Computational Ergo-Design for a Real-Time Baggage Handling System in an Airport

Abstract

Despite the growing importance of human-centered design and ergonomics in various fields, a significant gap exists in applying these principles to robotic systems in airport environments. This paper focuses on a real-time baggage handling monitoring system by proposing a computational ergo-design approach. It presents the optimal system architecture for real-time baggage handling. The proposed architecture, called ARTEMIS (ARchitecture for real-TimE baggage handling and MonitorIng System), is designed for real-time baggage handling and monitoring. The circuit modeling is carried out using a directed graph. Five strategies are simulated to test their effectiveness and evaluate their performance within the system. A simulation that generates key indicators enables preliminary visualization and analysis of AGV behavior through predefined scenarios. These results are presented through an intuitive and ergonomic user interface, designed with a focus on user–computer interaction as a problem-solving process centered on the user’s experience. The results show that, if the goal is to balance energy efficiency with effective baggage handling, the Mixed Advance/Delay Strategy appears to be the best overall choice, as it optimizes both energy consumption and baggage handling while maintaining relatively low waiting times. However, if minimizing queue time and maximizing baggage collection are the highest priorities (with less emphasis on energy efficiency), the Turnstile Strategy remains a solid option. In addition, the simulations show that the operator plays a central role in minimizing delays and ensuring the smooth operation of the system. Both local and global system failures depend heavily on the operator’s response time, decision-making, and overall efficiency. Therefore, operator efficiency and a well-designed support system are critical to maintaining a smooth and effective baggage handling process.

1. Introduction

The rapid advancement of automation in logistics and transportation has significantly transformed operational efficiency in airports. Today, airports face increasing pressure to improve efficiency while minimizing environmental impact. In the modern context of air travel, efficient baggage handling has become a foundation of both passenger satisfaction and operational quality. Airports worldwide are under increasing pressure to manage the growing volume of passengers and the associated complexity of logistical operations. To meet these demands, autonomous mobile robots (AMR) or automated guided vehicles (AGV) are increasingly deployed for different tasks that present challenges in optimizing their movement to ensure both energy efficiency and operational effectiveness. Airports, with their dynamic and often congested environments, also require these robots to operate alongside humans, further complicating the balance between task efficiency, safety, and adaptability. Moreover, the baggage handling system (BHS) is a network of technologies implemented at airports, representing a complex system for humans, as access becomes very difficult in the event of a breakdown.

Despite the growing use of autonomous robots in logistics and transportation, existing research often prioritizes task efficiency over energy consumption. Current studies overlook how these systems should function in dynamic, human-shared environments, where unpredictable events introduce additional complexity. While human-centered design and ergonomics are widely discussed in various fields [1,2,3,4,5,6,7,8,9,10,11,12,13,14] and with various creative methods [15,16,17,18,19], there is a clear gap in research when it comes to applying these principles to robotic systems—specifically in designing systems that balance task efficiency, energy efficiency, and human safety in both robot-only and human–robot shared environments.

Addressing these challenges requires innovative approaches. In this context, ergo-design—an emerging discipline that integrates ergonomic principles with design—emerges as a powerful tool for developing sustainable baggage management systems. At its core, ergo-design is about continuously incorporating the human factor into system design. Ergo-design also acts as the bridge between these two disciplines. By considering human interactions with both technology and infrastructure, ergo-design ensures that systems are not only operationally effective but also human-centric. The advancements in computational design [20,21] can bring a significant evolution in ergo-design, giving rise to computational ergo-design. This approach employs advanced simulations (e.g., coupled MATLAB and Simulink (R2024b Release Highlights—MATLAB and Simulink) and Python (Python Release Python 3.10.11) simulations [22]), real-time data processing, and computational modeling to design, test, and refine human-centered complex systems. Computational ergo-design is a key component of smart ergo-design, which integrates intelligent technologies and interconnected systems to meet the dynamic demands of modern environments such as airports [23,24,25].

Many methods have been proposed to model and simulate complex systems before their development, with the aim of efficiency, robustness and sustainability of these systems [26,27]: mainly, discrete-event modelling and simulation [28], system dynamics modelling and simulations [29], or agent-based modelling and simulation (ABMS) [30]. In ABMS, agents are modeled to perform tasks autonomously, to communicate and interact with each other, and to evolve and perceive dynamic environments [31]. Their simulation allows for us to visualize and analyze their adaptive behaviors in very varied fields, including transportation [32,33] and autonomous vehicles [34]. The simulation of a fleet of AGVs in an airport is, therefore, an excellent illustration of this [35].

Within the domain of airport baggage handling, ergo-design focuses on two primary stakeholder groups: (i) passengers and (ii) operators. By addressing the needs of both groups, ergo-design enables the creation of systems that are human-centered, functional, and sustainable. For passengers, the goal is to provide continuous access to information regarding the location and status of their baggage, ensuring reliable delivery to the intended destination and real-time updates along the way. User-friendly interfaces and timely notifications play a vital role in meeting these expectations. Meanwhile, airport operators require effective tools to track, trace, and manage baggage efficiently while minimizing errors or delays [36]. Recent literature is rich in publications on baggage tracking and tracing systems in the airport sector [23,24]. Some publications are based on QR code technology and blockchain, as illustrated in [37]. The idea behind this publication relies on assigning a unique QR code to each piece of luggage and on the fact that each QR code links the physical baggage to its digital twin on the blockchain. The security of this system is significantly enhanced using blockchain technology. The use of blockchain technology significantly enhances the security of the system. Other publications explore cognitive IoT approaches, such as in [38], where the authors leverage the principles of the Cognitive Internet of Things (Cognitive IoT) for intelligent baggage management in airports. The paper implements RFID and Bluetooth tags combined with a sensor network to achieve accurate and real-time tracking of baggage journeys. Finally, some publications are based on convolutional neural networks. In this category, reference [39] can be cited, where the authors deployed a set of cameras to capture aerial-view images. A self-supervised learning technique using a convolutional neural network is then applied to the segmented images for multi-object tracking. The key challenge is to provide operators with robust tools for securely and efficiently tracking and managing baggage within the complex infrastructure of airports. To address these needs, computing technologies are indispensable.

Several issues faced by airports are addressed using MATLAB and Simulink simulations, including the following:

• Baggage security control [40];
• Passenger flow management, either through macroscopic modeling [41] or using queueing systems [42];
• Evaluating fuzzy control strategies to reduce energy consumption while simultaneously ensuring passenger comfort in airport terminals [43];
• Analysis of energy load flow at the injection substation of Abuja Airport (Nigeria), highlighting the impact of decentralized production through the integration of genetic algorithms [44].

Carlson and Murphy [45] classified the failures into two main categories: human-related (e.g., design, interaction) and physical (e.g., communications, effector, sensors, power, control system, field-repairable and non-field-repairable, etc.). Expanding on this work, Ramesh et al. [46] introduced the concept of “robot vital signs”, defined as a set of parameters indicating a robot’s health and its ability to operate autonomously despite performance degradation. When such degradation is minimal or absent, these vital signs remain within a defined range of values, which may vary depending on the robotic hardware or environmental conditions. Variations in one or more vital signs can offer valuable insights into the specific causes of performance degradations. These failures have thus been the subject of specific studies for AGVs used in Industry 4.0 environments [47,48].

Aivaliotis et al. [22] propose an algorithm for estimating the remaining useful life (RUL) while considering the production plan. This approach enables the prediction of the robot’s future dynamic behavior. The execution of the RUL prediction algorithm relies on the Python-based RUL prediction executor, which can be operated either manually or automatically using a task scheduler. This module, in turn, iteratively runs MATLAB functions until a failure is predicted or the entire prediction horizon is analyzed without any failure being detected.

This paper focuses on a real-time baggage handling monitoring system (BHMS) by proposing a computational ergo-design approach. The goal is to model a wide range of scenarios, evaluate the monitoring system’s performance under various conditions, and analyze how specific changes might affect overall efficiency while including the final operators in the loop. The simulation of these scenarios allows for system designers and operators to collaboratively explore and test performance under realistic and challenging conditions [26]. The co-construction and the tests of these scenarios ensure that the computations are aligned with the specific requirements of the airport environment, enabling precise identification of weaknesses and opportunities for improvement.

Furthermore, one of the key issues is determining the optimal system architecture for real-time baggage handling. Additionally, identifying effective computing strategies is crucial for rigorously testing the system’s performance and resilience. These strategies must account for the real-world complexities of airport operations. Finally, the design of the human–computer interface (HCI) is essential for operational success. By integrating ergo-design principles with computational tools, this research aims to comprehensively address these challenges.

Particularly, energy consumption is a critical factor in our simulation as airports prioritize sustainability and efficiency. These complex environments have significant energy demands from multiple integrated systems. While baggage handling system data remain limited, optimizing energy use is essential to balance operational performance with environmental impact [49,50]. Airports worldwide are adopting energy-efficient designs to meet sustainability targets. Effective energy management reduces both environmental impact and operational costs, making it vital for modern airport system design and simulation [51,52]. Efficient energy management not only supports environmental sustainability but also reduces operational costs, making it a vital component of airport design and operation.

The structure of this article is as follows. We start by defining an architecture for a real-time baggage handling monitoring system that we named ARTEMIS. Then, we present the modelling and simulation of the ARTEMIS system. In Section 4, we compare five strategies aimed at optimizing baggage queue waiting times within ARTEMIS. A study and a discussion of system degradation mechanisms in AGV circuits are given in Section 5. Following this discussion on the numerical simulation of scenarios of system failures and the conclusion, we propose our perspectives on future works.

2. An Architecture for a Real-Time Baggage Handling Monitoring System

The proposed architecture for a real-time baggage handling and monitoring system called ARTEMIS (ARchitecture for real-TimE baggage handling and MonitorIng System) at an airport is composed of four main components: (1) database component, (2) real-time environment, (3) monitoring component, and (4) data analysis component (Figure 1).

2.1. Functional Description of Components of ARTEMIS

The database component is responsible for creating and storing performance-related data. Simulations run on MATLAB and Simulink are essential for creating a robust database that logs key performance data. This includes statistics such as the maximum number of AGVs on the circuit, bags processed per hour, and delayed baggage counts in the event of a system breakdown, depending on the duration of the outage. MATLAB and Simulink simulations are instrumental in testing how the system reacts under various conditions. For example, the system can simulate failures to assess response times, identify bottlenecks, and visualize the impact of outages on baggage flow. The database archives operational metrics, such as delayed baggage numbers correlated with outage durations, alongside other reliability and performance metrics.

The real-time environment integrates real-world elements, such as planes and AGVs, and provides real-time visualization and modeling of AGV behavior in critical situations. It is composed of operational components and a simulation platform. The operational components focus on real-world elements, like aircraft and AGVs, while the simulation platform—based on Python—offers real-time data visualization through labeled nodes (e.g., “On/Off Parking”, “Loading”, “Drop”). This platform enables modeling, simulation, and analysis of AGV behavior in a set of problematic situations defined in test scenarios. This visual interface helps operators in tracking baggage flow and understanding the status of different system components.

The monitoring component provides a comprehensive interface for displaying real-time statistics and supporting decision-making. It combines digital data streams with physical observations to offer precise situational awareness. A monitoring interface presents real-time statistics to operators, helping them make informed decisions by merging physical observations with digital data streams, ensuring an accurate and up-to-date snapshot of system performance.

Finally, the data analysis component evaluates system performance by generating detailed reports and visual representations of key metrics, including baggage throughput, AGV utilization, and energy consumption. The data analysis component performs three main roles: (i) performance metrics analysis; (ii) graphical representation, and (iii) time analysis. In performance metrics analysis, data are analyzed to generate reports on key metrics, such as baggage throughput (e.g., number of arriving baggage, baggage handled, baggage in process, delayed baggage), AGV utilization, and energy consumption based on speed and distance traveled. In graphical representation, graphs and charts depict time-series data, revealing trends like the number of AGVs used per hour and variations in energy consumption over days. These visualizations help identify performance trends and areas for optimization. Time analysis generates detailed charts that highlight baggage handling efficiency, identifying peak and off-peak operational periods.

2.2. Behavior of ARTEMIS

The components of ARTEMIS interact with each other, ensuring smooth and efficient baggage handling operations. This real-time baggage handling and monitoring system leverages simulations, real-time data processing, monitoring interfaces, and analytical tools. The interconnected components enable adaptive control, quick responses to system failures, and performance optimization, making it a robust and comprehensive solution for airport baggage management.

Tasks begin with the arrival of planes and the physical movement of baggage from aircraft to designated drop points. As soon as a plane lands, the monitoring system receives digital data on the number of bags. Leveraging the database created via MATLAB and Simulink, the monitoring system sends operational data from the Python platform in real time, such as the number of AGVs allocated based on incoming baggage. It also receives feedback on the number of processed and in-progress bags. Feedback loops between the real-time environment and monitoring tools create an adaptive system capable of responding to changes. This ensures high reliability and continuous improvement, facilitating seamless communication between the simulation platform and real-time data processing. The integration of physical data collection (e.g., baggage counting) with digital processing via the Python simulation module enables real-time updates and precise tracking. Alerts are generated for scenarios such as system failures or when performance thresholds are reached (e.g., maximum AGV capacity or baggage processing rates). These alerts ensure timely operator intervention.

3. Modelling and Simulation of ARTEMIS

3.1. Modelling of the Handling System

The simulation model of the baggage handling system is composed of four elements: a circuit, AGVs, baggage, and a human operator.

The operating space for the robots is defined as a two-dimensional plane representing a closed and oriented loop—a circuit with a defined direction of travel. Within this circuit, two baggage input flows and two baggage output flows are defined (Figure 2). In addition, the circuit includes a main lane and two branches: an upper branch that merges a secondary input flow with the main one and a lower branch that merges an additional output flow with the main output. The circuit design thus includes two divergence points (where the main lane splits into branches) and two convergence points (where the branches rejoin the main lane).

The circuit modeling is carried out using a directed graph with 17 nodes (Figure 3):

• Node P corresponds to the AGV parking lot.
• Nodes R1 and R2 are the two baggage collection points linked to the two entry flows.
• Nodes D1 and D2 are the two baggage drop-off points associated with the exit flows.
• Nodes P3 and P11 are the two circuit divergence points.
• Nodes P8 and P16 are the two convergence points.
• The remaining nodes (P1, P2, P4, P6, P9, P10, P12, P14) represent characteristic points of the circuit topology.

The AGVs’ mission is to transport baggage from the collection points (R1 or R2) to the drop-off points (D1 or D2). They are modeled as software agents to simulate their movements and baggage handling activities as well as to analyze their behavior in different situations defined by specific scenarios. The agent-based paradigm is indeed well suited for simulating and analyzing the actions of autonomous, active, and reactive entities within their environment [53,54].

Each baggage is modeled as an object containing elementary information, such as an identifier and a position on the circuit. Baggage is generated based on different input flow strategies defined in the simulation scenarios: continuous, discontinuous, random flows, or flows derived from real-world airport data used as test cases.

Finally, an office, not connected to the circuit graph detailed above, is located near the AGV parking lot. A human operator is in this office to supervise the system and intervene in the event of a problem in baggage handling or AGV movement.

3.2. HCI of ARTEMIS

Human–computer interaction (HCI) focuses on designing better computer interfaces by evaluating their usefulness through user trials, where real users assess the system’s effectiveness and ease of use [55]. The goal is to create interfaces that are intuitive, efficient, and user-friendly. According to Norman [56], devices and interfaces should be easy to use, be intuitive, and function correctly. He outlines six key design principles: visibility, feedback, affordance, mental mapping, constraints, and consistency. These principles help create effective designs. Additional principles, such as visibility, affordance, feedback, simplicity, structure, consistency, tolerance, and accessibility, are emphasized [57].

Interface design rules include maintaining consistency, enabling frequent users to use shortcuts, and offering informative feedback [58]. Additional guidelines suggest designing dialogue to provide closure, simple error handling, and easy action reversal. Supporting an internal locus of control and reducing short-term memory load are also essential [58]. These guidelines aim to create intuitive and user-friendly interfaces.

Usability is not defined by a single characteristic but rather by a combination of factors. According to Nielsen [59,60], a system is usable if it is easy to learn, easy to remember, error-free or error-forgiving, and pleasant to use. Additionally, the Nielsen Norman Group proposed 10 general guidelines (heuristics) to improve human–computer interaction. These heuristics are broad recommendations, not strict rules, intended to enhance the user experience.

The general principles of interaction design emphasize that a system should provide clear and timely status updates; present information in ways users can understand; and maintain consistency in language, icons, and other elements to prevent confusion. These principles should be mapped to heuristics or guidelines, which should then be linked to GUI design rules. From there, solutions should be developed, evaluated, and tested, ultimately converging toward the optimal interface. These mappings ensure that user–computer interaction is a problem-solving process centered on the user’s experience [61].

A simulation, based on the model presented above, enables preliminary visualization and analysis of AGV behavior through predefined scenarios. This allows for us, for example, to evaluate the impact of an AGV failure or to test different baggage arrival flow patterns. This type of simulation can also generate key indicators, such as the average waiting time and the average transport time, allowing for an in-depth analysis of the performance. These results are presented through an intuitive and ergonomic user interface, designed by considering user–computer interaction as a problem-solving process focused on the user’s experience. This approach ensures that the interface design accounts for the system’s functionality and behavior and the user’s experience. The user interface and AGVs circulating on the circuit are modeled in Python and developed using a system based on software agents [34,62].

The HCI is organized into four main sections (Figure 4):

• A simulation area. This space displays the circuit where the AGVs operate and shows the simulation in real time.
• A parameter table. This space allows for users to configure the simulation parameters.
• A dashboard. This area provides real-time monitoring of information related to the simulation, including essential data to analyze its evolution.
• The control panel contains buttons to start, stop, or reset the simulation as needed.

The parameter table allows for users to configure the following elements:

• The number of AGVs on the circuit, therefore participating in the simulation.
• The total number of bags managed in the simulation.
• The number of bags processed per minute in the simulation.
• The status of the baggage entry and exit processes.
• The selection of a predefined simulation scenario.

Improving operator efficiency also involves training. Training the human operator in techniques for managing complex situations they may face is indeed a positive approach in preparing this key actor in the baggage management system. Certain training methods for human operators in highly constrained production environments make use of augmented reality scenarios combined with a voice-based digital assistant to immerse the trainee operator in use cases with the desired level of complexity. A good illustration of this approach can be found in [63].

Beyond the training, which is undoubtedly valuable for the operator, it also seems useful to observe them (with their consent, of course) during their real-world interventions in baggage management. This observation should be followed by a debriefing and discussion with the operator to identify areas for improvement in their practical actions. This exchange also serves as an opportunity to raise the operator’s awareness of the risks of accumulating delays due to slow response times, which are themselves often caused by the occurrence of critical situations that were insufficiently anticipated.

It seems useful to point out that the objective here was not to highlight, in a comparative approach, the performance of the ARTEMIS system but rather to present its relevance, characteristics, and functioning. Moreover, to the best of our knowledge, literature does not offer any system that is perfectly comparable to ARTEMIS in terms of its features. This article positions itself on a dual concept that includes, on the one hand, baggage handling, and on the other hand, a particular focus on ergo-design aspects, which also determine important user-related dimensions such as acceptability and sustainability [64].

4. Computing Strategies for ARTEMIS

4.1. Strategies to Optimize Baggage Queue Waiting Time

Five strategies are simulated to assess their efficiency in the system:

• Turnstile Strategy: AGVs continuously circulate along the circuit and pick up a bag if one is available at R1 or R2, depending on their assigned path.
• On-Demand Strategy: AGVs are deployed only when baggage arrives and are available in the parking area, operating with an optimal number of units.
• Delay Parking Strategy: This is a variation of the On-Demand Strategy, where AGVs remain on the circuit slightly longer to collect all the baggage.
• Needs Prediction Strategy: The number of AGVs circulating on the circuit is periodically calculated based on predicted demand. This is a variation of the On-Demand Strategy, where AGVs enter the circuit before baggage arrival.
• Mixed Advance/Delay Strategy: This combines Strategies 3 and 4, deploying AGVs into the circuit in advance and delaying them towards the end as necessary.

4.2. Cost Function

Consider energy consumption when the robot moves along the trajectory. Assume that the robot moves with a different velocity for each segment, as the trajectory is represented by curved pieces rather than straight segments; then, the energy consumption is given by (1) [65]:

$$E_{i-1,i}=\underset{T_{i-1,i}}{\overset{n}{\int }}{K}_{s}dt+\underset{T_{i-1,i}}{\overset{n}{\int }}\mu mg{v}_{i-1,i}dt+\underset{T_{i-1,i}}{\overset{n}{\int }}d\left({\displaystyle \frac{1}{2}}m{v}_{i-1,i}^2\right)dt$$

(1)

To keep things simple, we express the energy consumption discreetly in the following way (2):

$$E_{i-1,i}=K_s{T}_{i-1,i}+m(\mu g+a_{i-1,i})v_{i-1,i}{T}_{i-1,i}$$

(2)

where:

• K_s represents the energy loss during the transformation from electrical to mechanical energy, expressed as a static coefficient.
• µ is the ground friction coefficient associated with each path segment, assumed to remain constant along the segment.
• g represents the standard acceleration due to gravity (m/s²).
• m is the robot’s mass and baggage (kg).
• v_i−1,i is the linear speed for each path segment (m/s).
• a_i−1,i is the acceleration for each path segment (m/s²).
• T_i−1,i is the time interval for each path segment (s).

4.3. Numerical Simulation Results

The queuing model is implemented in MATLAB and Simulink to manage traffic flow at the circuit intersection using a round-robin method. This queuing model is used to minimize the number of AGVs present in the On-Demand Strategy while maximizing the number of bags handled. The queuing system is characterized by the following assumptions:

• The arrival of baggage follows a Poisson process.
• The service time for each baggage loading and drop-off point is deterministically distributed, each service taking 4 s.

Using numerical simulations, the maximum number of AGVs for the circuit shown in Figure 4 is determined to be 31 AGVs (NbAGVmax = 31):

• Pick-up/Drop-off Service Rate (µ): the average service duration is µ = 2 s, as there are two services, each lasting 4 s.
• Baggage Arrival Rate (λ): the average baggage arrival duration satisfies 0 ≤ λ < 2.
• Numerical simulations lead to the following conclusions:
• If λ ≤ 0.5, the system converges (the AGVs collect all the baggage available in the queue within a given time).
• If λ > 0.5, the system diverges (the AGVs are unable to collect all baggage available in the queue within a given time).

Then, the highest rate of baggage handling is λmax = 0.5.

Thanks also to several numerical simulations, we can obtain the following: given an integer N > 0 (N = 1, 2, 3, …), the ideal number of AGVs according to the number of bags is defined by (3):

λ* = λmax/N, which implies that NbAGV* = NbAGVmax/N

(3)

As illustrated in Figure 3, several possible paths can be identified. Although all paths consume approximately the same amount of energy, the path with the highest energy consumption is the one where the AGV starts from the parking lot and returns to it by traveling along both secondary roads. The energy consumption corresponding to this path is E = 0.24 kWh, where $E=\sum _{i=1}^K{\mathrm{E}}_{\mathrm{i}-1,\mathrm{i}}$, with K representing the number of segments in the path.

As illustrated in Figure 3, twelve possible paths can be identified.

Table 1. Energy consumption corresponding to each path.

Path	Path1	Path2	Path3	Path4	Path5	Path6	Path7	Path8	Path9	Path10	Path11	Path12
Energy (kWh)	0.23	0.22	0.21	0.24	0.23	0.22	0.23	0.22	0.22	0.24	0.23	0.22

presents the energy consumption corresponding to each path.

• Path1: [P, P1, P2, P3, P8, P9, P10, P11, P16, P1, P]
• Path2: [P, P1, P2, P3, P8, P9, P10, P11, P16, P1]
• Path3: [P1, P2, P3, P8, P9, P10, P11, P16, P1]
• Path4: [P, P1, P2, P3, P4, P6, P8, P9, P10, P11, P16, P1, P]
• Path5: [P, P1, P2, P3, P4, P6, P8, P9, P10, P11, P16, P1]
• Path6: [P1, P2, P3, P4, P6, P8, P9, P10, P11, P16, P1]
• Path7: [P, P1, P2, P3, P8, P9, P10, P11, P12, P14, P16, P1, P]
• Path8: [P, P1, P2, P3, P8, P9, P10, P11, P12, P14, P16, P1]
• Path9: [P1, P2, P3, P8, P9, P10, P11, P12, P14, P16, P1]
• Path10: [P, P1, P2, P3, P4, P6, P8, P9, P10, P11, P12, P14, P16, P1, P]
• Path11: [P, P1, P2, P3, P4, P6, P8, P9, P10, P11, P12, P14, P16, P1]
• Path12: [P1, P2, P3, P4, P6, P8, P9, P10, P11, P12, P14, P16, P1]

Six scenarios are conducted to evaluate the five strategies previously presented, focusing on energy consumption, the number of bags in the queue, and the waiting time for the bags. The first two scenarios simulate variable baggage flows with one and then two peak periods. The next three scenarios simulate baggage flows in the form of a staircase with a baggage peak positioned successively at the beginning, in the middle, and then at the end of the simulated period. The last scenario simulates a baggage flow with two peak periods, one converging and one diverging. All simulations are run over a period of one hour using MATLAB and Simulink.

4.3.1. Scenario 1: Variable Baggage Flows with a Single Peak Period

Scenario 1 represents a variable baggage flow, starting with a single peak period (λmax = 0.5), followed by a decrease in large bags, and finally showing a slight increase towards the middle of the peak period (see the results in Figure 5).

Table 2. Energy usage in the absence of baggage retrieval of Scenario 1.

Strategies	Strategy 1	Strategy 2	Strategy 3	Strategy 4	Strategy 5
NbUT	800	98	100	90	86
Energy (kWh)	192	23.52	24	21.6	20.64
MBWT	23	100	65	50	23

presents the energy lost during operation without baggage retrieval, number of unnecessary turns per AGV (NbUT), and maximum baggage waiting time (MBWT) corresponding to Scenario 1.

4.3.2. Scenario 2: Variable Baggage Flows with Two Converging Peak Periods

In contrast to Scenario 1, Scenario 2 highlights the variability of the baggage flow through two peak periods. The first peak occurs at the beginning, followed by a decrease in the number of large bags. Subsequently, a second peak period arises, followed by a slight decrease towards the middle of this peak (see the results in Figure 6).

Table 3. Energy usage in the absence of baggage retrieval of Scenario 2.

Strategies	Strategy 1	Strategy 2	Strategy 3	Strategy 4	Strategy 5
NbUT	730	121	141	120	143
Energy (kWh)	175.2	29.04	33.84	28.8	34.32
MBWT	18	85	67	30	18

shows the energy consumption losses without baggage retrieval, number of unnecessary turns per AGV (NbUT), and maximum baggage waiting time (MBWT) for Scenario 2.

4.3.3. Scenario 3: Baggage Flows as a Staircase Descending with a First Peak Period

Scenario 3 presents baggage flow dynamics that are like those of Scenario 1 but with a staircase pattern. This indicates that there is a peak period at the beginning, during which the baggage volume is high, followed by a gradual decrease until there are no more bags in the queue (see the results in Figure 7).

Table 4. Energy usage in the absence of baggage retrieval of Scenario 3.

Strategies	Strategy 1	Strategy 2	Strategy 3	Strategy 4	Strategy 5
NbUT	800	94	92	96	90
Energy (kWh)	192	22.56	22.08	23.04	21.6
MBWT	25	87	65	40	25

shows the energy consumption in the absence of baggage retrieval operations, the NbUT per AGV, and the MBWT for Scenario 3.

4.3.4. Scenario 4: Baggage Flows as a Staircase, with Both Upward and Downward Movement, Peaking in the Middle

Scenario 4 is a combination of Scenario 2 and Scenario 5. It represents a baggage flow dynamic that gradually increases to a peak, followed by a slight decrease during the middle of this peak period (Figure 8). This pattern resembles a staircase curve, where a continuous rise is observed before reaching a maximum, followed by a slight decline.

Table 5. Energy usage in the absence of baggage retrieval of Scenario 4.

Strategies	Strategy 1	Strategy 2	Strategy 3	Strategy 4	Strategy 5
NbUT	850	54	70	70	70
Energy (kWh)	204	13	16.8	16.8	16.8
MBWT	10	56	56	12	10

presents the energy consumption in the absence of baggage retrieval operations, the NbUT per AGV, and the MBWT for Scenario 4.

4.3.5. Scenario 5: Baggage Flows as an Upward Staircase, Reaching a Peak During the Final Period

Scenario 5 represents the opposite of Scenario 3. It illustrates a baggage flow dynamic that gradually increases, reaching a peak, somewhat resembling a staircase curve. This pattern could indicate a period of high activity, where the number of bags increases in a controlled manner before reaching its maximum (see the results in Figure 9).

Table 6. Energy usage in the absence of baggage retrieval of Scenario 5.

Strategies	Strategy 1	Strategy 2	Strategy 3	Strategy 4	Strategy 5
NbUT	650	46	46	40	40
Energy (kWh)	156	13	13	9.6	9.6
MBWT	10	55	55	12	10

provides data on energy consumption without baggage retrieval operations, the NbUT AGV, and the MBWT for Scenario 5.

4.3.6. Scenario 6: Variable Baggage Flows with One Converging and One Diverging Peak Period

Scenario 6 is very similar to Scenario 2, except that the second peak period is represented by a divergent value (λ = 0.6), meaning it exceeds the highest baggage handling rate (λmax = 0.5). This scenario focuses on Strategies 1 and 5, specifically analyzing how varying the entry delay time (60 s, 120 s, 180 s) into the AGV parking area impacts Strategy 5 (see the results in Figure 10).

Table 7. Energy usage in the absence of baggage retrieval of Scenario 6.

Strategies	Strategy 1	Strategy 5 (60 s)	Strategy 5 (120 s)	Strategy 5 (180 s)
NbUT	550	101	121	137
Energy (kWh)	132	24.24	29.04	32.88
MBWT	130	90	65	50

outlines the energy consumption, the NbUT per AGV, and the MBWT observed in Scenario 6, excluding baggage retrieval operations.

4.4. Discussion

The ARTEMIS architecture introduces several features that distinguish it from conventional baggage handling systems, especially in terms of efficiency, energy consumption, and operational effectiveness.

Unlike traditional BHSs that rely on static rules and limited feedback, ARTEMIS employs simulation-driven strategies to optimize baggage flow. It integrates MATLAB and Simulink simulations to test performance under various failure and load conditions, allowing for proactive bottleneck identification and scenario planning. A distinctive feature is ARTEMIS’s suite of AGV dispatching strategies—such as Turnstile, On-Demand, Delay Parking, Needs Prediction, and Mixed Advance/Delay—which enable dynamic adjustment to real-time baggage demand. These strategies allow for ARTEMIS to minimize queue wait times and improve throughput efficiency more effectively than fixed-route AGV systems typically used in current BHSs.

The Turnstile Strategy (red color) is ideal in terms of the number of pieces of baggage and waiting time in the queue, as shown in Figure 5c, Figure 6c, Figure 7c, Figure 8c and Figure 9c and Figure 5d, Figure 6d, Figure 7d, Figure 8d and Figure 9d in Scenarios 1 to 5. On the other hand, regarding energy consumption, it is quite high due to the number of unnecessary turns it makes, as shown in Figure 5a, Figure 6a, Figure 7a, Figure 8a, Figure 9a and Figure 10a in all Scenarios 1 to 6.

In this context, the On-Demand Strategy (blue color) is proposed to reduce energy consumption, specifically by minimizing unnecessary turns, as shown in Figure 5a, Figure 6a, Figure 7a, Figure 8a, Figure 9a and Figure 10a for the six scenarios. However, the number of bags and the waiting time in the queue have increased significantly.

The Delay Parking Strategy (green color) is proposed in the context of allowing for some time (60 s) to collect all the baggage in the queue before the AGVs enter the parking lot while implementing an optimal strategy for the AGVs on the circuit. The number of AGVs is adjusted according to the baggage flow, like the On-Demand Strategy. This strategy helps to reduce energy consumption and improves the number of baggage items and the waiting time in the queue, as shown in Figure 5c, Figure 6c, Figure 7c and Figure 8c and Figure 5d, Figure 6d, Figure 7d and Figure 8d for Scenarios 1 to 4. On the other hand, for Scenario 5, there is no improvement due to the low flow of baggage at the beginning.

The Needs Prediction Strategy (black color) anticipates baggage arrival, reducing queue waiting times while keeping the number of AGVs optimal, as in the On-Demand Strategy. Figure 5c, Figure 6c, Figure 7c, Figure 8c and Figure 9c and Figure 5d, Figure 6d, Figure 7d, Figure 8d and Figure 9d in scenarios 1 to 5 show that the number of baggage items and the waiting time in the queue are too low compared to the On-Demand Strategy and the Delay Parking Strategy.

To achieve better conditions, we combined two strategies, the Delay Parking Strategy and the Needs Prediction Strategy, which together form the Mixed Advance/Delay Strategy (shown in pink). As demonstrated in Figure 5c, Figure 6c, Figure 7c, Figure 8c and Figure 9c and Figure 5d, Figure 6d, Figure 7d, Figure 8d and Figure 9d, the number of bags and the waiting time in the queue are like those of the Turnstile Strategy but with lower energy consumption, as illustrated in Figure 5a, Figure 6a, Figure 7a, Figure 8a and Figure 9a for scenarios 1 to 5. The Needs Prediction Strategy is quite sufficient for Scenarios 4 and 5, as shown in the figures, yielding the same results as the Turnstile Strategy and the Mixed Advance/Delay Strategy, since the flow of baggage is very low at the beginning.

Table 2, Table 3, Table 4, Table 5 and Table 6 demonstrate that Strategy 5 offers a good compromise between minimizing the energy consumption of AGVs operating without baggage retrieval and reducing the maximum baggage waiting time in the queue.

In Scenario 6, we present a simulation with a baggage flow value that diverges from the numerical study. We observe that the strategy with the best performance has a delay of 180 s instead of 60 s, as 60 s is insufficient (60 s for the other Scenarios 1 to 5). The number of bags and the waiting time in the queue are better compared to the Turnstile Strategy, even considering energy consumption. On the other hand, the Turnstile Strategy recovers more than 100 pieces of baggage, as shown in Figure 10e.

5. Simulation of AGV Failures and Operator Interventions

5.1. Different Types of AGV and Operator Faults During the Baggage Handling Process

The study of system degradation mechanisms in AGVs used in circuit applications involves analyzing various factors that contribute to performance decline, inefficiency, or errors, faults, and failures over time [66]. The following table presents the different types of faults during AGV circulation (

Table 8. Types of AGV failures.

Type of Failures	Sub-Type of Failures	Description
Mechanical Degradation	Wear and Tear on Components Misalignment Chassis Fatigue	Continuous use of AGVs can result in wear on wheels, bearings, and drive systems. Over time, components such as steering systems may lose precision due to vibration and mechanical stress. Structural components may develop cracks or deformations because of repeated loads or impacts.
Electrical and Electronic Failures	Battery Degradation Sensor Failures Control Board Malfunctions	Batteries gradually lose capacity and efficiency over time due to repeated charge–discharge cycles and exposure to thermal stress. Sensors (e.g., LIDAR, proximity sensors) may degrade because of dirt accumulation, miscalibration, or prolonged exposure to harsh environmental conditions. Electronic components are susceptible to issues such as overheating, corrosion, or interference from electronic noise.
Software and Control Systems	Aging of Algorithms Communication Failures Obsolescence	Control algorithms may become less effective as the physical characteristics of the AGV change over time. Signal degradation or interference can cause delays or result in a loss of control. Older software may be incompatible with newer hardware or updates in the operating environment.
Environmental Factors	Dust and Dirt Accumulation Temperature Extremes Humidity and Corrosion	Affects moving parts, sensors, and cooling systems. Can lead to material fatigue, lubricant breakdown, and electronic failures. Particularly impacts electrical contacts and exposed metal surfaces.
Operational Factors	Overloading Frequent Stops and Starts Improper Maintenance	Operating beyond the design capacity accelerates wear and stresses the system. This increases wear on brakes and motors. The lack of regular inspections or incorrect servicing can lead to accelerated degradation.

).

In a real-time baggage handling system in an airport, the operator’s role is essential in minimizing baggage delays and ensuring the smooth operation of the system.

Table 9. Types of operator failures in baggage pick-up or drop-off areas (zone R or zone D).

Type of Failures	Sub-Type of Failures	Description
Response Time Failures	Delayed Response to Local Failure Delayed Response to Global Failure	Operator fails to quickly respond to a malfunction in either zone R or zone D, leading to delays in baggage processing. Longer delays in response result in accumulating baggage, worsening overall system efficiency. Both zones (R and D) face simultaneous failures, and the operator does not react quickly enough, causing compounded delays across the entire BHMS.
Decision-Making Failures	Incorrect Diagnosis Failure to Prioritize Correctly	Operator misidentifies the source of the failure, causing the wrong course of action to be taken, which prolongs the delay and worsens the overall baggage handling process. Operator is unable to determine which zone (R or D) should be addressed first, leading to poor prioritization and extended delays in baggage handling.
Communication Failures	Lack of Coordination Between Zones Failure to Use Automated Alerts	Operator fails to communicate effectively with other operators working in different zones (R or D), causing delays and confusion in resolving system malfunctions. Operator ignores or fails to act on system alerts or diagnostic tools, which could help quickly identify and resolve issues, leading to unnecessary delays.
Stress/Pressure Failures	Impaired Decision-Making Under Pressure Failure to Manage Multiple Issues Simultaneously	Operator, under the pressure of time or a large failure scenario, makes rushed or incorrect decisions that extend the duration of the failure and lead to further baggage delays. In the event of a global failure, the operator struggles to resolve issues in both zones (R and D) simultaneously, leading to an inability to restore smooth baggage flow efficiently.

outlines different types of operator failures that can affect the performance of the system, particularly during local and global failures. These failures can lead to significant delays, inefficiencies, and disruptions in baggage handling if not managed effectively.

• Response Time Failures: The operator’s ability to quickly react to local or global failures determines how long baggage delays last. A slow response will result in increased baggage accumulation and longer processing times, affecting the efficiency of the baggage handling system.
• Decision-Making Failures: Quick and accurate decision-making is vital for determining the best course of action during failures. Misidentifying the cause of a failure or poor prioritization can exacerbate the issue and cause prolonged delays in baggage processing.
• Communication Failures: Effective coordination between operators or operators and AGVs in different zones is essential to resolve issues quickly. Failure to communicate or use automated system alerts can result in mismanagement, adding to delays and complicating the overall baggage handling process.
• Stress/Pressure Failures: The ability to manage stress and make calm, informed decisions under pressure are key for an operator. If an operator fails to effectively handle multiple issues simultaneously, the delays can become more extensive, worsening the baggage flow and overall system performance.

In this work, we are interested in failures (or breakdowns) occurring at baggage pick-up points or drop-offs. We distinguish two types of failures:

• Local failure: This occurs when one of the pick-up points or the depots fails but not at the same time.
• Global failure: This occurs when two pick-up points, both depots, or a pick-up point and a depot point fail simultaneously.

5.2. Numerical Simulations of the Failed System Components

Simulations conducted in MATLAB and Simulink allow for the system to simulate failures, evaluate response times, and assess the impact of these failures on baggage flow (Figure 11 and Figure 12). Additionally, simulations enable the visualization of effects based on the duration and type of failure, whether global or local, as illustrated in Figure 13, Figure 14 and Figure 15.

All simulations are run over a period of one hour using MATLAB and Simulink.

5.2.1. Local Failure

In this category, we are dealing with a local failure, which can occur either in zone R or zone D. The quantity of baggage delayed at both points—pick-up and drop-off—is illustrated in Figure 11, which features two pie charts. In these charts, number (1) represents zone R, and number (2) represents zone D. This situation is time-dependent, as the duration of the failure relies on how long it takes for a human to resolve the issue.

5.2.2. Global Failure

Unlike local failure, a global failure occurs when issues arise simultaneously at both zone R and zone D. The quantity of baggage delayed at these two points—collection and deposit—is represented in Figure 12, which features two pie charts. In these charts, number (1) corresponds to zone R, and number (2) corresponds to zone D. Like a local failure, the duration of global failure depends on how long it takes for a human operator to resolve the issue. However, in this case, if the operator takes more time to address the problem, it will result in even greater delays in the processing of baggage.

5.3. Discussion on the Numerical Simulation of System Failure

Figure 13, Figure 14 and Figure 15 illustrate the percentage of baggage lost at collection and drop-off points relative to the failure duration. The first figure (Figure 13) simulates a local failure based on its duration; unlike the global outage, if the pick-up or drop-off points operate for only half the time (i.e., 30 min), the loss at these points is reduced to less than a quarter of the merchandise, specifically, only 22% is lost at the various points.

The two other figures (Figure 14 and Figure 15) simulate a global failure as a function of its duration. Figure 14 shows an AGV or baggage pick-up or drop-off points breaking down over a period ranging from 1 min to 30 min, while Figure 15 shows two AGVs or two baggage pick-up or two drop-off points breaking down at different times and levels, alternately, during continuous periods. These two figures indicate that the breakdowns depend solely on time, not on the position of the AGVs or the baggage pick-up or drop-off points. Since the simulation spans an hour, if the breakdown lasts for half of the time (i.e., 30 min), nearly half of the goods are lost at the pick-up and drop-off points (i.e., 50%), as shown in Figure 14 and Figure 15.

The duration of the breakdown depends on how long the operator takes to resolve the issue at the collection and/or drop-off points as well as the location of these points in relation to the office. For example, the operator will typically take less time to access the drop-off points than the collection points.

The role of the operator is central to minimizing delays and ensuring the smooth operation of the system. Both local failures and global failures within the system depend heavily on the operator’s response time, decision-making, and overall efficiency. These failures can occur in critical zones such as zone R (pick-up) and zone D (drop-off), and their impact can significantly disrupt the entire baggage processing flow.

In a local failure, problems arise within either zone R or zone D but not simultaneously in both. These zones are key to the movement and sorting of baggage, and even small disruptions in these areas can lead to delays in processing. The operator’s reaction time is critical in local failures. When an issue is detected in either zone R or zone D, the operator must act swiftly to identify the root cause and resolve the problem. In a well-designed system, computational support (like automated alerts) can assist the operator by highlighting where the failure occurred. However, the operator’s timely intervention is essential to minimize delays. The longer the operator takes to react, the more baggage will accumulate in the affected zone, leading to extended delays and potentially complicating the baggage sorting process. In addition to speed, the operator’s decision-making is essential. The operator must quickly determine whether to reroute baggage, reset machinery, or engage manual intervention to clear the blockage. Since baggage processing in a real-time system is complex and time-sensitive, the operator’s ability to make the right decision at the right time is key to preventing a small issue from escalating into a larger delay.

A global failure occurs when problems arise in both zone R and zone D simultaneously. This type of failure is more complex because it involves disruptions at two critical points in the baggage handling process, which can have a much greater impact on overall system efficiency. In the event of a global failure, the operator’s response time becomes even more critical. Since both zones are affected, the operator must address issues across multiple points in the system, which increases the complexity of the situation. With the right computational support, operators can be provided with real-time diagnostic information to identify the nature of the failure in both zones. However, operator efficiency is key to preventing a more severe backlog from developing. A delay in addressing issues in either zone can lead to prolonged delays in baggage processing, increasing the time it takes for baggage to reach its destination. The operator’s decision-making is also paramount in a global failure scenario. The operator must assess whether one zone requires more immediate attention than the other and prioritize actions accordingly. A failure in zone R might take precedence if it is affecting the timely departure of flights, while a failure in zone D could be prioritized if it is preventing baggage from being correctly sorted. However, failure to prioritize effectively could result in significant delays and more serious consequences for baggage handling, further complicating the overall system’s performance. In global failures, the impact on the baggage handling system can be far-reaching. Since both zones are disrupted, delays in processing baggage can increase exponentially if the operator fails to resolve the situation quickly. If system alerts and computational support are not efficiently used, the operator’s slow response time can cause significant disruptions, affecting the baggage processing for multiple flights and possibly leading to missed connections or customer dissatisfaction.

The computation system should assist the operator by suggesting priority actions or providing alternative solutions to minimize downtime and ensure smooth operation across both zones. Despite computational assistance, the operator’s actions in responding to failures (both local and global) are critical in preventing further delays. The system may be designed to assist the operator, but it is their response time and decision-making abilities that determine how quickly baggage delays are minimized. Slow or incorrect actions on the part of the operator can lead to compounding delays, affecting the entire baggage flow through the airport.

We previously observed that the human operator’s reaction time is crucial in the event of a local failure in the baggage handling system. This reaction time becomes critical in the case of a global system failure. One way to improve the operator’s responsiveness is through a baggage tracking system. While routing is essential in baggage management, tracking also plays a central role, as it enables tracing and archiving the history of a bag’s journey. This provides valuable support for maintaining the operational condition of the baggage management system, especially in cases where delays are caused by the human operator. It is, therefore, essential that all resources and organizational measures of the baggage routing system be implemented before deployment to assist the operator in their task, ensuring optimal responsiveness in the event of a failure. Consequently, automated baggage tracking, due to its ability to prevent failures caused by insufficient operator responsiveness, holds significant potential for optimizing baggage management and ultimately resolving this category of problems in airports. Furthermore, predictive maintenance methods, such as observers [67] or Bayesian techniques [68], could be used to strengthen the prevention of faults and their effective detection and localization.

6. Conclusions

This study presents a real-time baggage handling monitoring system using a computational ergo-design approach, leading to the development of the ARTEMIS architecture. By modeling the structure of the baggage handling system as a directed graph and then running simulations, five baggage handling strategies were evaluated to determine their effectiveness in balancing efficiency, queue times, and energy consumption. The results indicate that the Mixed Advance/Delay Strategy offers the best overall performance, maintaining low energy consumption while ensuring smooth baggage flow and relatively short waiting times. On the other hand, the Turnstile Strategy remains the best choice when the main goal is to minimize queues and maximize baggage collection, even though it requires more energy.

Among the other strategies, the On-Demand Strategy saves energy but results in longer queues and more baggage buildup. The Delay Parking Strategy finds a middle ground between energy use and queue management, while the Needs Prediction Strategy does a good job of reducing waiting times but is not as effective at handling baggage. Interestingly, in Scenario 6, increasing the delay time outperformed the Turnstile Strategy.

To better assess the performance of these strategies, a simulation environment was developed to generate key indicators, enabling the preliminary visualization and analysis of AGV behavior in predefined scenarios. The results are presented through an intuitive and ergonomic user interface, designed with a strong focus on human–computer interaction as a user-centered, problem-solving process.

Beyond the system itself, this study also highlights the important role of the human operator in ensuring the smooth functioning of baggage handling processes. Simulations show that an operator’s reaction time and decision-making skills significantly influence overall system performance. Any delay in responding to issues can cause baggage pileups, leading to disruptions across the entire airport. Therefore, the implementation of a well-designed support system is essential to assist operators in making timely and effective decisions.

Unlike traditional optimization methods, which typically focus on efficiency and performance without integrating human factors, computational ergo-design emphasizes the interaction between humans, technology, and infrastructure. It continuously integrates human factors into the design process, ensuring that systems are not only effective but also aligned with human needs and behavior. By incorporating real-time data processing and feedback, it enables dynamic adaptation to changing conditions and evolving human interactions. This approach accounts for both human behavior and system feedback as well as environmental factors, facilitating ongoing optimization. While traditional optimization methods may lack this dynamic, human-centered approach, computational ergo-design brings a more comprehensive and adaptable solution for complex transport systems.

Future research could explore ways to improve operator decision-making through intelligent agents (mainly with reinforcement learning-based techniques) and automation, helping to make baggage handling even more efficient.