cloud server via the ThingSpeak application. A prototype was designed for the evaluation in the laboratory. Results on the accuracy of the ultrasonic sensor were reported for food, paper, plastic, and metal waste. No other relevant indicators were studied. No real evaluation or deployment was reported.

Joshi et al. [152] proposed an IoT architecture combining a Wireless Personal Area Network and cloud-assisted technologies for real-time monitoring of solid waste. The smart bins used IoT devices to monitor the filling level and communicated to an edge node using the Xbee protocol. Edge nodes had decision-making capabilities to reduce latency. A fog gateway served as a central hub, leveraging fog computing capabilities to efficiently transmit data to a cloud server. Data processing was performed in the cloud server, and the results were left available for queries and visualization. The system was only evaluated via simulation using the LabVIEW tool. No real evaluation or deployment was presented.

Wong et al. [153] described an IoT WMS integrating a Raspberry microcontroller for collection and waste classification, an overflow alarm via ultrasonic, and tracker sensors to trigger waste collection. A simple workflow was applied for computing the fill level of bins using infrared emission and distance measurement. Machine learning (RF and CNN) was applied for waste classification. A very simple evaluation of the overflow alarm was reported for filling levels of 7 and 3 cm below the top border. The classification methods were validated over an own dataset from the authors, to classify waste into one of four categories: recyclable, kitchen waste, hazardous waste, and other. CNN computed better accuracy results than RF. No other classification or learning metrics were evaluated. No real evaluation was presented, other than the prototype for studying the overflow alarm mechanism.

Chauhan and Gargrish [154] applied Artificial Intelligence of Things (AIoT) to design a system combining smart bins with GPS, collecting vans, and a central database. for real-time monitoring of bins. The design was rather simple, and although authors claimed that AI played an important role in selecting proper locations for bins, no details or algorithms for this goal were presented. The approach was generic, and no validation was presented. Later, Chauhan et al. [155] applied a standard CNN for object recognition and waste categorization to improve decision-making in WMS. The evaluation was performed over the standard waste classification dataset from Kaggle, augmented using random re-sized crop and horizontal flip. The proposed method showed improvements in accuracy over AlexNet, VGG16, and ResNet34. No real development or evaluation were reported.

Thaseen et al. [156] presented an hybrid combining GA and smart waste management. Using data collected from sensors, a Mamdani fuzzy inference system was employed to assess the likelihood that each smart bin was approaching full capacity. Waste segregation was also applied, via image classification. The GA was used for finding a near-optimal set of rules for the proposed fuzzy inference system. The GA-fuzzy method was implemented in a cost-effective, small hardware component using Arduino Board, applied for detecting the waste level within each bin, and waste classification. The system was evaluated on synthetic experiments over the Proteus simulator. Accurate results were reported for standard machine learning metrics. No real scenario was considered in the evaluation.

Brouwer et al. [157] compared ultrasonic sensors and visual observations by drivers of collection vehicles for monitoring waste bins. A Gaussian predictive model was applied for trade-off analysis between the number of collections and the number of overflows. The evaluation was performed by simulations, using real data on bin fill levels from a waste management company in Portugal. Significant improvements were achieved when applying the monitoring approaches, compared to the current situation. Visual observations proved valuable in complementing automatic approaches based on IoT technologies.

5.2.6. Computational Intelligence for Other Relevant Problems in Waste Management

Toutouh et al. [158] applied soft computing methods for locating waste accumulation points in smart cities. The aim was to reduce investment costs, increase the population served by installed bins, and improve system accessibility. Single- and multi-objective heuristics based on the PageRank method and two multi-objective evolutionary algorithms were proposed for the problem. The evaluation considered real scenarios in Montevideo, Uruguay and Bahia Blanca, Argentina. Accurate results were reported, with appropriate trade-offs between the problem objectives, improving over the current planning strategies in both cities. The approach did not include real-time information.

Mokale [159] studied the awareness of people and the role of local governments and non-governmental organizations in smart waste management. The study analyzed primary data from a survey conducted in Mumbai, India, and secondary data gathered from research articles, newspapers, and other sources. The differences between smart waste management in small towns and metropolitan cities were commented on, and specific recommendations for improving solid waste management were provided for the studied scenario. The obtained results were not properly organized, reported, and commented on in the article.

Ramalho et al. [160] presented a software architecture for waste collection based on multi-agent systems. The main requirements for a smart WMS were studied to define important features for the proposed architecture, including a user-friendly and comprehensive dashboard interface, community participation, IoT technologies, decision support, and continuous monitoring. Multi-agent simulation techniques were applied to model social relationships, and a strategy was devised to generate improved collection routes, taking into account the locations of waste bins and their filling levels, monitored through IoT smart meters. Simple methods were applied for bin location and shortest path calculations using the Floyd–Warshall algorithm. The proposed architecture was instantiated and integrated into the smart city agenda of Natal, Brazil. No comparison with other methods was reported.

Sharma et al. [161] developed a framework to analyze the barriers to the adoption of IoT solutions for waste management in smart cities of developing countries. A multi-attribute decision-making approach was applied to identify and analyze barriers for IoT adoption in WMS projects in India. The application of the Total Interpretative Structural Modeling approach enabled the construction of a structural framework to address the identified barriers. The evaluation was performed over data from a panel of nine experts in different smart city and WMS domains from Mussorie Dehradun Development Authority, Uttarakhand, India. Several bottlenecks were identified, including: operational costs and payback; absence of standardization, regulations, directions, and policy norms; limited technical knowledge among policymakers; challenges related to internet connectivity, privacy, and security; and issues regarding mobility, transparency, and the absence of ICT infrastructure. The possible extension to other developing countries was not explained.

Mishra et al. [162] also applied a hybrid multi-attribute decision-making approach to study barriers to the adoption of IoT solutions for waste management in India. A framework was presented for building an index applying the combined compromise solution methods and Fermatean fuzzy sets to model the uncertainty of decision-makers. The framework was evaluated on an undisclosed scenario in India, presented as representative of WMS on smart cities in “developing nations”. Sensitivity analysis and a comparison against previously proposed approaches were performed. Results indicated that the proposed method provided useful assistance for the decision-making process.

Torkayesh et al. [163] studied the application of smart technologies to design smart medical WMS. Several factors were analyzed and evaluated following a multi-criteria approach, focusing on issues that have prevented the application of IoT and blockchain technologies for medical waste. A case study was presented for Istanbul, Turkey. The main factors that hindered the acceptance of new technologies in that case were the difficulties surrounding training, and the lack of stakeholders, professionals, and technicians. These shortcomings outperformed the benefits of the considered technologies in terms of security and privacy.

Duela et al. [164] proposed a routing method for waste collection using IoT and fog computing. A simple shortest path algorithm was applied, supported by the Google Maps API from Fog Layer, considering four major parameters: the location of the collection vehicle, the distance between the collection vehicle and the bin, the filling level of the bin, and the capacity of the collection vehicle. An infrastructure was proposed for a smart bin, including an ultrasonic sensor, a weight sensor, and sensors to evaluate the concentration of methane and other harmful gases. A RFID tag was included to track the collection timestamp. LoRaWAN was used for communications. Logic, computations and storage were provided by a fog server, taking advantage of low latency, high bandwidth, and improved security capabilities. No experimental validation was reported.

Bazrbachi et al. [165] proposed a framework for studying the intention of households to engage in smart solid WMS. The theory of planned behavior was applied to assess the dimensions of latent variables of a model for evaluating the intentions of waste separation and recycling. Relevant determinants were studied, including attitude, subjective norm, and perceived behavioral control. The proposed approach was evaluated in a case study in Shah Alam, Malaysia, using questionnaires regarding awareness, opinions, attitude, and behavioral intention. The socioeconomic characteristics of households were also evaluated. The result of the evaluation indicated that the main drawback of the smart WMS were the high routing cost, since no optimization or use of real-time information was applied. This drawback resulted in an overall increase in the budgetary cost of managing solid waste. As a consequence, a rise in the financial expenses associated with solid waste management was observed.

Shahsavar et al. [166] proposed a MCDM framework for management and bio-recovery of plastic waste, using fungus for bio-degradation. The system applied Shannon entropy, ordered weighted aggregation, analytic hierarchy process, order preference by similarity to the ideal solution and elimination/choice translating reality. Relevant factors were considered in the analysis, taking into account economic criteria (investment cost, consumed materials, human resource, and energy consumption) and other criteria related to the fungus bio-degradation process. The framework was evaluated for a case study in Mashahd, Iran, considering different economic, energy, and environment-related situations. Results indicated that implementing plastic waste bio-recovery led to an increase in the satisfaction of citizens with the urban management system from 49% to 64%. Other relevant indicators increased too, e.g., the development of waste industries, the achievement of smart city goals, and the reduction in hazardous material emissions from solid waste. The need to consider other pertinent social factors (job opportunities, the impact on employment, and the number of working days lost during WMS implementation) in order to develop a more comprehensive model was acknowledged. The integrated model was presented as applicable to other domains, including supply chains and energy-based systems.

Ogawa et al. [167] analyzed smart waste collection technologies worldwide and performed participant observation to assess the waste collection system in Japan. Building upon the identified trends and current situation, a contactless waste collection system was developed, using a combination of smart bins and adapted collection vehicles. A practical evaluation was conducted using data from Fukuoka City, Japan. The demonstration successfully showcased the capability of the proposed system to safely lift a waste container weighing 212 kg onto a collection vehicle without any human contact. As a result, the number of laborers required to operate the collection truck was reduced from two to one. No real evaluation or deployment was reported. The study revealed the potential of smart bins in promoting the reduction of packaging waste consumption.

5.3. Topic: Social Aspects of WMS in Smart Cities

Social aspects play a crucial role in waste management, particularly in the context of smart cities. Smart WMS not only focus on technological advancements but also prioritize citizen engagement and behavior change. Waste management in modern cities involves educating and empowering citizens to actively participate in waste reduction, segregation, and recycling efforts. This often includes initiatives such as public awareness campaigns, community-based programs, and incentives for sustainable waste practices. By fostering a sense of responsibility and ownership among residents, smart cities effectively reduce waste generation and improve recycling rates. This section aims to categorize articles that emphasize social aspects rather than technological innovations. It is noteworthy that classifications may overlap, with some articles encompassing multiple subcategories. However, each work has been categorized based on its most salient contributions. See Appendix A.

5.3.1. Smart WMS and Circular Economy

A relevant topic that appears regularly in the bibliography of waste management in smart cities is the concept of circular economy. In a general view, the goal of a circular economy is to replace the traditional “take-make-dispose” linear paradigm of economy with a more sustainable paradigm in which resources are used for as long as feasible while waste and pollution are reduced [42]. This involves designing products for reusability, recyclability, and repairability and creating closed-loop systems where waste from one process becomes a resource for another. Aceleanu et al. [168] performed a high level analysis to assess the degree of application of circular economy in Romania. After a careful revision, the authors concluded that major efforts needed to be addressed from companies in order to produce easier to recycle and/or re-manufacture products, for governments to enact laws that encourage sustainable policies and, for citizens to change their consumer habits and increase source classification of waste. Cooperation between the three mentioned stakeholders was identified as really important to achieve a sustainable smart WMS.

Franchina et al. [169] presented a general discussion on how smart technologies contribute to implementing the principles of a circular economy in cities. As an example, a previous case study from the literature related to waste management was reported. Dincă et al. [170] examined the factors related to circular economy that impact the environmental condition of smart cities in European countries. Several critical factors of waste management connected to circular economy that influence the environmental outcomes of smart cities were highlighted: the implementation of taxes and restrictions on landfilling and incineration of waste, taxation based on the quantity of waste generated, economic incentives for local and regional authorities to encourage waste reduction, the enhancement of separate waste collection programs, strategic planning of investments in waste management infrastructure, research and innovation in advanced waste treatment, recycling, and remanufacturing technologies, public education and awareness campaigns to mitigate environmental pollution from waste, and the promotion of selective waste collection.

Formisano et al. [171] studied the relationship between smart sustainable cities and circular economy in 391 large European cities. A correlation analysis between two metrics was performed. One metric measured the ratio of the number of scientific articles connected to “smart cities” and the number of scientific articles on any topic in Google Scholar. The second metric was similar but using the keyword “circular economy”. A total of 193 out of the 391 cities, most of them located in Germany and Italy, showed a high correlation between these two metrics. Maiurova et al. [172] analyzed the capacity of implementing circular economy to enhance the Moscovian WMS, which has been under a lot of pressure due to the increment of waste generation. Data were collected from semi-structured interviews with relevant stakeholders, along with literature revision. The drawbacks and bottlenecks of the system were identified, comparing with the successful case of Berlin, Germany. Digitalization was set as a key element to fill the gap between the two cities, including different improvements, such as using IoT technology for smart bins, route optimization schedulers for collection, digital tracking, and charging citizens for the exact amount of disposed waste.

Muheirwe et al. [173] analyzed the impact of regulations in WMS of different Sub-Saharan African countries, highlighting the importance of contextualization in public policy making. Their findings indicated that the mere adoption of best practice regulations from developed nations did not necessarily enhance the sustainability of WMS in Sub-Saharan Africa. Rather, the effectiveness of such regulations hinged on their adaptation to the specific socio-economic, cultural, and infrastructural dynamics of the region where they were implemented. Rena et al. [174] performed a similar study to demonstrate how different regulatory initiatives have helped India enhance its complex WMS to make it more sustainable. According to the authors, another beneficial factor in enhancing sustainability in waste management in India was the dynamism of the private sector, mainly the startups related to the development of eco-innovations.

Colangelo et al. [175] performed a cost–efficiency analysis to select the best technology for converting waste to energy: incineration, gasification, or flameless oxy-combustion, to identify the best option for implementing a circular economy in Bari, Italy. The environment impact of the process was only considered as an operational cost to acquire the rights for generating the carbon emission. Considering the current prices of electricity and carbon, gasification was found to be the best option. However, after performing an extensive sensitivity analysis on key parameters of the analysis, the authors concluded that in a future, possibly greener economy where extracted carbon dioxide can be sold, the newer technology of flameless oxy-combustion might prove to be more profitable.

Kurniawan et al. [176] studied the application of circular economy practices already implemented in Naning, China to the Indonesian city of Malang. Their findings underscored the potential for enhancing WMS through digitalization initiatives. Specifically, the study emphasized the importance of leveraging data-driven approaches to efficiently classify waste at recycling plants, thereby reducing processing costs. Additionally, the adoption of smart bins emerged as a key strategy for optimizing waste collection processes. Listiningrum et al. [177] studied how laws and regulations serve as a basis for implementing circular economy strategies in waste management of a smart city in Indonesia. The article concluded that while the law is relatively strict regarding waste generators, it is lax regarding the responsibilities of the authorities. Additionally, there is not an integrated management system in Indonesia, leading to confusion and a disclaimer of responsibilities among the actors that are in charge of the different stages of the system.

Moslinger et al. [178] studied questionnaires issued to stakeholders in a set of 362 cities that expressed interest in the Horizon Europe 100 Climate-Neutral and Smart Cities Mission. One of the main findings was that in order to enhance the circularity of economies in these cities, governments need to improve their expertise in the fields of circular economy, digitization, and emissions accounting and need to set clear key performance indicators that are easy to understand and can be used to measure the transition from linear to circular economies. Moreover, governments should take into account the trade-offs of different solutions, e.g., the rise in the budget expenses that these strategies may represent at the beginning, encouraging the formation of collaborative networks between the industry and research institutes, and implementing specific legislation.

Onesmo et al. [179] analyzed the scrap market in Arusha, Tanzania, using data from documents by the local government and interviews with stakeholders, scrap collectors, scrap dealers and government officers. Gathered data included the quantities of the different scrap materials collected in the city (mainly metals but also other minerals), the usual selling prices of each material, and the main selling markets. The article also studied the social impact of the scrap material logistic chain in the city, which is important as a labor source for waste pickers. Finally, the article analyzed the reduction in the environmental impact of the local industry due to the presence of an important scrap business in the city. Scrap material serves to feed the regional industry with large amounts of iron ore, coal, and bauxite reducing the usage of virgin material. Furthermore, the scrap material reduces the need to process raw material and, thus, contributes to an up to 56% reduction in energy consumption in the industry.

Yadav et al. [180] analyzed the critical barriers faced to implement a smart WMS and circular economy principles in Indian cities. The data collection included a revision of the related literature and questionnaires distributed to experts from both government representation and technology industrial sector. The main identified barriers were the lack of strict regulatory policies and the lack of proper financial and benchmarking processes.

5.3.2. Field Studies of WMS

Singh and Leena [181] performed a forecast based on mathematical equations surrounding the emissions of greenhouse gases by WMS in Faridabad, India. Two scenarios were compared: an scenario in which waste is directly deposited in the landfill, and another scenario in which waste is transformed into energy based on incineration, with the second scenario being more beneficial in terms of greenhouse gas emissions. The scenarios were solved considering only deterministic parameters. Thus, the results could be enhanced with uncertainty using a simulation model. Mingaleva et al. [182] presented a descriptive analysis of WMS in Perm, Russia, to assess its sustainability. Citizens with environmentally friendly attitudes were identified as major contributors. Building recycling plants were proposed to increase the recycling rate, and it was recommended that they be distributed throughout the district instead of building a single large plant, in order to minimize waste transport costs.

Rai et al. [183] conducted a survey on both stakeholders and common citizens to analyze citizen satisfaction with WMS in the Nepali city of Bharatpur. Results showed that more than half were dissatisfied with the system due to waste collection issues, mainly the inconvenient timing of collection and the lack of community bins in the streets for disposing of daily waste. The study also revealed that citizens were willing to pay up to 30% more than the current rate of the service to improve these negative aspects of the system.

Onoda [184] analyzed the use of smart technologies to improve waste management post-COVID-19 in Japan. Different initiatives were discussed: virtual reality for remote sustainability education, traceability for medical waste, robotic arms in waste facilities, and self-driving cars for waste collection. The initiatives exemplified innovative approaches implemented in Japan to address the challenges impeding more sustainable WMS. Rahmayanti et al. [185] conducted a survey involving 440 elementary school students from two Indonesian cities to assess their attitudes towards implementing a smart WMS. The findings revealed a strong inclination among younger generations towards embracing smart waste solutions.

Popova and Sproge [10] performed surveys of citizens and interviews with authorities to examine the factors influencing household waste classification. The research revealed a notable divergence in waste sorting behavior between residents of one-family houses, who displayed greater willingness to sort waste, and those in multi-family dwellings, who were less inclined to do so due to space constraints for multiple bins. Moreover, the surveys indicated a lack of awareness among citizens regarding the benefits of waste sorting. In contrast, responses of authorities tended to prioritize cost-efficiency concerns. Results suggested a need for targeted educational initiatives to raise awareness of waste sorting among citizens, alongside policy measures to address logistical challenges, particularly in multi-family housing. Measures were proposed for enhancing waste sorting in the region.

Saptadi et al. [186] performed a survey of 49 citizens across five districts in Makassar, Indonesia, to analyze the willingness of people to correctly dispose of their waste in a rudimentary WMS. Using the k-means clustering algorithm, two main groups were identified: a group of 39 people who followed the authorities’ suggestions about waste disposal, and a group of 10 people who were less responsive to waste disposal guidelines.

Whiteman et al. [187] introduced a conceptual framework aimed at transitioning from traditional WMS to a “waste and resource management” system, aligning with the sustainable development goals of the United Nations. The framework categorizes WMS into nine categories based on their level of development. To illustrate the application of this framework, the article analyzed two case studies, in Bo, Sierra Leone and Muncar, Indonesia; Mohanty et al. [188] presented a brief descriptive analysis of WMS in Bhubaneswar, India, highlighting its vulnerabilities exacerbated by a recent population increase. The study primarily outlined the shortcomings of the system, due to limited financial resources and inadequate planning. Kurniawan et al. [189] examined WMS in St. Petersburg, Russia, to draw lessons from the successful experience of Taiwan in improving their system. The article stated that St. Petersburg should prioritize initiatives to promote waste classification at the household level, digitizing the recycling industry, and optimizing operational costs through improved system planning in order to increase the efficiency of WMS.

Suryawan and Lee [190] introduced an evaluation framework for estimating the willingness of citizens to pay for implementing a smart WMS in Jakarta, Indonesia. Several key factors influencing the willingness of citizens to pay more for enhanced waste services were identified, including socioeconomic status, age, and environmental awareness. Among the citizen demands from the proposed system, the ability to classify waste and provide easily understandable information emerged as primary requirements.

Zhao et al. [191] studied the impact of demographic variation and smart city development on waste generation in the Chinese Region of Circum-Bohai-Sea. From this study, it was observed that population decline and aging, alongside advancements in smart city infrastructure, were significant factors leading to a decrease in waste generation rates in urban areas. The article emphasized the importance of extending smart city initiatives to rural areas as a means of further reducing waste generation, thereby promoting more sustainable waste management practices across the region.

5.3.3. Efficiency Analysis of WMS

Cheela et al. [192] performed a comparative analysis in six Indian smart cities, considering different indicators related to smart waste management, i.e., total budget expended on the system, number of treatment plants, number of landfills and open dumps, transfer stations, size and type of vehicle collection fleet, use of waste accumulation collection points, and waste generation rate. The authors discussed how the indicators varied from city to city and suggested some practical strategies to improve these indicators in cities. However, the article did not analyze relative efficiency, leaving this issue as a consideration for future work.

Jonek [193] studied the efficiency of WMS of sixteen Polish capitals based on a multicriteria analysis. The indicators to determine the smartness of a waste system of a city were the waste generation rate per capita, the annual variation rate of generated waste per capita, the proportion of mixed (unclassified) waste that is collected in the city over the overall collected waste, and the cost-to-effectiveness ratio calculated as the expense of the system divided by the amount of waste collected. The WMS of the largest Polish cities (Warsaw and Wrocław) were found to be relatively inefficient and, thus, public policies should focus on reducing the waste generation rate per capita and increasing cost efficiency. However, the article did not discuss how to achieve these improvement in WMS.

Thakur et al. [194] identified 13 factors of sustainable WMS based on a literature review, surveys provided to citizens, and the opinions of stakeholders. The opinions of stakeholders were analyzed using interpretive structural modeling to propose a hierarchy between the factors. Then, these factors were used to analyze the Indian city of Odisha as a case study. A final discussion studied a set of strategies to make WMS of the city more sustainable.

Öztaş et al. [195] presented a life cycle assessment analysis of WMS in Istanbul, Turkey. After gathering data from field work and literature, a simulation model was used to estimate the environmental passives produced by the system considering different scenarios. The data gathering process was briefly described. Based on simulations, the most environmentally friendly scenario was selected, which comprises an integrated management of the system with a balanced disposal of waste through incineration, composting, and recycling.

Rafiquee and Shabbiruddin [196] assessed the WMS in Patna, India, using TOPSIS, based on the opinion of three experts in the field. Various evaluation criteria were considered, including capital investment requirements, social acceptance, quantity of collected and processed waste, and operational complexity. Notably, social acceptance emerged as the most critical factor for the suitability of WMS and the door-to-door collection system with the addition of smart bins was regarded as the best system for the city.

5.3.4. Other Social-Related Aspects of WMS

Fatimah et al. [197] investigated the issues and opportunities to develop a sustainable and smart WMS using Industry 4.0 technologies in four Indonesian cities. Data were collected through a comprehensive systematic literature review, intensive focus group discussions, and direct observation. The article mainly presented a classification of WMS into social, governmental, technological, environmental, and economic dimensions and five different levels of maturity of “smartness” for each dimension. Almalki et al. [198] noted a specific drawback associated with IoT technologies: their tendency to generate a significant quantity of electronic waste. Consequently, the authors reached the conclusion that a meticulous selection of IoT technologies is imperative to mitigate the accumulation of electronic waste. However, this conclusion was not supported by specific examples.

Peoples et al. [199] proposed a business model aimed at identifying the prerequisites for the implementation of a WMS within a city. A survey was performed to asses how compromised citizens would be with smart waste management program. The survey collected information whether citizens would accept modifications in collection frequency, source separation of waste, collaboration with other neighbors, and sharing data with the stakeholders in order to enhance WMS. A method for analyzing the cost efficiency of the program depending on the number of citizens willing to participate was proposed. However, no application of the proposed business model to a real case study was presented in the article. Vrabie [200] studied waste-to-energy conversion technologies in four European cities with advanced WMS. A supply chain was introduced, where waste generation is the first stage of the chain and energy production is the last stage. The advanced WMS was studied to extract useful strategies to implement waste-to-energy technologies in Romania.

Costa et al. [201] conducted a large-scale survey to evaluate the awareness of cities regarding governmental initiatives aimed at enhancing system sustainability. The findings revealed significant support for source classification programs, particularly among female and elderly populations. However, there was poor reception of smart waste management initiatives, indicating a low impact on citizens, many of whom exhibited a high level of ignorance regarding these initiatives. Zérah et al. [202] presented a study on how digitalization has affected WMS in Mangaluru, India. Through an informal survey conducted on the streets, the authors found significant mistrust among citizens regarding how the collected data would be used, highlighting the necessity for strong information privacy regulations from authorities. Additionally, the authors emphasized the need for proper training for workers to gather, process, and use information effectively.

6. Discussion

This section discusses the main findings in the reviewed articles, to identify emerging trends and opportunities for further research and development in the studied field.

Many published strategies for analyzing and optimizing waste collection management systems involve computational intelligence, IoT, data analytics, spatial analysis, and data-driven approaches. These strategies are based on TSP and VRP models, with few comparisons to state-of-the-art benchmarks being reported. Most proposed methods utilize computational intelligence, e.g., metaheuristics or machine learning, to optimize and estimate collection vehicle routes. Most proposals use real-time IoT devices in smart bins, validated through real-world case studies. However, only some comprehensive systems are combined with intelligent systems for smart cities. Likewise, only one article proposes a routing approach for waste collection using electric vehicles, but that in the future will require managing waste from electric car batteries, which opens another area of research. There is no specific waste management benchmark to test optimization methods for a general comparison. Most cases compare local real-world waste management procedures with experimental results obtained by the optimization methods. Thus, future research in smart WMS should focus on applications incorporating electric vehicles in waste collection and developing collection plans based on computational intelligence. These approaches are crucial for establishing institutional structures, policies, plans, and regulations for waste collection services in local and national contexts to minimize costs and maximize collection. Another research gap is addressing the processing of end-of-life batteries from electric vehicles, aligning with circular economy principles. This is a significant area for future research, especially when considering the next generation of Industry 5.0, automation, flying vehicles, autonomous vehicles, AI, human–robot collaboration, and electric vehicles. All these factors play a key role in supporting sustainability and the circular economy.

Some articles have presented blockchain approaches for smart waste management, with direct application in smart planning and collecting, finding the best routes for disposal, promote recycling and circular economy, and for fostering trust among stakeholders. Blockchain technologies have been often used in conjunction with other tools such as Business Process Model, smart contracts, and data privacy methods, applied to improve the traceability, interoperability, end efficiency of the waste management process. The main gaps in blockchain technologies are the incompatibilities in sharing data and the need for adoption and compatibility for private and public institutions and organizations involved in waste management. However, the potential of blockchain technology to revolutionize waste management is immense. It is also essential to consider technological infrastructure and design applications, decentralized and transparent strategies, and constraints in the finance process, blockchain-based transactions, supply chain, automation, and security. These research gaps highlight the need for further studies to explore the applications and implications of blockchain technology in WMS and its legislative adoption.

Regarding characterization and forecasting of waste generation, several variables have been considered in the reviewed studies, including demographics and socioeconomic variables. Other specific features of WMS have also been considered, such as the types of waste and the specific collection practices. Most articles have studied open data for characterization and forecasting, but the most accurate results have been obtained using multivariate approaches considering both open and proprietary data properly correlated with solid waste generation. Several statistical metrics have been used for results comparison. Few articles have proposed generalized models to be applied to different scenarios. Accurate prediction of waste generation is essential for the sustainable development of a city. This makes it difficult for management to develop a multipurpose goal for a sustainable pathway to minimize the gaps in existing systems. Also, the lack of a waste management plan and historical documents hinders the development of local and national public policies for medium and long-term periods. Thus, gaps in forecasting and characterization of waste management should be analyzed using statistical methods and AI estimation methods to help manipulate data for decision-making.

Waste recognition and classification have been approached using semi-automated methods, mostly using CNN for object recognition and classification. A significant number of proposals have described integrated systems to take advantage of smart bins with IoT capabilities. Several published approaches for waste recognition and classification are rather simple and few comparisons with state-of-the-art methods have been reported. Most proposals using ANN and deep learning for waste recognition and classifications have been validated over standard datasets, e.g., from Kaggle or Google Images repositories. Few implementations have been validated on real scenarios. Current waste classification models still have several issues, such as poor classification accuracy or lengthy run times, and limitations surrounding household waste generation using time-series data, high computational demands for computer vision algorithms, IoT integration for data acquisition, adequate model selection and architecture, precision of pollution detection sensors, among others. Thus, future research should focus on closing the gap between the different classification models, aiming towards standard models that facilitate the implementation of AI in the classification of organic, domestic, medical, and recyclable waste towards a sustainable and circular economy.

Many articles have proposed infrastructures for smart bins, using different components. The most included have been ultrasonic sensors for determining the filling level, servo motors for automatically open and closing lids, and detectors for waste and humans. GPS devices have also been often used to gather information. Arduino and Raspberry Pi system-on-a-chip microcontrollers have been used for controlling the smart bins. Thus, future research must focus on emerging and modern technologies based on sensors, IoT, and autonomous vehicles, aiming towards efficient systems of waste storage and collection. Additionally, smart bin materials and hardware technologies for enhancing durability and compaction techniques must be studied, including real-time data to monitor the fill levels and automatize the collection based on intelligent schedules generated by computational intelligence and electric autonomous vehicles. This way, dynamic approaches to optimize waste collection routes and minimize unnecessary trips will contribute to reducing the carbon footprint.

In turn, IoT-based WMS have relied on smart bin infrastructure, often integrated with lightweight and low-power consumption protocols for data communications. Most systems have proposed centralized data processing. Some used cloud or fog computing, and a few have relied on edge computing technologies to defer specific light data processing tasks to distributed devices. Thus, future research must focus on automated waste management using IoT technologies via dynamic IoT systems, real-time analytics, IoT services, IoT-powered waste monitoring in residential areas and public spaces, embedded sensor networks and services, and integration within smart cities, moving towards sustainable urbanization.

A specific drawback of the proposed systems is that they have not been evaluated on real scenarios, so it is not easy to determine their effectiveness and contributions to impact on the citizens lives. Overall, few articles (27 out of 79) have addressed real-world scenarios. Most of the proposals were evaluated via simulations or in the laboratory. A specific line of interest in future research should be developing real-world case studies, proposing real benchmark scenarios that consider relevant features of nowadays cities and up-to-date data.

Regarding social-oriented articles, several research studies have focused on the circular economy. Circular economy policies have been reported as an asset to encourage waste reduction in smart cities, and particularly in special waste fractions such as e-waste. Several published approaches considered the legal framework for enforcing circular economy policies as a key element for smart WMS. Few articles included the opinions of citizens with representative surveys when performing their studies. A total of 12 out of 38 articles presented field studies in which the main focus was on describing a specific WMS. Several articles presented a brief section of strategies to enhance the particular WMS analyzed. However, only a few articles went deeper in the analysis of the implementation of these strategies. A few articles performed efficiency studies of different WMS (5 out of 38). Some of them compared different cities in the same country to pick the most efficient WMS, and others compared different policies to select the best to be implemented in a city. Finally, six articles studied other social related aspects of WMS, two of them without analyzing a real case study.

Regarding the application in social-oriented subjects, 32 out of 38 articles used real case studies to illustrate their analysis. Thus, the usage of real case studies is much more common in social-oriented articles than in the technology-oriented articles. Among the articles that dealt with case studies, the most frequent countries were India (eight articles) and Indonesia (seven articles). A few articles (five) dealt with case studies from more than one country. Three articles considered many European cities/countries in the same study. However, these articles tended to arrive to general conclusions about WMS without considering the fact that they were analyzing only European case studies.

Based on the literature review, several future research directions are to be considered relating to social-oriented aspects. There has been limited development of large-scale surveys of citizens in the related works. Although conducting such surveys was extremely costly and time-consuming in past decades, the development of digital and smart technologies now facilitate this task. Integrating the opinion of citizens with the expert knowledge of decision-makers and stakeholders provides a broader view of the system from the perspective of users. Additionally, there is a certain need for stronger integration between analysis of WMS and action plans to improve the weak points detected in the analysis. Several articles only present a diagnosis of the current situation of WMS and very little discussion about how to improve WMS. Providing insightful diagnoses of specific WMS is useful, but including fruitful discussions regarding possible action plans and their potential implementation is much more helpful for practitioners that have to manage WMS. The academic community should strive to integrate diagnoses and proposals to provide more comprehensive and actionable insights. A few articles have emphasized the need for policies and regulations to promote sustainable smart waste management systems (WMS). Current regulations primarily focus on waste generators within the circular economy framework, with limited attention paid to the responsibilities of authorities. However, some studies caution against directly replicating legislation from other cities, as this approach may not ensure success in different cultural contexts. Instead, regulations should be case-sensitive and tailored to local conditions. Further discussion on implementing tailor-made policies that foster sustainable and integrated smart WMS is essential. Additionally, these regulations should encourage private sector participation in WMS.

In moving forward, a comprehensive research agenda is essential for advancing waste management practices through innovative solutions. Future efforts should focus on the integration of innovative solutions with institutions and government bodies to establish policies that promote sustainability and address the unique needs of urban environments. Additionally, active participation from various stakeholders—including local communities, environmental organizations, and the private sector—will be crucial in fostering a holistic approach to waste management. Collaboration with private industry can play a significant role, not only by providing resources and technology, but also by encouraging a shared responsibility in reducing environmental impact. By addressing these aspects, future research can contribute to developing adaptable and effective waste management solutions that align with sustainability goals, ultimately benefiting both local and regional communities.

7. Conclusions

Waste management is a crucial activity in modern cities, with a significant impact on the everyday lives of citizens. Consequently, it has traditionally been a focus of attention for both authorities and the research community, aiming to develop more sustainable and efficient WMS. With the advent of smart cities, these efforts have benefited from new computational-aided technologies that support the decision-making process in waste management. In this line of research, this article presented a thorough review of the main recent advances in innovative computational tools for waste management in smart cities.

Following a systematic research methodology to select the most relevant articles retrieved from prestigious databases, more than 190 scientific articles were analyzed. Overall, this article offers an updated description of waste management practices in modern smart cities, a bibliometric analysis of the related literature, a review of existing reviews on related topics, and a categorization of application articles into three main categories: computational optimization methods in waste management, computational intelligence and IoT in waste management, and social aspects of waste management in smart cities. The analysis of each of these three categories includes further classification into subtopics, the identification of the main findings and limitations of each article, and a description of the main case studies. Finally, a discussion outlined the main emerging trends and opportunities for further research and development in this rapidly evolving field.

After the revision of the related literature, different research gaps were identified in each category. Regarding computational optimization methods in waste management, special efforts must be made to integrate real-world and intelligent systems for smart cities. Route optimization problems for waste collection using electric vehicles and managing waste from electric car batteries, drones, or other vehicles must also be addressed. Regarding computational intelligence and IoT in waste management, special effort must be paid to developing real-world implementations of the proposed techniques, in order to properly validate and guarantee the effectiveness of the proposed approaches. Regarding the social aspects of waste management in smart cities, there has been limited development of large-scale surveys of citizens in related works and insufficient integration between the diagnosis of WMS and action plans to address the weak points identified during the diagnosis phase. Additionally, as highlighted in several reviewed articles, a detailed discussion on how policy and regulatory frameworks can support the successful integration of computational intelligence technologies would be a valuable addition to the context of smart waste management systems. Thus, future research should address these research gaps and strive to further integrate computational-aided technologies in the planning and operation of efficient and sustainable smart cities.