Requirements Engineering for a Drone-Enabled Integrated Humanitarian Logistics Platform

Abstract

The pursuit of ameliorating humanitarian logistics (HL) through the integration of cutting-edge technologies has received significant attention in recent years. AIRDROP is a visionary platform conceived to offer a cohesive disaster management approach spanning from preparedness to recovery of a wide range of natural and human-made disasters. AIRDROP aims to be a scalable, modular and flexible solution, employing an array of drones of different sizes and payload capabilities, able to provide different HL services to first responders and operational decision-makers. This study aims to elicit, specify and validate the requirements for AIRDROP to ensure their applicability across a broad spectrum of disaster scenarios and the entire disaster management continuum. This research utilized a thorough literature review and expert consultations to systematically elicit and specify the AIRDROP requirements, ensuring they were grounded in both academic foundations and practical industry standards. The validation process involved a questionnaire survey administered to 26 participants from various professional backgrounds. The requirements were prioritized using the MoSCoW methodology, and significant differences among participant groups were identified through the Kruskal–Wallis H and Mann–Whitney U tests. Furthermore, two critical requirements emerged from open-ended responses. As a result, 276 out of the initially defined 335 requirements in total advanced to the design phase. It is worth noting that the dynamic nature of requirements in HL necessitates ongoing assessment and adaptation to keep AIRDROP at the forefront and aligned with evolving needs.

1. Introduction

The indisputable digital transformation is disrupting entire sectors with the emergence and development of new business models that rely significantly on digital technologies [1,2]. This trend not only refers to the commercial sector but also extends to the pursuit of ameliorating disaster management through the integration of cutting-edge technologies, which has received significant attention in recent years. Acknowledging the impossibility of preventing all disasters, the utilization of such technologies holds the potential to transform humanitarian logistics (HL) operations at pre-, intra-, and post-disaster stages by making them more accurate, agile, and resilient [3,4,5]. The integration of unmanned aerial vehicles (UAVs) in humanitarian operations, as evidenced by their revolutionary capabilities in communication, monitoring, and transportation, has played a crucial part in this advancement [6]. Moreover, a diverse array of disruptive digital technologies, including big data analytics [7], Internet of things (IoT) [8], cloud [9], edge [10], fog computing [11], artificial intelligence (AI) and machine learning (ML) [12], social media and crowdsourcing [13], robotics and cyber–physical systems [9], blockchain [14], and extended reality [15], has also been harnessed to support HL initiatives.

A systematic literature review was previously conducted, with a view to addressing the need to understand the complementarity and interoperability between humanitarian drones and these technologies [16]. This investigation resulted in the formulation of a novel framework, elucidating a plethora of dynamic synergies in a universal disaster context. In the pursuit of this framework, we laid the cornerstone for conceptualizing AIRDROP, a platform that aims to offer a cohesive disaster management approach spanning from preparedness to recovery of a wide range of natural and human-made disasters. AIRDROP aims to be a scalable, modular and flexible solution, employing an array of drones (UAVs, unmanned ground vehicles (UGVs), unmanned surface vehicles (USVs)) of different sizes and payload capabilities, able to provide different HL services to first responders and operational decision-makers.

The need for a holistic platform that addresses the full spectrum of disaster management underscores the importance of a meticulous requirements engineering process and is further amplified by the complex interplay between diverse technologies and the multifaceted nature of HL operations. The manner in which requirements engineering is carried out in a given setting is a vital component in successful software development projects [17]. Visualized as the bedrock upon which the entire project edifice is erected, requirements intricately bind the various stages of project development, from inception to fruition. Requirements in the literal sense can be referred to as anything that is discovered with a view to developing products [18]. They encompass the critical features a product has to successfully address so as to fulfill its purpose, while meticulously taking into consideration a spectrum of factors, including technical aspects, resource availability, risk factors, and budgetary constraints [19]. A lack of robustness in, or non-abidance by, the predetermined requirements may lead a system astray from its intended purpose, ultimately culminating in its failure.

While the reviewed literature highlights various initiatives in the field, it is evident that no single system comprehensively addresses the full spectrum of disaster management needs. Acknowledging the deficiencies in current solutions and AIRDROP’s capacity to address them, the necessity for a rigorous requirements engineering process for such a holistic HL platform is underscored. This recognition forms the basis of our investigation, leading us to formulate two critical research questions (RQs):

• RQ1: What are the foundational requirements for developing a holistic HL platform, such as AIRDROP, considering both functional and non-functional aspects?
• RQ2: How can the relevance and adequacy of the identified requirements for the AIRDROP platform be ensured to meet the complex dynamics of HL operations?

Addressing this literature gap, this paper makes a significant contribution to the field of HL and disaster management technology integration. Specifically, it uniquely provides the first comprehensive framework for the elicitation, specification and validation of the requirements necessary for the development and operational success of a holistic HL platform, such as AIRDROP. Through expert engagement and rigorous analysis, this work not only identifies and prioritizes key functional and non-functional requirements but also aligns them with the complex needs of humanitarian operations. Such a systematic requirements engineering process is pivotal for bridging theoretical concepts with practical applications, thereby offering a robust foundation for the effective design and implementation of technology-driven HL solutions.

The paper is organized as follows: Section 2 presents a comprehensive literature review of previous research efforts in the domain of humanitarian drone applications, with the aim of elucidating their areas of focus, including the technologies used, the disaster phases of application, and the range of services offered, as well as highlighting the research gap AIRDROP attempts to address. Section 3 provides a meticulous exploration of AIRDROP’s diverse array of proposed services and functionalities. Section 4 details the research design employed in this study, explaining the adoption of an abductive research approach and the integration of design science research (DSR) methodology alongside survey research. In Section 5, a structured methodology for requirement elicitation and specification is presented, followed by requirement validation, which is conducted through a structured questionnaire. Section 6 delves into the presentation of questionnaire analysis results, including descriptive statistics and the outcomes of non-parametric tests. Central to the validation process is the MoSCoW (Must Have, Should Have, Could Have, Would Have) methodology, which is employed to prioritize these requirements. The significance of differences in participant responses is assessed through the Kruskal–Wallis H test and Mann–Whitney U test. Section 7 offers an in-depth discussion of the implications drawn from the results, focusing on their significance for the design of AIRDROP. Finally, the paper concludes in Section 8 and Section 9, summarizing the key findings and insights obtained throughout the study and addressing research limitations.

2. Literature Review

The drone industry has witnessed unprecedented growth and technological advancement over the past decade. Initially developed for military applications, UAVs have transcended their original purposes and are now integral to various sectors, including commercial enterprises, scientific research, agriculture, and emergency response [20]. The global drone market is projected to grow at a compound annual growth rate (CAGR) of 15.3%, reaching a valuation of USD 16.7 billion by the end of 2033, while the humanitarian drone market, accounting for nearly 14% of the global drone market at the end of 2023, is expected to expand at a CAGR of 13.1% [21]. In disaster management, UAVs have transformed operations by providing rapid and precise situational awareness, high-resolution mapping, search and rescue (SaR), delivery of supplies, communication support, and training [22]. Furthermore, the integration of complementary technologies further improves their effectiveness and efficiency. This section reviews various projects that employ drones alongside other advanced technologies for real-life disaster management applications.

In wildfire scenarios, UAVs equipped with thermal imaging cameras are used to detect hotspots and track fire spread, aiding in effective resource allocation and evacuation planning. For instance, the ResponDrone project employs UAVs for real-time monitoring and communication relay during forest fires, improving the coordination of first responders and reducing response times [23]. The STEReO project combines unmanned aircraft systems (UAS) autonomy with resilient communication systems to expedite disaster response in wildfire scenarios, facilitating extensive aerial operations and ensuring continuous connectivity [24,25]. Furthermore, UAVs assist in flood response by capturing aerial imagery for obstacle identification and route planning, as utilized in the DRIVER+ project [26,27,28].

Complementary technologies such as UGVs, augmented reality (AR), virtual reality (VR), ML and social media sensing augment the capabilities of UAVs in disaster management. UGVs, such as those used in the INTREPID project, navigate through hazardous environments to perform tasks, such as rubble removal and survivor identification in structural collapse scenarios, enhancing the safety and efficiency of rescue operations [29]. In SaR operations following earthquakes, UAVs with high-resolution cameras and sensors along with miniaturized robotic equipment locate trapped victims under debris, as demonstrated by the CURSOR project, which has successfully contributed to decreasing victim detection time and increasing localization accuracy in such scenarios [30]. Similarly, the SEARCH & RESCUE project supports the convergence of UAVs and UGVs for early victim localization and first responder safety in SAR operations [31]. VR technology, as integrated into the ASSISTANCE project, provides immersive training simulations for first responders, improving their preparedness and proficiency in managing large-scale disasters [32,33]. The FASTER project utilizes UAVs and UGVs for mapping, transportation, and resource delivery, while AR tools provide real-time information, increasing responders’ effectiveness and safety in various emergencies [34]. The RESPOND-A project ensures continuous connectivity and improved situational awareness for first responders using intelligent wearable sensors, AR and VR, as well as drone-assisted smart monitoring [35].

Zipline International utilizes an intelligent automated system to deliver blood, vaccines, and medical devices to people in need, highlighting the potential societal benefits of ML technologies in the drone delivery industry [36]. The INGENIOUS project employs a combination of UAVs, UGVs, and smart wearable sensors to enhance first responder safety and operational capacity in various disaster scenarios, from earthquakes to chemical spills [37,38]. The TeamAware platform leverages UAVs for visual scene analysis and integrates social media sensing for situational awareness in both natural and human-made disasters, improving coordination and response effectiveness [39]. The AiRMOUR project aims to enhance the safety and security of passenger drones in emergency medical services [40]. The TOAS project focuses on precision in herbicide application and crop weed management using UAVs, addressing pest infestations as a form of disruption [41]. Lastly, the PathoCERT project employs satellite data, social media analysis, and data processing to detect and assess water contamination, enhancing situational awareness in events such as earthquakes and floods [42,43].

This exploration of past efforts has illuminated the research gap that AIRDROP aspires to address. While the reviewed literature highlights various initiatives in the field, it is evident that no single system comprehensively addresses the full spectrum of disaster management needs. AIRDROP’s primary objective is to bridge these existing gaps by consolidating insights from previous efforts and offering a holistic disaster management approach that spans from preparedness to recovery for a wide range of natural and human-made disasters, effectively aligning itself with the academic endeavor for holistic methodologies in HL [44].

Table A1. AIRDROP and pertinent drone-assisted HL research projects: a comprehensive comparison of disaster stage foci, technologies used and services provided.

Pertinent Drone-Assisted HL Projects	AIRDROP	ResponDrone	ASSISTANCE	INTREPID	STEReO	FASTER	CURSOR	RESPOND-A	PathoCERT	INGENIOUS	DRIVER+	TeamAware	SEARCH & RESCUE	AiRMOUR	Zipline International	TOAS
Reference No.	This	[23]	[32,33]	[29]	[24,25]	[34]	[30]	[35]	[42,43]	[37,38]	[26,27,28,101]	[39]	[31]	[40]	[36]	[102]
Disaster Stage	Study
Mitigation	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓			✓
Preparedness	✓		✓					✓	✓				✓
Response	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓
Recovery
Technologies Used	✓
UAV	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓
UGV	✓		✓	✓		✓	✓	✓					✓
USV	✓
AR	✓		✓	✓		✓		✓	✓	✓		✓	✓			✓
VR	✓		✓	✓			✓	✓
IoT	✓	✓	✓	✓	✓	✓	✓	✓	✓			✓	✓
AI (ML)	✓	✓		✓	✓	✓	✓	✓	✓		✓	✓	✓		✓
Social Media Sensing	✓					✓			✓	✓		✓
Blockchain						✓
Services Provided
Area Scanning	✓
Fire Detection	✓	✓	✓	✓	✓	✓	✓	✓		✓		✓
Smoke Detection	✓	✓	✓	✓		✓				✓		✓
Missing Person Detection	✓	✓		✓	✓		✓	✓			✓	✓	✓
Flood Detection	✓								✓		✓
Weather Monitoring	✓													✓
EMF Detection	✓
Hazardous Gas and Radiation Detection	✓			✓		✓						✓	✓
AR Situational Awareness	✓		✓	✓		✓		✓	✓	✓		✓	✓
Damaged Infrastructure/ Blocked Roads Recognition	✓			✓							✓	✓
Pest Monitoring	✓															✓
24/7 Aerial Monitoring	✓	✓					✓
Social Media Sensing	✓					✓			✓	✓		✓
Prediction	✓		✓							✓	✓
DSS	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓		✓
Control		✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓
MPS	✓
Routing for a set of locations	✓		✓	✓	✓	✓	✓		✓		✓			✓	✓	✓
Trajectory De-confliction	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓		✓	✓	✓
Autonomous Landing	✓								✓					✓
Communication
Communication Relay	✓	✓				✓	✓	✓
Crowd Counting	✓
Connectivity and Data Offloading	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓			✓		✓
Bandwidth Management and Efficiency	✓		✓			✓	✓	✓						✓
Position Accuracy	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓		✓
Recharging	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓
Transparent Funding	✓
Rescue
Package Drop	✓	✓													✓
Symptom Monitoring and Medical Care	✓												✓	✓	✓
Evacuation	✓		✓								✓		✓
Drone Extinguishing	✓
Risk Status Monitoring	✓		✓			✓		✓	✓	✓		✓	✓
Pest Control	✓															✓
Physical Manipulation	✓		✓	✓		✓	✓
Sampling	✓								✓
VR Training	✓		✓
Reconstruction	✓
Warehouse Inventory Management	✓

in the Appendix A demonstrates a comprehensive juxtaposition of AIRDROP with the aforementioned relevant projects, demonstrating its broader range of disaster aspects and holistic approach across all disaster stages, distinguishing it from its predecessors.

Recent research emphasizes the necessity of a strategy for utilizing technology across different HL operations and deployment phases to achieve real benefits [45]. Additionally, there is a growing body of research assessing the complementarity of drones with other digital technologies. However, comprehensive strategies that integrate these technologies across different HL operations and deployment phases are still lacking [5,6,16]. AIRDROP addresses these needs by providing a comprehensive platform that supports both pre- and post-event phases, ensuring continuity and effectiveness throughout the entire disaster management cycle across diverse disaster scenarios. In summary, the selection of AIRDROP as the focus of this research is driven by its potential to serve as a prototype for future disaster management systems, offering a scalable and integrated solution that addresses current gaps in both theoretical and practical aspects of HL.

3. AIRDROP Overview

AIRDROP is comprised of sixteen (16) services, which are described below.

3.1. Control System

The Control System (Figure 1) encompasses the Command Center Web Application (CCWA) and First Responder Mobile Application (FRMA). On the one hand, the function of CCWA is to facilitate mission planning, management, and supervision. It integrates multiple source data into an aggregated operational picture to be used by a mission commander with the authorization of making mission-relevant decisions. It is responsible for managing drone missions, first responder task assignments, and incident reports for the provision of actionable business insights. It provides real-time information on vehicles, first responders, and critical facilities (e.g., hospitals, bridges, roads).

On the other hand, FRMA aids first-response units with situation visualization, task assignment, incident reporting, live video streaming, and drone mission requests. Its interface provides tailored information to the first responder team (e.g., firefighters, paramedics, police officers, volunteers, etc.) which uses it each time. First responders can interact via a chat function, ensuring effective communication with operational decision-makers. When a drone mission is requested via the FRMA, it needs to be approved by the CCWA-operator before it is executed by the drone.

3.2. Mission Planning System (MPS)

MPS plays a vital role in the AIRDROP platform by processing missions initiated through CCWA or FRMA. It effectively utilizes information from the mission-relevant components to compute highly detailed flight plans, subject to approval by the CCWA-operator. Subsequently, the MPS verifies de-conflicted trajectories and transmits them to the ground control station (GCS). The three key subservices of MPS include routing for a set of locations, trajectory de-confliction, and autonomous landing. Notably, MPS prioritizes drone locations based on specific needs and available resources, effectively preventing conflicts in 4D trajectories (where each point represents UAV presence at a specific 3D location and time). Similarly, this de-confliction process applies to UGVs and USVs. Moreover, MPS ensures safety by enforcing geo-fencing restrictions. Furthermore, it empowers UAVs to perform precise landings on designated targets through the interpretation of visual cues using ML algorithms. Finally, MPS enables emergency autonomous flight and the removal of geo-fence restrictions when necessary.

3.3. Area Scanning System

The area scanning system is a critical AIRDROP service, leveraging sensors and cameras on UAVs, UGVs, and USVs for specialized disaster monitoring. In fire outbreaks, UAVs and UGVs, equipped with ML-assisted electro-optical and infrared cameras as well as dedicated gas sensors, swiftly detect fire, smoke, and gas leakage (CO and CO₂).

For SaR operations, real-time video analytics on these platforms locate missing persons or vehicles, while acoustic sensors identify human sounds. Miniature ground vehicles (MGVs) with chemical sensors complement UAVs and UGVs in detecting human presence through the detection of specific organic compounds found in breath and body odors. In the event of structural collapse, ground-penetrating radar (GPR) sensors on UGVs locate survivors buried under rubble. Furthermore, UAVs and UGVs, with ML-enhanced electro-optical and infrared cameras, identify locations (e.g., voids) where victims may be trapped. The presence of live cables covered with debris and electrical devices that can remain energized in spite of power being turned off due to poor wiring poses a substantial electrocution hazard for first responders. For that reason, UGVs are equipped with electromagnetic field (EMF) sensors designed to identify high voltages.

In floods, UAVs and USVs work synergistically; UAVs capture aerial images to identify obstacles, landmarks, and victims, while USVs provide surface information via acoustic sensors for route re-planning.

In nuclear and industrial disasters, UAVs and UGVs equipped with specialized sensors detect hazardous chemicals and conduct radiological mapping.

For infestation disasters, the system facilitates UAV-intensive searches using hyper-spectral cameras and real-time video analytics to assess vegetation size and detect pest infestations.

It is important to monitor and record weather parameters on every flight in order to achieve a successful area scanning mission. Therefore, UAVs are equipped with weather sensors to aid mission planning.

AR devices are given to first responders to enhance their situational awareness by visualizing important data from the above-mentioned area scanning subservices.

Last, a tethered “Mothership” UAV acts as a continuous “Eye in the Sky”, equipped with high-definition cameras, flood lights, megaphone, and Wi-Fi access for seamless data transmission via optical fiber to the GCS at disaster zones.

3.4. Communication System

The communication system within the AIRDROP platform encompasses various critical aspects, including robust security measures, communication links, communication relaying, crowd counting, bandwidth management, network connectivity, and position accuracy. It employs robust security measures to safeguard sensitive data and establish secure channels between UAVs, UGVs, USVs, and the GCS to prevent unauthorized access and data breaches. It prioritizes command and control (C2) messages over telemetry messages while maintaining low latency between vehicles and the GCS. The system includes communication relaying capabilities, both tethered and non-tethered, to extend communication range and coverage when direct links are not feasible. Additionally, it enables crowd counting services through RSSI (received signal strength indicator)-based techniques and Wi-Fi connectivity, allowing for the estimation of crowd sizes and densities in affected areas, with a view to providing insights for resource allocation and emergency planning. With redundancy in data storage and bandwidth management strategies, such as data compression, the communication system ensures reliable and optimized data transfer. Integration of Wi-Fi access points, connectivity interfaces, and cloud server support fosters seamless information sharing and real-time data exchange, enhancing situational awareness and collaborative decision-making. Leveraging multiple GNSS (global navigation satellite system) receivers, RTK (real-time kinematic) GPS technology, and multi-frequency receivers, the system ensures high position accuracy for vehicles and first responders, thereby facilitating precise localization and mapping of affected areas for effective response coordination and mission planning.

3.5. Recharging

AIRDROP utilizes far-field wireless power transfer (WPT), employing specialized transmitter UAVs (tUAVs) equipped with multiple input–multiple output (MIMO) antennas to wirelessly transfer power to receiver UAVs (rUAVs) during missions. The tUAVs optimize energy transfer gain by adjusting their distance to the rUAVs, while MIMO antennas enhance energy reception by focusing the energy beam. Additionally, separate charging stations for UAVs, UGVs, and USVs are provided, enabling autonomous navigation and docking for recharging.

3.6. Social Media Sensing

This service incorporates social media sensing, where real-time observations from online users on various social media platforms are collected to gather information about the disaster status. The selected social media platforms encompass blogging social media, such as Twitter, facilitating information sharing, social networking media, such as Facebook and Instagram, aimed at connecting individuals, and content sharing media, such as YouTube. This service also includes location extraction functionalities, enabling the identification of geospatial information from social media sensed data to enhance disaster situational awareness and response coordination.

3.7. Prediction

The prediction service entails the utilization of data acquired through the area scanning and social media services to predict disaster outcomes with the use of ML algorithms. Data from relevant centers and councils are also meant to be utilized for wildfire, weather, hurricane, tsunami, flood, and public health emergency prediction. For disaster demand forecasting, ML shall initially combine historical information to predict resource and funding needs, complementing data from the aforementioned services. As more information becomes available, resource demand and specific requirements can be accurately calculated and prioritized, enabling decision-makers to allocate and deploy resources efficiently and effectively.

3.8. Rescue

The rescue service incorporates various functionalities to optimize resource allocation, provide medical assistance, monitor disease symptoms for epidemics, aid evacuation and fire extinguishing, and ensure the well-being of first responders during their operations.

This service facilitates optimal distribution of resources by the UAVs, which can deliver essential supplies and aid packages to areas of interest (AoIs) without landing or upon landing via parachute. For medical care during emergencies, specialized UAVs are equipped with an electrocardiogram (ECG), automated external defibrillator (AED), and temperature and respiratory sensors. These UAVs analyze patients’ conditions before the arrival of ambulances. Moreover, life-saving vaccines can be delivered through cold transport drones. This service also employs UAVs equipped with thermal and visual image sensors to remotely detect disease symptoms, including respiratory infections, through thermography and machine vision algorithms. This enables effective epidemic monitoring and prevention.

Moreover, UAVs assist in evacuation efforts using megaphones to guide and facilitate the evacuation of stranded individuals within AoIs. In extinguishing operations, water-based fire extinguishing agents are deployed by drones to combat forest or building fires.

Lastly, an indispensable facet of this service involves the surveillance of first responders’ well-being. UAVs gather crucial data from wearable sensors affixed to their uniforms, encompassing GPS location, temperature, radiation levels, hazardous chemicals, and live video feed. These data play a vital role in identifying first responders in need of immediate assistance and expediting timely support during disaster scenarios.

3.9. Physical Manipulation

This service entails a UGV equipped with a seven-degree-of-freedom (DoF) arm featuring a specialized three-finger gripper designed to perform tasks, such as opening doors, toggling lights, controlling elevators, and deploying equipment in the disaster field. Moreover, the UGV incorporates foldable 5-DoF robotized legs with wheels, enabling it to maneuver through rolling, walking, crawling, and stair climbing to overcome obstacles in the field.

Collaboration between UAVs and UGVs is also possible as the two platforms can work symbiotically for the execution of tasks (e.g., close valves) in disaster-affected industrial environments. The UAV transports the UGV, thereby enabling it to overcome challenging obstacles. Conversely, the UAV can perch on the UGV to conserve battery life, leveraging the UGV’s mobility to approach target in close proximity.

3.10. Sampling

This service is inspired by the work presented in [46] and involves the collection and examination of water samples to determine their drinkability or potential hazardous substances. The process begins with the UGV autonomously navigating towards the point of interest (PoI) using stereo vision and dexterous manipulation to clear any debris blocking its path. If the UGV encounters space constraints, a UAV is deployed from its base platform on the UGV. The UAV autonomously navigates through the debris, identifies the PoI using ML algorithms, and collects the sample, which is then returned to the UGV. The UGV utilizes an on-board handheld chemical sensor to analyze the sample’s chemical composition. If the UAV cannot access the PoI due to limited space, a human operator remotely deploys an MGV through the UAV as a relay. The MGV collects the sample and returns to the UAV, which then brings the MGV back to the UGV.

3.11. Pest Control

This service includes a UAV equipped with specialized nozzles intended for the controlled dispersion of biological and chemical pesticides. This solution prevents humans from venturing into potentially dangerous, remote areas. The timely detection and intervention facilitated by the UAV’s capabilities allow for early treatment of pests while their population is still limited, consequently minimizing the quantity of environmentally harmful pesticides required for effective pest control.

3.12. Transparent Funding

The integration and embedding of a portal within humanitarian agencies’ website interfaces offer a means to facilitate data sharing with external organizations. The proposed portal serves as a tool for enabling donations and external funding, with the support of blockchain technology, which contributes to the prevention of fraud and corruption. By employing blockchain, donors and external organizations gain the ability to trace the utilization of their contributions to disaster relief efforts, introducing elements of visibility, traceability, and transparency to all transactional activities. Consequently, this enhanced accountability fosters improved stewardship and trust in the disaster relief process.

3.13. Reconstruction

This service pertains to the recovery phase, where concerted efforts are undertaken to address the aftermath of the disaster by restoring the affected regions’ living conditions through the reconstruction and rehabilitation of damaged infrastructure.

3.14. VR Training

This service encompasses an integrated approach employing 3D VR, cloud computing, UAVs, UGVs and USVs. The data sets derived from observations made by these robotic systems are utilized to construct a comprehensive and lifelike VR environment, accurately representing real-world conditions. This facilitates the development of rescue scenarios within a feasible timeframe to enhance first responder preparedness and response efficiency.

3.15. Warehouse Inventory Management

This service pertains to the utilization of drones for inventory management of emergency relief warehouses. It encompasses essential tasks, such as stock taking, cycle counting, locating missing items, and retrieving hard-to-reach stocks, while also conducting warehouse capacity calculations. The system incorporates a blockchain and distributed ledger to store specific inventory data acquired by UAVs, ensuring their validation, trustworthiness, and accessibility to relevant stakeholders.

3.16. Decision Support System (DSS)

The DSS serves as a central hub, capable of integrating and exchanging data with the aforementioned services (Figure 2). It includes a GIS (geographic information system)-based system for geospatial analysis and can determine danger zones and hazard positions, calculate uncertainty for all its results, optimize sensor positions, conduct scenario planning, and evaluate performance and progress. Moreover, it can validate social media sensed data and predictions, determine optimal resource allocation, and provide insights for reconstruction, communication, and pest control efforts.

4. Research Design

Research design is the overarching framework that guides the implementation of the research, encompassing various layers, including the research approach, strategy, choice, techniques and procedures [47]. A research approach is defined as the path of conscious scientific reasoning [48]. The research approach adopted in this paper aligns with abductive reasoning, i.e., a synthesis of the deductive and inductive approach. While deductive reasoning commences with established theories or hypotheses, testing them against specific instances [49], and inductive reasoning builds general theories from specific observations [50], abductive reasoning navigates between these poles [51], acknowledging the necessity of theory to guide observation and the critical role of observation in refining theory. The abductive research approach starts with a real-life phenomenon and observation. Still, previous theoretical knowledge is significant, even if it is not able to explain the phenomenon [52,53]. Hence, the researcher initiates a creative iterative process [54], engaging in a continuous cycle of hypothesis generation, observation, and analysis.

In the case of AIRDROP, this process begins with the observation of significant research gaps within the current state of drone-enabled HL, highlighting the need for a more integrated solution. Recognizing the limitations of existing theoretical frameworks to fully address these challenges, the research leverages previous theoretical knowledge as a foundation, while acknowledging its inadequacy in offering a complete solution, setting the stage for an exploratory process, where hypotheses about the potential for more holistic integration and operationalization of drones and emerging digital technologies within disaster management are continuously generated, tested, and refined. This cycle of hypothesis generation, observation, and analysis, where the collection of empirical data and the building of theoretical frameworks occur simultaneously, in a reinforcing loop [52,54], allows for the dynamic development of the AIRDROP platform, ensuring that each iteration is informed by empirical data and closer to addressing the real-world needs identified.

The research strategy introduces the overall direction, plan or scheme of research [55]. The categorization of research strategies varies significantly in the literature; Robson [56] proposed experiment, survey and case study, while Yin [57] added history and archival analysis to these three strategies. Saunders et al. [58] further diversified the spectrum by introducing action research, grounded theory and ethnography. On the other hand, Bryman and Bell and Kumar [47,59] categorized research strategies solely on the basis of qualitative and quantitative paradigms. Though relatively newer in the context of management research [60,61,62] and widely used in the fields of information and technology research [63,64], DSR has quickly established itself as a pivotal strategy for developing and evaluating artifacts—ranging from models and methods to constructs and design theories—that aim to solve practical problems [65].

The intersectionality of this research, which resides at the confluence of management research and information and technology research, necessitates an approach that is both flexible and robust. This interdisciplinary nature is driven by the unique challenges and RQs posed, which seek to explore and resolve real-world problems through the lens of both managerial implications and technological innovation. The integration of DSR methodology—especially its first three steps [66]—alongside survey research is, therefore, a deliberate and strategic choice, allowing for an exploration that not only leverages DSR’s strengths in artifact creation and evaluation but also capitalizes on the empirical insights afforded by survey research.

The initial step involves problem identification, where the research gaps are defined and the foundation for the study’s contributions is established. The research starts with an exploratory phase, wherein qualitative insights obtained from literature review and consultations with experts establish a foundation for comprehending the intricate requirements that AIRDROP ought to encompass. While this approach incorporates elements reminiscent of the case study strategy [57,67], such as engaging with experts and analyzing the literature to grasp particular requirements, it diverges from adhering strictly to the conventional case study framework.

The second step, objective definition, builds on the identified problems to outline specific goals for the research. Even though decision-making in requirements engineering has traditionally relied on stakeholders’ intuition and experience as well as rational schemes, the recent rise of data-driven approaches [68] puts practitioners under strain to integrate quantitative data in order to automatically decide what requirements should be added or removed. Several methods can be used to validate requirements, among which questionnaire surveys stand as an effective approach. In general, surveys play a pivotal role in presenting researchers with valuable insights into the specific challenges and characteristics of humanitarian contexts; as such, methodologies are able to offer a more profound understanding of the complex dynamics and requirements within them [69]. Nevertheless, it is remarkable that survey research has not been thoroughly utilized in studies focused on HL [70,71] despite their dominance in commercial logistics studies [72]. Thus, to advance knowledge in the HL domain, there is an increasing need for the adoption of surveys [73], with the aim of contributing to the development of more effective and responsive solutions, such as the AIRDROP platform. This confirmatory approach facilitated the efficient collection of data from a diverse population, providing a statistical foundation to measure the extent to which the platform’s features align with user needs and expectations.

Many scholars view quantitative and qualitative research as having different foundations, particularly in three key areas: principal orientation to the role of theory in relation to research, epistemological orientation and ontological orientation [47]. Qualitative research emphasizes perceptions rather than quantification in the collection and analysis of data and utilizes an inductive approach to the relationship between theory and research. Conversely, quantitative research emphasizes quantification in the collection and analysis of data that can be statistically analyzed and entails a deductive approach [74]. The research choice for AIRDROP embodies a mixed methods approach, meticulously combining qualitative and quantitative research paradigms to address the RQs. Mixed methods research has strengths that can offset the weakness of adopting one method and provide more evidence in resolving RQs [75].

Last but not least, the research techniques and procedures employed in this study, i.e., the elicitation, specification, and validation of AIRDROP’s requirements, are elaborated upon in the subsequent Section 5.

5. Research Methodology

Elicitation, specification, and validation of requirements are inextricable components of the requirements engineering process and essential determinants of software and system quality [76]. Requirement elicitation is the process of acquiring information from different sources about the system at hand as well as existing pertinent systems from which the requirements are distilled to enhance the understanding of the system to be specified [77]. Requirement specification is the process of translating the information collected during requirement elicitation into a document that defines a set of requirements, describing how the product to be delivered functions and behaves, including user interaction [78]. Requirement validation ensures that the proposed requirements are realistic, consistent, and thorough. Throughout this process, issues in the requirements document are detected and addressed via updates [78]. Additionally, during this stage, requirements are often ranked and prioritized in order to allocate resources effectively and facilitate well-informed decision-making about addressing key aspects of a project and ensuring harmonization with essential objectives.

In the context of AIRDROP, the role of requirements engineering encompasses the elicitation and specification of requirements, involving the careful consideration of how to skillfully translate the broad array of the above-described services into particular needs, followed by the validation of these needs through a meticulous questionnaire analysis.

5.1. Requirement Elicitation and Specification for the AIRDROP Platform

In this section, the initial stages of the elicitation and specification of requirements are tackled, which act as prerequisites, forming the path for our extensive analysis and investigation in the validation phase. Τhis process is crucial for laying the groundwork for our analysis, directly contributing to answering RQ1 by gathering and documenting the wide array of AIRDROP services into specific requirements/needs.

According to the work in [79], in requirements engineering, there is a multitude of elicitation and specification techniques, which are often used in a hybrid manner. In this research, during the elicitation process, the investigation of pertinent systems from the existing literature (e.g., case studies, academic outlets, project documentations), including their requirements documents, when available, played a role of paramount importance. In fact, the reuse of existing requirement artifacts makes requirements engineering more prescriptive and systematic [80]. Moreover, useful insights were provided through consultations with a leading drone technology integrator company (https://altus-lsa.com/, accessed on 20 May 2024), notably in identifying and cross-checking technical requirements. These efforts provided critical inputs that were used to elicit the AIRDROP requirements and subsequently exploited in the requirement specification process. This involved considering infrastructure limitations, geographical challenges, communication barriers, resource constraints, technological compatibility, logistical challenges and more.

As seen in

Table 1. Summary of the requirements defined with respect to each category.

Category of Requirement	# of Requirements Defined
AIRDROP	6
Drones	30
Control System	27
Mission Planning System	21
Area Scanning System	29
Communication System	29
Recharging	19
Social Media Sensing	15
Rescue	29
Prediction	7
Decision Support System	32
Physical Manipulation	5
Sampling	9
Pest Control	5
Transparent Funding	12
Reconstruction	6
Virtual Reality Training	15
Warehouse Inventory Management	12
User Interface	27
Total	335

, the process of requirement elicitation and specification resulted in the identification and documentation of a total of 335 requirements, thoughtfully allocated across the spectrum of 19 services. Notably, three supplementary services were included, namely general AIRDROP requirements, drone specifications encompassing the distinct requirements of UAVs, UGVs (comprising regular UGVs, heavy-lift UGVs, and MGVs), USVs, and user interface (UI) requirements. This first category was introduced to capture overarching prerequisites pertaining to the entire AIRDROP platform, including foundational elements, essential for its effective operation. Given the diverse range of unmanned vehicles utilized by the platform, a dedicated category was established to articulate the distinctive requirements of each vehicle type. Finally, the UI category was introduced to delineate the specific requirements regarding the design, usability, and accessibility of the platform’s UIs.

AIRDROP’s requirements include functional and non-functional categories. The former pertain to the specific capabilities, functionalities, or tasks that the platform must perform to accommodate user needs and accomplish desired objectives, as opposed to the latter, which focus on the system’s qualitative characteristics and behavior beyond basic functionalities. This distinction is not only dominant within research but also exceptionally influences how requirements are managed in practical implementations. Still, it is argued that most non-functional requirements, including those described in the present research, can be approached similarly to functional requirements [81]. AIRDROP’s functional requirements include data, technical, and operational requirements, whereas its non-functional requirements include reliability, security, performance, compliance, usability, scalability, compatibility, and localization requirements.

Regarding functional requirements, data requirements relate to the platform’s data management and processing capabilities. Data collection, storage, retrieval, processing, and integration with other systems or sources are all included. The underlying technical infrastructure required to provide AIRDROP’s functionality is the subject of technical requirements, encompassing hardware, software, networking, and any third-party tools or services needed for the platform to function properly. Operational requirements include explicit criteria that govern the platform’s functional capabilities and practical limits required for it to operate effectively. With regard to non-functional requirements, reliability requirements refer to the platform’s capacity to fulfill its duties devoid of errors or disruptions on a constant basis. It addresses issues such as system uptime, fault tolerance, and error management. Security requirements dictate that the platform and its data must be protected from illegal access, breaches, and vulnerabilities, including features, such as user authentication, data encryption, access restrictions, and adherence to applicable security standards. The platform’s speed, responsiveness, and efficiency are addressed through performance requirements. Compliance requirements ensure that the platform adheres to applicable regulations and standards, including data protection rules, privacy legislation or best practices in the industry. Usability requirements focus on making the platform as user-friendly as possible, including elements, such as UI design, navigation, and overall user experience. Scalability requirements guarantee that the platform can manage the growing workload and user demand over time without experiencing substantial performance deterioration. It takes into account both vertical and horizontal scaling. The platform’s ability to function seamlessly with diverse devices, operating systems, and other software components is addressed through compatibility requirements. Finally, the specific criteria that AIRDROP must meet in order to accurately determine the precise geographic location of its entities, ensuring a predefined level of accuracy, are referred to as localization requirements.

In Figure 3, the distribution of AIRDROP’s requirements is visualized, illustrating the composition of functional and non-functional aspects, along with their respective categories. It is pertinent to acknowledge that the categorization of requirements is inherently subjective, potentially entailing instances of overlapping. Nevertheless, the purpose of this classification is to provide a basic understanding of the multifarious spectrum of AIRDROP requirements, allowing for a realistic appraisal of their relevance and significance in relation to their intended aims. The subsequent section delves into the process of validating the identified requirements.

5.2. Requirement Validation for the AIRDROP Platform

As we transition to addressing RQ2, which seeks to understand how the relevance and adequacy of the identified requirements for AIRDROP can be ensured within the complex dynamics of technology-enabled HL, we employed survey research using Google Forms. Google Forms is a popular, broadly used online survey platform, which offers multiple advantages, including its user-friendly interface, easy access, seamless sharing options, data management, and real-time data analysis, as well as its notable benefit of being cost-free, rendering it an excellent choice for gathering information and feedback from respondents. The questionnaire used for the validation of the requirements of the AIRDROP platform is provided in Supplementary Material File S1. In the next subsections, more detail is provided concerning the design of the questionnaire, the data collection process, sampling, and participant recruitment, and lastly, the analysis of the questionnaire’s results.

5.2.1. Questionnaire Design

Structure and Content 

The questionnaire designed for the validation of the AIRDROP platform is divided into 21 parts. The first part is introductory, and its aim is to gather respondent information relevant to their occupation (industry), their knowledge, expertise and background with regard to several domains, including logistics and/or humanitarian and disaster relief operations, their experience level in using similar platforms and/or technologies, and the challenges present in the existing solutions they are currently using or considering to use. The next 19 parts refer to the requirement validation, 1 for each category of requirements. Finally, the final part—comprised of three open-ended questions—works as an invitation for general feedback, suggestions and concerns pertinent to the platform and its requirements.

Requirement Ranking through the Importance Ranking Scale 

The significance and impact of each requirement of the AIRDROP platform must be clearly understood in order to evaluate it successfully. A thorough importance ranking scale has been created in order to do this. This scale seeks to give respondents a methodical way to assess the importance of particular requirements in light of their knowledge and experience. The importance ranking scale provided in the questionnaire ranges from 1 to 5. Each number on the scale carries a specific meaning, as described below:

• Score 1: A score of 1 indicates that the requirement is deemed unimportant, meaning that its inclusion or exclusion does not have any effect on the solution. In other words, the solution can function equally well with or without this requirement.
• Score 2: A score of 2 suggests that the requirement is of low importance. While it is not essential for the core functionality of the solution, its inclusion can provide additional value and enhance user satisfaction.
• Score 3: A score of 3 signifies an important requirement. Without this requirement, the solution’s functionality may be limited or only partially useful. Its inclusion significantly improves the effectiveness and usability of the solution.
• Score 4: A score of 4 indicates a serious requirement. If this requirement is not fulfilled, the solution remains usable but may lack certain crucial features or capabilities. Its inclusion is paramount for enhancing the overall usefulness of the solution.
• Score 5: A score of 5 represents a critical requirement. Without meeting this requirement, the solution becomes completely ineffective and unusable. Its inclusion is absolutely necessary for the solution to be considered successful.

In addition to the five-point scale, a “Not Sure” (N/S) category is also provided to accommodate respondents who may have uncertainty regarding the importance of a particular requirement. This option allows participants to indicate their lack of confidence in assessing the importance of a requirement.

5.2.2. Data Collection

The questionnaire was accessible to responders for a period of two months, i.e., from the 1st of July to the 31st of August 2023. Particularly in studies based on online surveys, choosing a suitable data collection timeframe is an essential component of the study methodology. The two-month period of time was chosen to guarantee an adequate number of responses. The sample size is a crucial factor in survey-based research for providing statistically significant results. The AIRDROP questionnaire, due to its length resulting from the wide range of HL topics covered, demanded a large time and energy commitment from respondents, which may have reduced the number of prospective responders. In order to prioritize the quality and depth of responses over the number of participants, the research team decided to select a relatively small sample size of 20–30 people. As a general guideline, researchers are urged to identify the minimum necessary sample size according to the particular research objectives [82,83,84,85,86]. The selected sample size is also aligned with a similar study that validated requirements for a comparable HL robot-assisted platform, which demonstrated the suitability of 24 participants to obtain diverse and in-depth insights [87].

Overall, there were a total of 26 responses to the AIRDROP survey during the course of the two-month data gathering period, a response rate deemed satisfactory, falling within the expected range. Reaching this goal sample size allowed the study to collect a diversified dataset, encompassing a wide variety of nuanced viewpoints from respondents with varying backgrounds and roles, thus ameliorating the findings’ representativeness and bolstering their overall robustness.

Several measures were taken during the development of the questionnaire and data collection, with a view to ensuring the validity of the data collected. First, the participants were informed of the confidentiality of their responses, thus creating a safe environment for them to express themselves. Second, a pre-test was carried out with a small group of individuals (4) to collect feedback on the questionnaire’s overall coherence and comprehensibility. This step allowed for the identification of any potential ambivalences or intelligibility issues, making sure that the questionnaire was easy to understand for all participants. After this preliminary test, the team decided to limit the technical jargon to the minimum extent possible and rephrase or eliminate several requirements, deemed confusing or ambiguous. Moreover, clear instructions were given to guide the respondents on how to complete the questionnaire accurately and effectively.

5.2.3. Sampling and Participant Recruitment

The questionnaire did not target a particular group of professionals; instead, it sought to gather the opinions of academics, practitioners, and technical experts from a broad range of logistics- and technology-related fields (e.g., robotics, information technology, telecommunications and networking, data science, and ML, etc.). Hence, rather than selecting random sampling, a multimodal recruiting method was employed to reach a diverse and representative sample of participants. A chosen mailing list of experts received emails asking them to complete the online questionnaire. Requests were also made via social media (Facebook and LinkedIn). According to the roadmap proposed in [88] for higher research quality in humanitarian operations, the inclusion of many stakeholder groups enables researchers to present a more comprehensive and unbiased view of the research topic. In the case of AIRDROP, scholars, as well as representatives of non-profit organizations (NGOs), commercial logistics firms, other private sector businesses, disaster management organizations, and other national government agencies, including emergency responders, were among those who responded.

5.2.4. Questionnaire Analysis

Descriptive statistics and non-parametric tests were employed to analyze the collected data via Microsoft Excel and IBM SPSS Statistics 20 (Statistical Package for the Social Sciences) software. Combining these tools has been proved suitable and reliable for analyzing data gathered through questionnaires, including the application of descriptive statistics and nonparametric tests.

Descriptive Statistics 

With the aim of gaining valuable insights from the responses and providing a comprehensive analysis, descriptive statistics were employed to summarize and understand the data. For starters, descriptive statistics were used for the introductory part of the questionnaire and included visualizations to showcase the distribution of professional backgrounds among respondents along with their years of experience. This not only provided valuable context for interpreting the subsequent data but also laid a foundation for the execution of non-parametric tests. Additionally, their expertise in different disaster management as well as general domains was also visualized to better comprehend their background and knowledge, thus enabling a clearer insight into their views and potential influence on the requirement prioritization process.

The process of requirement prioritization plays a crucial role in assessing the results of the questionnaire. By assigning priorities to the identified requirements, valuable information can be gained, with regard to the needs and preferences of potential stakeholders, who, through their responses, shall contribute to pinpointing the most important features and functionalities. This is a structured approach to decision-making, ensuring that the most vital requirements are given the highest priority and receive the necessary attention during the next implementation stages, while the less important ones are postponed or even ignored.

Requirement prioritization techniques range from simple and qualitative procedures to advanced analytic prioritization approaches that fall under the genre of optimization algorithms [24]. For instance, the application of the Kano model involves the assessment of participant contentment and discontentment to classify requirements into distinct scoring groups [89,90]. Amongst quantitative techniques, simple ranking involves the meticulous ranking of requirements based on predefined criteria [91], while in more elaborate approaches, such as the analytic hierarchy process (AHP) and the prioritization matrix, numerous criteria are integrated, such as effort, value, and risk, to facilitate requirement prioritization [92]. Employing computational techniques in conjunction with expert insights, search-based software engineering (SBSE) acts as an valuable tool, effectuating the automation of requirement selection and providing dynamic assistance across the software engineering lifecycle [93]. Another common prioritization strategy in requirement prioritization, selected to be employed for the prioritization of the AIRDROP requirements, is the MoSCoW methodology [94]. This strategy aids in classifying requirements according to their urgency and priority by providing a simple and easy-to-understand framework, which is particularly well-suited for the case of AIRDROP due to its ability to handle a large volume of requirements. “MoSCoW” is an acronym for:

• MUST: These are the requirements that, in order for the solution to be declared successful, must be fulfilled. In this research, requirements with scores between 5 and 4 were assigned in this category.
• SHOULD: These requirements need to be met as closely as possible due to their critical nature. Although not essential, like the MUST requirements, their inclusion considerably increases the project’s overall value and effectiveness. In this research, requirements with scores between 4 and 3 were assigned in this category.
• COULD: These requirements are desirable but not obligatory. They encompass additional features or capabilities that could be taken into account if time and resources allow. These requirements may be deprioritized or deferred to a later stage. In this research, requirements with scores between 3 and 2 were assigned in this category.
• WOULD: These requirements are optional and may be considered for future improvements or iterations. Since they hold the lowest priority, they can be delayed or disregarded without adversely affecting the project’s essential functions. In this research, requirements with scores between 2 and 1 were assigned in this category.

The ratings provided by the respondents serve as valuable inputs for the prioritization process, allowing for a comprehensive understanding of the relative importance and urgency of each requirement. The average score for each requirement was calculated, and then the MoSCoW technique was applied to categorize them into Must, Should, Could, and Would categories. This approach ensures that the most critical and essential requirements are identified and given the highest priority, while still considering the importance of other requirements.

Non-Parametric Tests 

The Kruskal–Wallis H test for One-way Analysis of Variance (ANOVA) by Ranks is a rank-based nonparametric test that is used to determine statistically significant differences between groups of responders (for three of more groups) [95]. It is an omnibus statistical test and can only indicate that at least two groups are statistically different, without disclosing which ones [96]. In the present research, given the non-parametric nature of the data and the ordinal scale of response options ranging from 1 to 5, this test was used for multiple-group comparative inferential analysis, with a view to evaluating whether there were statistically significant differences in the respondents’ perceptions of the importance of the requirements across their different professional backgrounds. The test enabled a thorough understanding of how their backgrounds may influence their perspectives and priorities in the context of the features and functionalities of the AIRDROP platform. A number of Mann–Whitney U tests were run as post hoc analyses in cases where significant differences existed among participant groupings, with the aim of identifying the specific groups where these differences are present.

Managing Participants’ Uncertainty 

In the dataset, responses denoted as “Not Sure” by participants were handled as missing values by SPSS during statistical analysis. Consequently, during the analytical process, these responses were excluded from the calculations pertaining to the corresponding requirements. By managing these responses as missing data, the acknowledgement of respondents’ genuine lack of knowledge regarding certain requirements was ensured, thereby preventing any undue influence on our analysis. However, excluding “Not Sure” responses from the statistical analysis, while helping streamline our data processing, may have also limited our insight into areas of participant uncertainty regarding specific requirements. This decision was made with the aim of simplifying our analysis, but we recognize it as a deliberate trade-off rather than an oversight.

Analysis of Open-Ended Responses 

The open-ended questions in the questionnaire were intended to capture any critical requirements that were not included in the predetermined list and solicit participants’ familiarity with such platforms as well as their general opinions and concerns. Because such responses were limited, qualitative analysis software was not required. Instead, these replies were manually scrutinized to unravel any recurrent themes, important insights, or potential new requirements developed from the participants’ feedback. In this way, more complete knowledge of the participants’ opinions was provided, and the overall analysis of the requirements was supplemented.

6. Results

6.1. Descriptive Analysis

6.1.1. Distribution of Participants’ Professional Backgrounds and Experience Levels by Year Range

As shown in Figure 4a, the majority of participants are academics (34.62%), including researchers, scholars, educators, etc., followed by technology experts (23.08%), i.e., professionals familiar with drones, robotics, sensor systems, data analysis, software development, ML, etc. We also had participation from end-users (19.23%), i.e., first responders, including firefighters, paramedics, police officers, volunteers, etc.; disaster management professionals (7.69%), e.g., professionals working in government agencies, NGOs, etc.; as well as stakeholders (7.69%), e.g., representatives of previously affected communities, community leaders, etc. The least represented category is comprised of subject matter experts (3.85%) (i.e., professionals with in-depth knowledge and expertise in specific areas related to disaster management). Moreover, one participant selected the “Other” option to describe their affiliation and specified a background in supply chain management with limited experience in HL drones.

Most participants have experience within the 5 to 10 years range (N = 8), followed by those with more than 10 years (N = 7), those from 3 to 5 years (N = 7) and those from 1 to 3 years (N = 4). Figure 4b sheds light on the distribution of each year range throughout the seven professional backgrounds groups discussed, revealing that subject matter experts, technology experts and stakeholders tend to have greater levels of experience, which is a reasonable observation.

Before proceeding to the next section, it is worth noting that professional background categories are mutually exclusive, which means that a participant can only belong to one category at a time. Furthermore, there are no repeated measures on the subjects and each response within each group is deemed independent, which means that it is neither affected nor reliant on the responses of other participants.

6.1.2. Participants’ Expertise in Different Disaster Management or General Domains

The participants were asked to evaluate their level of familiarity with several disaster management and general domains on a scale ranging from 1 (Not Familiar) to 3 (Very Familiar). As anticipated, the distribution appears to be more consistent across the different general domains (Figure 5a) when compared to the disaster management domains (Figure 5b). In fact, the graphs in Figure 5a show a significant overlap, with data and technology emerging as the most familiar domain among participants, closely followed by management and leadership, risk and compliance and logistics and supply chain; this is understandable given that general domains apply to a broader spectrum of professions and thus are expected to be more evenly distributed.

On the other hand, disaster management domains are specialized and may not be as widely understood by those without direct knowledge and expertise. Upon analyzing the graph illustrated in Figure 5b, it seems that subject matter experts are the most familiar with disaster mitigation and preparedness as well as disaster response and recovery, i.e., all four disaster stages, which is a highly expected result given their in-depth knowledge and expertise in specific areas related to disaster management. Stakeholders indicate the highest level of familiarity with disaster management strategies and frameworks, which corresponds to their responsibilities and involvement in disaster management activities, including policymaking, decision-making and strategy development. As expected, technology experts are the most familiar with data analysis, technology, and communication in disasters. Last but not least, disaster management professionals ranked a relatively high familiarity score across all disaster management domains with no specific domain standing out, which can be attributed to their well-rounded capability to address a wide variety of challenges within the field of disaster management and HL.

6.1.3. MoSCoW Requirement Prioritization

The results of the requirement prioritization for all AIRDROP services after applying the MoSCoW methodology to participants’ rankings are provided as a table (Table S1) included in the Supplementary Material File S2. Table S1 includes several key metrics: mean value, standard deviation, the percentage of missing values (“Not Sure” answers), and the distribution of participants’ replies across the five ranking categories for every requirement. The last two metrics are visually presented in the form of heat maps. Figure S1 (in Supplementary Material File S2) displays the color legends, which are employed to represent the ranking scales and missing value percentages within these heat maps.

Most requirements ended up being ranked as “MUST” (N = 277, 82.69%) and “SHOULD” (N = 57, 17.01%). Only one requirement (N = 1, 0.3%) was ranked as “COULD”. None of the requirements were ranked as “WOULD”.

Five out of the six AIRDROP requirements were ranked as “MUST”, with AIRDROP_003 ranked as a “SHOULD” requirement. This requirement stresses out the need for compatibility with common hardware, and its “SHOULD” ranking may indicate the need for adaptability in case specialized configurations are necessary.

Transitioning to drone requirements, 19 out of 30 were ranked as “MUST”, ten as “SHOULD” and one as “COULD”. A concern about several vehicle technical specifications was expressed by three technological experts in our participant sample, who believe that such requirements might restrict the potential holistic use of the AIRDROP platform across the various missions described. This concern was most likely shared by all participants as it was reflected in the results. For instance, UAV_004, UGV_002, HLUGV_003, MGV_001, and USV_003 were all ranked as “SHOULD”, all addressing maximum and minimum vehicle speeds. The decision to classify them as “SHOULD” acknowledges the diverse nature of disaster response scenarios, where precision and slower operations are sometimes more fitting while in others higher speed capabilities are preferred. The rationale is similar for the drones’ size (UGV_006, USV_006, i.e., the only requirement ranked as “COULD”) or weight (MGV_003, UAV_002, UGV_005, USV_005). All in all, categorizing these requirements as “SHOULD” accommodates mission-specific needs.

In the context of the control service, 20 out of 27 requirements were ranked as “MUST” and seven as “SHOULD”. Participants assigned “SHOULD” rankings to the CCWA_008, CCWA_018, CCWA_019, CCWA_017, and FRMA_006 requirements, acknowledging that functionalities, such as enabling area scanning requests through drawing a polygon on the CCWA map, future timeframe visualizations, 3D map extensions, and chat functions between the CCWA and FRMA may not be as necessary as other requirements. Correspondingly, FRMA_007 and FRMA_008, which offer “Track & Follow” and “Follow path” options to FRMA interface for requesting vehicle missions, were also ranked as “SHOULD” and not deemed as mandatory compared to their operational decision-maker counterparts, which were ranked as “MUST”.

The assessment of the area scanning system service resulted in 26 “MUST” and 3 “SHOULD” requirements. ASCAN_007, which refers to AIRDROP sensors being quickly mountable without tools, might be important in scenarios where rapid deployment is crucial; however, in situations where time is not a critical factor, this requirement could be relaxed. WEA_001, involving equipping the UAV with weather sensors, is important for collecting meteorological data during disaster response. However, in some cases, external weather data sources or ground-based weather stations might provide sufficient information, rendering on-board sensors a “SHOULD” requirement. This statement is further supported by a stakeholder who provided a relevant comment. EMF_001, related to equipping the UGV with an EMF sensor to detect high voltages, is critical in situations where electrical hazards are present. In scenarios without such threats, this requirement may not be as crucial, justifying its classification as “SHOULD”.

Remarkably, all MPS requirements were universally ranked as “MUST”, a reasonable consensus given the criticality of these requirements for successful mission execution, leaving no room for compromise.

In the communication service, 26 requirements were ranked as “MUST” and 3 as “SHOULD”. LINK_004 specifies latency restrictions for telemetry messages, which are intrinsically less important than C2 messages. BME_004 focuses on video streaming bandwidth restrictions, which, while crucial for certain tasks, might have been recognized by participants as less essential for some disaster management operations that do not require extensive video streaming. Similarly, CROW_001, which pertains to RSSI-enabled crowd counting, was categorized as “SHOULD”. An end-user noted the multitude of missing person detection mechanisms already presented by AIRDROP, which led to a lower priority for this particular requirement in the overall system.

Regarding the recharging requirements, 17 out 19 were ranked as “MUST” and 2 as “SHOULD”. ARECH_011 recognizes that requiring dedicated UAV charging infrastructure on UGVs may not be as essential, since alternative charging methods (MIMO recharging configuration and dedicated UAV recharging infrastructure) are available as described by the requirements from ARECH_001 to ARECH_10. As pinpointed by an end-user, mandating UAV battery charging when it drops below 50% state-of-charge (ARECH_013) might not always be practical, especially in time-sensitive missions.

Eight social media sensing requirements were ranked as “MUST” and seven as “SHOULD”, including all location extraction methods (from SOCI_007 to SOCI_012). This ranking emphasizes the need for flexibility in extracting locations from multiple sources for enhanced system robustness. SOCI_013, i.e., real-time social media data pre-processing and de-noising, was also ranked as a “SHOULD” requirement, acknowledging its potential resource-intensive nature while allowing for flexibility in data processing.

Shifting towards the prediction service, four requirements were ranked as “MUST” and three as “SHOULD”. PRE_001 emphasizes the use of social media data for prediction, which might not be feasible or necessary in all cases, depending on data availability and reliability. According to an end-user, AIRDROP already incorporates a plethora of powerful disaster insight mechanisms, and the addition of the less reliable social media data may not provide as significant value. A similar rationale applies to PRE_006 and PRE_007, which involve predicting initial demand patterns with regard to funding as well as evacuation and shelter needs, respectively, using historical data. Such data, according to a subject matter expert, are not important for all hazards. Some disasters may have predictable patterns that can be informed by historical data, while others may not, rendering the use of historical data less critical.

Looking into the DSS service, 26 requirements were ranked as “MUST” and 6 as “SHOULD”. Some requirements, i.e., accessing video streaming (INT_015) and photos (INT_016) through the DSS, the use of Docker images (INT_017) and the ability to conduct communication effectiveness and network traffic analyses (ANA_014) were rated lower in importance. These rankings could be attributed to the complexity and resource-intensive nature of these tasks as well as their limited applicability in certain disaster scenarios. For example, real-time video streaming might be critical in a wildfire response but less so in a flood situation. Additionally, resource constraints, such as limited bandwidth, can affect practicality. Furthermore, in scenarios involving sensitive data, such as medical emergencies, the use of Docker images and network analyses may raise privacy and security concerns. Providing the VR training (INT_012) and pest control (INT_010) services with data was also considered less essential. Unlike the other services, for which data are inherently relevant to the specific disaster scenario, VR training content could potentially be generated from various alternative sources, including historical disaster data, or simulations created explicitly for training purposes. Similarly, information about pest control efforts specifics can be acquired using a combination of methods and data sources, such as environmental sensors, local reports, historical data, and satellite imagery.

MED_002 is the only requirement of the rescue service that was not ranked as “MUST” and was ranked as “SHOULD”. This requirement describes the need for the UAV to use ML algorithms to recognize human actions (e.g., coughing and sneezing). According to a technology expert specialized in AI and ML, this feature might not provide significant advantages but could add unnecessary computational complexity and costs. Additionally, they emphasized the potential risk of false positives, in which the ML algorithms might mistakenly interpret non-threatening motions as coughing or sneezing, leading to unnecessary alerts and potentially diverting resources from more critical tasks.

As far as the sampling service is concerned, six “MUST” and three “SHOULD” requirements were reported. Less emphasis was placed on enabling the teleoperation of the MGV via a UAV relay (SAM_006), enabling the UAV to transport the MGV (SAM_007) as well as equipping the MGV with a water capturing mechanism (SAM_008). This ranking is reasonable since these requirements complement the other sampling requirements. In particular, the UGV is already equipped with a camera and water capturing mechanism. Also, the need for the UAV to transport the MGV would only be applicable to sampling scenarios where the MGV needs to reach AoIs it cannot navigate on its own, e.g., hard-to-access locations or disaster-affected areas with debris.

Analyzing the transparent funding service, eight requirements were ranked as “MUST” and four as “SHOULD”, which include the TF_003 (smart contracts), TF_004 (tokenization), TF_006 (real-time donation tracking), and TF_010 (governance and consensus mechanisms) requirements. While these features enhance transparency and efficiency, their applicability varies based on the scale and complexity of the disaster management operation. Also, a disaster management professional commented that TF_010 might introduce unnecessary overhead in smaller-scale operations.

In the VR training service, 11 requirements received a “MUST” and 4 a “SHOULD” ranking. Among the “SHOULD” requirements, VR_010 emphasizes that training and materials should be exclusively in English, which might not align with local needs, as mentioned by two academics. VR_011 necessitates specific training equipment and room conditions, thus potentially limiting adaptability in resource-constrained areas. VR_002 specifies that training cannot be excessively lengthy and VR_013 describes the need for modelling additional VR objects to make visualizations more representative of real disasters. Both requirements were deemed to be of lesser importance for the service’s success.

All 12 warehouse inventory management, 5 pest control, and 5 physical manipulation requirements were ranked as “MUST”. Concerning reconstruction, five out of six requirements were evaluated as “MUST”, with RE_003, which pertains to tethering a UAV to a construction machine, categorized as a “SHOULD” requirement.

Finally, 25 UI requirements were ranked as “MUST” and 2 as “SHOULD”. This ranking diminishes the importance of specifying a click limit in UI interactions (UI_005), considering that the complexity of interfaces can vary, and certain actions may inherently require more clicks, as noted by a technical expert among the participants, with experience in such interfaces. Similarly, adhering to external standards (UI_015) may not always be practical in cases where standards conflict or customized interfaces are required.

Overall, requirements had missing values ranging from 0% to 35%, with the highest values, exceeding 20%, observed in the drone, communication and recharging requirements. One possible explanation for this pattern is the innate technical nature of these requirements, which could have potentially obstructed participants’ understanding, and thus led to a decrease in the number of answers collected. Thankfully, only approximately 6% of the requirements (N = 19) displayed such remarkably high missing value rates. The range of missing values across participant groups spanned from 3.43% for disaster management professionals to 13.40% for academics, which is a reasonable result given that the academics reported the least expertise in disaster management domains. Two academics left 45% and 53% of the questionnaire unanswered, significantly impacting the response rates.

Of particular concern were the standard deviation rates, which surpassed 1.00 in 74 requirements (22.1%). The highest proportions of high deviation rates were observed in the transparent funding (50% of requirements), reconstruction (50% of requirements), social media sensing (53% of requirements), and warehouse inventory management (67% of requirements) services. It is worth noting that while a standard deviation of more than 1.00 is an initial indication of variability, it alone does not determine statistical significance. This is why non-parametric tests were subsequently carried out, to confirm whether the observed standard deviations are statistically significant and provide more robust evidence of differences among participant groups.

6.2. Non-Parametric Tests

The results of the non-parametric tests are presented in the following sections.

6.2.1. Kruskal–Wallis H Test

Null Hypothesis: Equation (1) presents the null hypothesis of the Kruskal–Wallis H test.

$$H_o = \mathrm{There} \mathrm{is} \mathrm{no} \mathrm{significant} \mathrm{difference} \mathrm{among} \mathrm{the} \mathrm{distributions} \mathrm{of} \mathrm{the} \mathrm{seven} \mathrm{groups}.$$

(1)

Alternative Hypothesis: Equation (2) presents the alternative hypothesis of the Kruskal–Wallis H test.

$$H_a = \mathrm{There} \mathrm{is} \mathrm{a} \mathrm{significant} \mathrm{difference} \mathrm{among} \mathrm{the} \mathrm{distributions} \mathrm{of} \mathrm{the} \mathrm{seven} \mathrm{groups}.$$

(2)

For this test, the overall differences among all groups for each requirement were analyzed. The Chi-square, df and p-values for all requirements are provided in Supplementary Material File S3. The significance used depends on the sample size. As a rule, when the sample size is larger than 30, the asymptotic significance is used, while in the opposite case, exact significance is used. Therefore, for this test (N = 26), exact significance was used. The significance level used for the Kruskal–Wallis H test was 95%. Requirements in which significant differences among groups were present are highlighted, and Mann–Whitney U tests were used to explore pairwise differences between groups. The null hypothesis was rejected for a total of 25 requirements (7.46%) and retained for the rest.

6.2.2. Mann–Whitney U Test

Mann–Whitney U tests were performed individually for all 25 requirements with significantly different responses. The number of pairwise comparisons among the groups of participants is given by the following formula, where n = 7 and r = 2:

$$C\left(n,r\right)={\displaystyle \frac{n!}{\left[r!\left(n-r\right)!\right]}}$$

(3)

Therefore, each requirement was tested 21 times, as C(7,2) = 21, and 525 Mann–Whitney U tests were performed (21 × 25 = 525) to evaluate pairwise comparisons among all seven groups.

Null Hypothesis: Equation (4) presents the null hypothesis of the Mann–Whitney U test applied to each pair of groups (e.g., academic vs. disaster management professional, academic vs. end-user, disaster management professional vs. end-user, etc.):

$$H_o = \mathrm{There} \mathrm{is} \mathrm{no} \mathrm{significant} \mathrm{difference} \mathrm{in} \mathrm{the} \mathrm{distributions} \mathrm{of} \mathrm{responses} \mathrm{of} \mathrm{the} \mathrm{two} \mathrm{groups}.$$

(4)

Alternative Hypothesis: Equation (5) presents the alternative hypothesis of the Mann–Whitney U test applied to each pair of groups (e.g., academic vs. disaster management professional, academic vs. end-user, disaster management professional vs. end-user, etc.):

$$H_a = \mathrm{There} \mathrm{is} \mathrm{a} \mathrm{significant} \mathrm{difference} \mathrm{in} \mathrm{the} \mathrm{distributions} \mathrm{of} \mathrm{responses} \mathrm{of} \mathrm{the} \mathrm{two} \mathrm{groups}.$$

(5)

The sequence of Mann–Whitney U tests demonstrated significant differences among the seven groups in various requirement categories (

Table 2. Significant differences identified by the Mann–Whitney U tests.

Req.	Groups Compared	Median Group A	Median Group B	U	Z	p-Value
MGV_001	Academic vs. End-User	4	3	4	−2.214	0.045
	End-User vs. Stakeholder	3	5	0	−2.16	0.048
CCWA_007	Academic vs. End-User	4	3	4.5	−2.383	0.024
	Disaster Management Professional vs. End-User	5	3	0	−2.029	0.048
CCWA_014	Academic vs. End-User	4.5	3	2.5	−2.754	0.007
	Disaster Management Professional vs. End-User	5	3	0	−2.449	0.048
FRMA_008	Academic vs. End-User	4	3	4.5	−2.531	0.013
ARECH_002	Academic vs. End-User	4	5	4.5	−2.249	0.034
ARECH_003	Academic vs. End-User	4	5	2.5	−2.785	0.005
SOCI_001	Academic vs. End-User	4	5	7.5	−2.277	0.031
SOCI_002	Academic vs. End-User	4	5	6	−2.411	0.035
RS_001	Academic vs. End-User	5	3	8.5	−1.993	0.036
	Disaster Management Professional vs. End-User	5	3	0	−2.160	0.048
REL_001	Academic vs. End-User	4	5	2.5	−2.559	0.015
PRE_002	Academic vs. End-User	4	5	5	−2.56	0.015
SAM_004	Academic vs. End-User	5	3	0	−3.071	0.003
	Disaster Management Professional vs. End-User	4.5	3	0	−2.415	0.048
	End-User vs. Stakeholder	3	5	0	−2.449	0.048
TF_005	Academic vs. Disaster Management Professional	4	5	0	−2.582	0.022
INT_017	Academic vs. Disaster Management Professional	3	5	0	−2.049	0.048

). These groups were input to SPSS as follows: Group 1 (academics), Group 2 (disaster management professionals), Group 3 (end-users), Group 4 (others), Group 5 (stakeholders), Group 6 (subject matter experts), Group 7 (technology experts). For each pairwise comparison, Mann–Whitney U, Z score, and p-value were calculated for all requirements where significant differences between groups were found via the Kruskal–Wallis H test. The two-tailed exact significance (exact p-value, corrected for ties) is preferred for the Mann–Whitney U test and was also selected to be used in our analysis. Median values (Md) for each requirement within a group were also documented. The significance level used for the Mann–Whitney U test was 95%.

Taking into consideration the results above, the Mann–Whitney U tests showed significant differences among participant groups in 14 out of the 25 requirements pinpointed by the Kruskal–Wallis H test, and the null hypothesis was rejected for 20 out of the 525 Mann–Whitney U tests run in total.

6.2.3. Interpretation of the Non-Parametric Test Results

Out of the 14 requirements identified by the Mann–Whitney U tests as significantly different rankings, only 3 had been ranked as “SHOULD”, with the rest of them being “MUST” requirements. Regarding the “SHOULD” requirements, end-users rated FRMA_008, specifically the “Follow path” option in the FRMA interface, lower than academics did. This divergence in ratings aligns with the comments from end-users who emphasized the importance of UI simplicity, stressing that UIs should be user-centric, allowing users to concentrate on their tasks without being distracted by the interface. Similarly, disaster management professionals rated INT_017 higher than academics. This disparity is relevant to the former group’s need for quick and efficient access to specific tools, in this case, Docker images, as substantiated by a relevant comment. MGV_001 was ranked significantly lower by end-users compared to academics and stakeholders. End-users in roles like firefighters, medical professionals, and police officers may prioritize factors directly related to their operational needs during disaster response. They might consider the specific speed of the MGV less critical because it may not directly impact their immediate tasks in a disaster scenario.

Moving on to “MUST” requirements, CCWA_007 (the “Follow path” option in the CCWA), CCWA_014 (warnings raised when sensors are unavailable), RS_001 (installation of personal video cameras for first responders), and SAM_004 (equipping the UAV with a water capturing mechanism for sampling) were all ranked significantly higher by academics and disaster management professionals than end-users. Notably, stakeholders also ranked SAM_004 higher than end-users. Similar to FRMA_008, academics and disaster management professionals value CCWA_007 higher because it offers advanced capabilities for planning and executing disaster response missions, whereas end-users prioritize simplicity and ease of use, therefore considering this functionality as less critical. For CCWA_014, academics and disaster management professionals place a higher emphasis on system reliability, as they recognize the importance of timely warnings in case of sensor malfunction to maintain data accuracy. In contrast, end-users prioritize other aspects, such as system UI simplicity, as mentioned earlier. Academics and disaster management professionals find more value in RS_001 than end-users for several tasks, such as documenting response operations, post-incident analysis, and training improvements. On the contrary, end-users prioritize immediate safety concerns during a crisis, focusing on real-time localization and communication for guidance and personal safety, as well as the detection of immediate threats, such as hazardous gases, radiation, or temperature changes. The majority of end-users raised practicality concerns regarding the use of personal video cameras, citing potential issues such as mobility hindrance. In contrast, the other sensors proposed for risk status monitoring (RS_002: CO detection sensors, RS_003: temperature sensors, RS_004: GPS sensors, RS_005: gas sensors, RS_006: radiation sensors) provide critical real-time data without complicating tasks. Additionally, two end-users expressed reservations regarding privacy and ethical aspects related to video recording during response operations. Academics, disaster management professionals, and stakeholders rated SAM_004 higher than end-users did. Again, end-users may prioritize operational simplicity during real-time disaster operations and consider the addition of a UAV with a water capturing mechanism duplicative, since the UGV already has one. Academics and disaster management professionals could have rated SAM_004 higher because they have a deeper understanding of the technical capabilities and potential benefits with regard to adaptability of deploying an UAV for collecting samples in hard-to-reach areas.

ARECH_002 and ARECH_003, both related to MIMO recharging antennas, SOCI_001 and SOCI_002, addressing real-time social media sensing and text analytics, PRE_002, focused on hazard footprint prediction, and REL_001, which requires the inclusion of a tethered UAV communication relay, were ranked higher by end-users than by academics. End-users have rated ARECH_002 and ARECH_003 higher than academics on account of their recognition of the practical benefits of MIMO recharging antennas in real disaster scenarios, where uninterrupted communication is crucial. Academics, on the other hand, may have considered these requirements from a broader technological standpoint, taking into consideration the intricate technical complexities and challenges, such as interference management, alignment and beamforming, etc., involved in deploying such technology in real-life disaster scenarios. Similarly, academics approached PRE_002 with an emphasis on the challenges and complexities associated with combining multiple data sources, including real-time meteorological data, topographic data, and real-time sensor information, for precise hazard footprint prediction. This stance is further reinforced by a comment provided by an academic. The requirement of tethered communication relay (REL_001) raised controversy among end-users and academics, in contrast to the non-tethered relay proposed in REL_002, where no significantly different responses were observed across participant groups. This discrepancy can be attributed to end-users, who, having encountered situations where traditional communication methods failed or were unreliable, have a greater appreciation for the practical significance of a tethered communication relay, as emphasized by an end-user’s comment. End-users prioritize SOCI_001 and SOCI_002 due to their immediate impact on real-time situational awareness and operational effectiveness during disasters. Academics may assess these requirements with a more research-oriented perspective, beyond immediate disaster response. As expressed by an academic, the extraction of real-time observations from online social media users (SOCI_001) requires sophisticated web scraping or data crawling techniques, robust APIs, and mechanisms to ensure data accuracy and integrity. They also added that multilingual text analysis (SOCI_002) poses complex challenges requiring natural language processing (NLP) expertise and access to comprehensive language resources. Both requirements may involve the use of ML models for tasks such as sentiment analysis, topic modeling, and content classification. Building, training, and fine-tuning these models, as well as ensuring their scalability and adaptability to different disaster scenarios, can also present technical complexities. Additionally, the extraction and analysis of social media data for disaster response must address concerns related to data privacy and ethics.

Last but not least, TF_005, i.e., the integration of identity verification procedures for transparent funding, received higher ratings from disaster management professionals compared to academics, a discrepancy reasonably attributed to the fact that disaster management professionals likely have a heightened awareness of potential fraudulent activities within the HL domain, thanks to their practical experience in frameworks and strategies across all disaster phases. This familiarity offers a plausible explanation for their elevated rating of this security requirement—compared to academics’ rating.

6.3. Analysis of Open-Ended Responses

Regarding the open-ended questions in the questionnaire, participants provided valuable insights into their familiarity with similar platforms and challenges they encountered while using them. They also offered justifications for their ratings and provided insightful feedback.

In terms of familiarity, approximately 30.7% of the participants, dispersed across all professional backgrounds, had prior experience with similar disaster management platforms and functionalities. When requested to provide details about the challenges they faced during their experience, subject matter experts highlighted the need for real-time monitoring of hazardous agents during chemical, biological, radiological and nuclear (CBRN) accidents and the use of drones for flood events. Technology experts mentioned challenges related to data standards and coordination as well as security and scalability during disasters. Some end-users raised concerns about interoperability and security, while others did not report any challenges despite their familiarity. Neither stakeholders nor disaster management professionals specified any challenges, despite their experience with pertinent systems.

The justifications provided for the ratings played a role of paramount importance in adding qualitative context to the quantitative analysis of the results. Some participants even looked ahead to the subsequent design and development phases, raising concerns about changing regulations and legal frameworks governing the use of drones in specific geographical regions and emphasizing the need for adaptability. Participants also addressed the configurability of the platform. Providing customizable configurations within the platform would enable different departments to tailor the system to their specific operational needs. Such comments were made by both end-users and disaster management professionals.

Finally, some participants suggested new requirements. A firefighter expressed the desire for real-time updates and notifications related to firefighting equipment and gear. They envisioned a system that could monitor the status of breathing apparatus, firefighting suits, and safety gear. In case of equipment malfunctions or resource depletion, such as low air levels in the breathing apparatus, they suggested implementing an emergency distress signal feature. This feature would trigger an immediate response from the command center and nearby first responders to provide prompt assistance during life-threatening situations. Moreover, an end-user made a comment emphasizing the necessity for UIs to be designed to accommodate first responders wearing gloves and other protective gear.

7. Discussion: Implications for AIRDROP

This study revolved around conducting an exhaustive requirements engineering analysis for the AIRDROP platform. Our mission encompassed a thorough exploration of AIRDROP and its multifaceted services. Our overarching objective was to ensure that AIRDROP emerged as a holistic solution, capable of addressing a diverse spectrum of disaster scenarios across all four stages of disaster management.

Within this scope, we embarked on the task of formulating RQs that guided the development process and ensured a focused and effective inquiry into the platform’s capabilities and needs. The first question focused on the elicitation and specification of foundational requirements for AIRDROP, seeking to identify both functional and non-functional aspects essential for the platform’s operation and success, whereas the second question moves us into the validation process, examining how the relevance and adequacy of these identified requirements can be affirmed to meet such complex HL dynamics. The effective execution of the proposed requirement engineering process is expected to significantly bolster the subsequent development stages and enhance the quality of the final software product to be developed.

Addressing RQ1, this paper started with the elicitation and meticulous specification of AIRDROP’s requirements. This involved a rigorous process that culminated in the collection and validation of a total of 335 functional and non-functional requirements, carefully distributed among 19 services. In response to RQ2, a questionnaire survey was carried out to validate the significance of these requirements. A panel of 26 experts from various professional backgrounds was asked to assess their importance, using a rating scale ranging from 1 to 5. Leveraging the renowned MoSCoW methodology, these requirements were then prioritized, ensuring that the most critical aspects were given their rightful precedence in the subsequent design phase. Our exploration did not halt at quantifiable assessments; we delved deeper to understand how respondents’ diverse professional backgrounds influenced their perspectives. This analysis, facilitated through non-parametric tests, illuminated the nuanced dynamics within our participant groups. Moreover, recognizing the richness of qualitative insights, we solicited open-ended responses from our participants. Their feedback, concerns, and suggestions enriched our understanding and served as a vital complement to the quantitative findings.

Most requirements ended up being ranked as “MUST” (N = 277, 82.69%) and “SHOULD” (N = 57, 17.01%). Only one requirement (N = 1, 0.3%) was ranked as “COULD”. None of the requirements were ranked as “WOULD”. In total, the Mann–Whitney U tests showed significant differences among participant groups in 14 (11 being “MUST” and 3 “SHOULD” requirements) out of the 25 requirements pinpointed by the Kruskal–Wallis H test.

In the subsequent phase of AIRDROP design, none of the “SHOULD” or “COULD” requirements will be considered. This decision aligns with the purpose of the MoSCoW prioritization method, which is designed to guide teams in concentrating their efforts on the most high-priority requirements. In this context, “MUST” requirements are deemed indispensable for the system’s core functionality and overall success, and this approach ensures that the core functionality is thoroughly established before entertaining the inclusion of supplementary, less critical elements. Additionally, this emphasis on “MUST” requirements in the initial design phase underscores a user-centric approach to software development. It guarantees that the most critical features, as determined by user needs and priorities, receive primary attention.

Our requirement exclusion process for the design phase extends beyond this point, guided by the results of non-parametric tests. Out of the 11 “MUST” requirements identified as significantly different through the Mann–Whitney U tests, 4 were deemed necessary to exclude. Firstly, ARECH_002 and ARECH_003, despite their “MUST” ranking, have to be omitted from the design phase. This decision was based on the fact that academics, with their extensive expertise and experience, recognized the inherent technological complexities associated with these requirements and provided significantly lower rankings than end-users. The challenges posed by these requirements are not proportional to the benefits offered with respect to operational effectiveness. Addressing energy efficiency is a valid concern, but diving into antenna design intricacies is not warranted.

Similarly, for PRE_002, end-users rated this requirement higher for its impact on situational awareness during disasters. However, academics considered the technical complexities in data integration and analysis, resulting in significantly lower rankings. Again, the complexity of dynamically predicting hazard footprints based on multiple data sources is not proportional to the benefits it offers in the context of AIRDROP when compared to real-time monitoring and assessment. In contrast, SOCI_001 and SOCI_002, despite academics’ concerns about technical complexity, align with end-user priorities for real-time situational awareness. This alignment ensures that AIRDROP meets the operational needs of end-users, justifying the retention of these two requirements.

RS_001 directly pertains to end-users and was identified as a potential mobility issue for them, evident in their significantly lower rankings compared to academics and disaster management professionals, as well as their open-ended responses. Given the focus on simplicity and usability for end-users, RS_001 had to be excluded from the design phase. On the contrary, CCWA_007 and CCWA_014, although rated lower by end-users compared to academics and disaster management professionals, were retained, as CCWA is tailored to cater to the needs of operational decision-makers, not end-users.

As for SAM_004, in spite of lower end-user ratings, it was retained due to its potential recognized by academics and disaster management professionals for enhanced adaptability in constrained spaces. The decision to retain REL_001, a tethered communication relay, is grounded in end-users’ practical experiences where traditional methods failed during disasters, thus addressing their real-world needs for effective communication. Finally, keeping TF_005 for identity verification, rated higher by disaster management professionals, aligns with their practical expertise in fraud prevention within HL.

Thus, following the MoSCoW analysis and non-parametric tests, 273 requirements were identified as eligible for the design phase. These requirements sufficiently address the challenges expressed by participants familiar with similar platforms. AIRDROP includes 14 security requirements and seven scalability requirements. It also covers CBRN-related requirements, including hazardous gas and radiation monitoring. These requirements are crucial not only for environmental concentration monitoring through drones (GASRA_001, GASRA_002) but also for protecting first responders (RS_005, RS_006). Additionally, AIRDROP encompasses requirements for symptom monitoring and medical care, both applicable to biological hazards (MED_001, MED_003 to MED_005), as well as sampling for chemical hazards (SAM_001 to SAM_005, SAM_009). The platform also incorporates drones for flood detection and monitoring (FLOOD_001). In line with these requirements, AIRDROP prioritizes the establishment of robust data standards to ensure seamless data exchange among the numerous components of the system (45 data requirements). The platform also places a strong emphasis on interoperability (TF_001, INT_009), enabling it to integrate with existing disaster management systems.

In addition to these requirements, two critical requirements, informed by participant feedback, were also incorporated. The first essential requirement, which pertains to the rescue module, is the “Real-time Monitoring and Emergency Distress Signal”. This functional (operational) requirement entails the integration of a real-time monitoring system for fire-fighting equipment. This system continuously monitors the status of breathing apparatus, firefighting suits, and safety equipment. In the event of malfunctions or resource depletion, the platform promptly activates an emergency distress signal. The second requirement, “Adaptability for Protective Gear”, is a non-functional (usability) requirement, which pertains to the UI module. Recognizing the paramount importance of user-friendliness, this requirement focuses on enhancing the adaptability of all AIRDROP UIs. The objective is to make them glove-friendly, facilitating seamless interactions for users, particularly first responders who wear protective gear.

In conclusion, in the upcoming design phase, a total of 276 requirements, out of the initial 335, will be considered.

8. Conclusions

This study aimed to elicit, specify and validate the requirements for AIRDROP, a proposed integrated HL platform, ensuring their suitability across diverse disaster scenarios and all stages of disaster management. Following a thorough process of elicitation and specification, which involved gathering insights from the literature and consultations with industry experts, the validation process involved a questionnaire survey administered to 26 participants from various professional backgrounds. The requirements were prioritized using the MoSCoW methodology, and significant differences among participant groups were identified through the Kruskal–Wallis H and Mann–Whitney U tests. The results guided the selection of requirements for the design phase. This process entailed the exclusion of “SHOULD” and “COULD” requirements, with a focus on prioritizing essential “MUST” requirements, aligning with a user-centric approach. Further refinement considered factors like technological complexities and practicality, leading to the exclusion of four “MUST” requirements, specifically focusing on those where significant differences were present among participant groups. Additionally, two critical requirements were introduced, one for the rescue and another for the UI modules. As a result, 276 requirements advanced to the design phase. It is important to recognize that requirements in the field of HL, as in many dynamic domains, are subject to continuous evolution. What is considered a critical requirement today may evolve as technology and regulations change over time. Therefore, ongoing assessment and adaptation of requirements will be integral in the design phase to ensure that AIRDROP remains cutting-edge and aligned with the evolving needs.

Despite the ever-increasing need for this integration, it remains a largely underexplored area of research [6,97], particularly in humanitarian contexts, contrasting with their widespread use in civilian applications [98]. With its sixteen unique services, AIRDROP emerges as a cutting-edge tool, devised for a broad range of disaster scenarios and phases, from mitigation and preparedness to response and recovery, effectively aligning itself with the academic endeavor for holistic methodologies in HL [44]. Unlike other platforms discussed in the literature, AIRDROP uniquely covers all aspects of disaster management, offering an unprecedented solution that addresses both the operational and logistical challenges across its full spectrum.

Methodologically, this work utilizes DSR, identifying critical features and technical components that enhance drone performance in HL, in line with the recommendations of recent studies focusing on drones in supply chain management and logistics [97]. Furthermore, this research underscores the critical role of stakeholder engagement in the technology development process. By actively involving—through a questionnaire survey—disaster management professionals, tech experts, academics, end-users, and community representatives in the co-design of AIRDROP, the research highlights the value of collaborative technology development and sets a standard for involving diverse perspectives in creating relevant and responsive disaster management solutions [99]. The theoretical importance of this research lies in addressing a significant gap in the existing HL literature. The study not only fills the void of empirical research in HL [70,71,73]—a field dominated by commercial logistics studies [72]—but also contributes to the theoretical understanding of how integrated technological solutions can enhance disaster management effectiveness. The use of mixed methods, combining qualitative and quantitative approaches, provides a robust framework for developing comprehensive, real-world solutions that are both theoretically sound and practically viable. This approach aligns with the broader need for empirical research in HL [71,100], advocating for the continuous evolution and adaptation of requirements to keep pace with technological advancements and regulatory changes.

9. Limitations

Of course, it is crucial to reflect on the inherent limitations that accompany this research and the implications they hold for the future of AIRDROP. The deployment of a questionnaire survey faced the dual challenge of navigating the complexities of an exhaustive questionnaire while also capturing a momentary snapshot of stakeholder needs. This expansive survey may have contributed to questionnaire fatigue, potentially affecting the depth and quality of the responses. Efforts were made to alleviate this through careful survey design, including the use of Google Forms. This platform allows respondents to save their progress and return to complete the questionnaire at their convenience, thus preventing fatigue. Moreover, as already addressed in Section 5, excluding “Not Sure” responses from the statistical analysis may have limited our insight into areas of participant uncertainty regarding specific requirements. This methodological choice, while intended to streamline the analysis, represents a limitation in capturing the full spectrum of stakeholder perspectives, particularly regarding the nuanced understanding of participant indecision. Future studies could benefit from exploring these patterns of uncertainty in greater detail, potentially offering valuable insights into stakeholder engagement and requirement clarification needs.