A Hybrid Method to Solve the Multi-UAV Dynamic Task Assignment Problem
Abstract
:1. Introduction
- Clustering for Large-Scale Environments: We use a clustering approach that segments the operational environment into smaller, manageable areas and distributes UAV resources in a balanced manner based on the task density within each cluster. Unlike existing global assignment approaches that allocate UAVs without considering localized workload variations, our method ensures that task allocation is performed locally within each cluster in a decentralized manner. This significantly improves scalability by reducing the computational overhead, and enhances responsiveness by enabling the UAVs to focus on nearby tasks, reducing travel time and improving real-time adaptability.
- Partial Reassignment Strategy for Fast and Adaptive Task Handling: The proposed solution employs a partial reassignment strategy that selectively engages only the nearest subset of UAVs when a dynamic event occurs. These UAVs are released from a portion of the farthest assigned tasks, prioritizing proximity to the new task to minimize travel delays and maximize reassignment efficiency. Unlike traditional reassignment approaches that either reset all task assignments (e.g., the Consensus-Based Bundle Algorithm (CBBA) full reset) or rely on static heuristics, our method ensures that reallocation is both targeted and efficient, reducing unnecessary disruptions. This enhances robustness, allowing UAVs to adapt quickly to dynamic changes, while maintaining high operational continuity and minimizing interruptions to ongoing tasks.
- Intelligent Idle UAV Management for Efficient Resource Utilization: The proposed solution introduces a centralized idle UAV management mechanism to ensure efficient resource utilization and maximize task coverage. When a UAV becomes idle, the central controller first checks for any unassigned tasks and directs the UAV to handle them immediately. If no unassigned tasks exist, the controller reallocates the idle UAV to assist the most heavily loaded UAV within its cluster, ensuring workload balancing. If the cluster is fully covered, the idle UAV is relocated to the nearest uncovered cluster to minimize travel time and enhance mission efficiency. Unlike existing methods that either leave idle UAVs underutilized or lack a structured relocation strategy, our approach proactively redistributes UAVs based on real-time task demand. This optimizes resource allocation, accelerates task completion, and enhances overall system scalability and responsiveness in dynamic environments.
- Hybrid Centralized-Distributed Communication for Efficient Coordination: The proposed communication topology optimally balances centralized coordination with distributed execution, significantly reducing communication overhead. Within each cluster, UAVs exchange data using a distributed structure, leveraging decentralized task assignment algorithms such as CBBA, Performance Impact Algorithm (PI), or similar approaches to minimize excessive inter-UAV communication. At the same time, the central controller manages inter-cluster coordination, ensuring seamless high-level decision-making, while restricting unnecessary global data exchange. This dual-layer approach accelerates responses to ongoing tasks and dynamic events, avoiding the bottlenecks associated with fully centralized methods, while maintaining better organization and efficiency than purely decentralized systems. As a result, the proposed communication strategy enhances responsiveness, reduces delays, and improves overall mission execution in complex operational environments.
2. Literature Review
2.1. Market-Based Algorithms
2.2. Intelligent Optimization Algorithms
2.3. Clustering-Based Algorithms
3. Problem Formulation
4. Proposed Solution
4.1. Description
- Appearance of new tasks: When a new task emerges and is detected by a UAV, the central controller manages the reallocation process among neighboring UAVs within the same cluster. The first step in this process is to check if there are idle UAVs available within the cluster. If an idle UAV is found, it is immediately assigned to the new task without disrupting the ongoing operations. This ensures a quick response to dynamic events, while minimizing the need for task reassignments.If no idle UAVs are available, a partial reassignment process is triggered to accommodate the new task. In this step, the central controller selects a sub-team of UAVs that are closest to the newly detected task to participate in the reassignment process. Each selected UAV then releases a subset of its assigned tasks, specifically those that are farthest from its current location and make the lowest contribution to the global objective. By releasing only a portion of their workload, these UAVs free up resources to accommodate the new task without causing a complete reset of assignments, ensuring continuity in ongoing operations.Once the tasks have been released, the participating UAVs execute a decentralized CBBA task allocation algorithm locally within the cluster. During this process, the UAVs engage in distributed communication and progressively share and update their decisions until a final allocation has been reached. This localized decision-making ensures rapid task allocation, while avoiding unnecessary global communication overhead.Once the decentralized assignment process has converged, the participating UAVs update their task lists and resume their operations. The central controller oversees the process but does not intervene, allowing the UAVs to coordinate and make local decisions, while maintaining efficiency and scalability.As illustrated in Figure 4, when three new tasks are detected in different clusters (green, blue, and yellow), the system measures distances between the UAV locations and the detected tasks locally within each cluster. The two nearest UAVs (np = 2) in each cluster, represented by solid red lines, are selected for reassignment. Each UAV then releases two of its farthest assigned tasks (nr = 2), before the locally distributed assignment algorithm reallocates both the released tasks and the new task within the cluster. This approach ensures fast responsiveness to dynamic events, minimizes task disruptions, and optimizes the workload distribution within the cluster.
- UAV failure: In the event of a UAV failure, the central controller initially manages the reassignment process by assessing all unperformed tasks previously assigned to the failed UAV. If an idle UAV is available, the central controller immediately assigns it to take over the tasks, ensuring continuous operation, without triggering a reassignment process. However, if no idle UAV is available, the unperformed tasks are treated as new tasks and integrated into the Partial Reassignment Process, which operates in a distributed manner within the failed UAV’s cluster.At this stage, a subset of UAVs closest to the failed UAV’s last known position are selected to participate in the reassignment. These UAVs independently release a portion of their assigned tasks, prioritizing those that contribute the least to the global mission objective or are geographically distant, to accommodate the newly assigned tasks. The distributed CBBA assignment algorithm is then executed locally among the participating UAVs, allowing them to autonomously exchange task availability, workload status, and bid values. This decentralized decision-making ensures that tasks are efficiently reallocated, without requiring intervention from the central controller. By combining centralized intervention for failure detection and initial task management with a distributed reassignment process within clusters, the system demonstrates a hybrid nature, ensuring scalability, responsiveness, and resilience in dynamic environments. Figure 5 shows a failed UAV in the green cluster and another failed UAV in the yellow cluster. All of their unperformed tasks (red points) are considered new tasks.
- UAV Idleness: In a dynamic operational environment, a UAV may enter an idle state after successfully completing all its assigned tasks. To ensure optimal resource utilization and task coverage, the central controller manages idle UAVs through the following steps:
- Immediate Task Allocation: If a new task is detected or unassigned tasks remain, the central controller immediately assigns the idle UAV to these tasks, ensuring a fast response, without requiring task reassignment from active UAVs.
- Intra-Cluster Assistance: If there are no new or unassigned tasks but the current cluster still requires coverage, the central controller directs an idle UAV to assist the most heavily loaded UAV in its cluster. This ensures efficient workload balancing within the cluster, optimizing resource utilization, without unnecessary UAV downtime. As shown in Figure 6, the idle UAV in the blue cluster (highlighted circle) is directed by the central controller to support the busiest UAV in the same cluster.
- Inter-Cluster Migration: If all tasks within the cluster are fully addressed, the central controller directs the idle UAV to migrate to the nearest uncovered cluster that requires additional support. As illustrated in Figure 6, an idle UAV from the yellow cluster moves to the pink cluster, ensuring continuous coverage where UAVs are needed.
This centralized resource redistribution allows UAVs to maximize task coverage and improve overall mission efficiency in large-scale environments.
4.2. Algorithms
Algorithm 1 Central Controller |
Input: N: Number of UAVs.
|
|
Algorithm 2 Initial . Offline Tasks-UAVs Assignment |
Input: N: Number of UAVs.
|
|
Algorithm 3 Fully . Distributed Dynamic Task Execution |
Input:
|
|
Algorithm 4 Generate Tasks |
Input:
|
|
Algorithm 5 Initialize UAVs and Sensors |
Input:
|
|
Algorithm 6 Balance UAVs Across Clusters |
Input:
|
Step 1: Compute the Initial Number of UAVs Per Cluster
Step 2: Balance and Distribute Extra UAVs Among Clusters
|
Algorithm 7 Dynamic Event Handler |
Input:
|
|
Algorithm 8 Partial Reassignment Process |
Input:
|
|
Algorithm 9 Consensus-Based Bundle Algorithm (CBBA) for Distributed Task Assignment |
Input:
|
|
4.3. Novelty of the Proposed Solution
- Communication Bandwidth: As a dynamic assignment problem, the interaction and information sharing among UAVs must be rapid and restricted. Therefore, the proposed solution establishes a centralized-distributed communication topology connecting UAVs both within and between clusters. This approach serves to reduce communication overhead, reduce delay, and ultimately enhance overall performance.
- Communication Inside Clusters: The proposed solution employs a distributed communication strategy within clusters to ensure efficient task allocation and minimal network overhead. This occurs in two decentralized scenarios, where a distributed CBBA assignment algorithm is executed autonomously by the UAVs.
- (a)
- First, before the mission starts, the UAVs within each cluster run the distributed assignment algorithm to generate the offline task allocation. The UAVs communicate locally within their cluster, following the mechanism of the selected distributed algorithm to determine task assignments. Once the allocation has been finalized, the UAVs operate autonomously and independently, without requiring further coordination, as shown in Figure 7.
- (b)
- Second, when a dynamic event occurs (e.g., new task, UAV failure), the central controller selects the nearest UAVs, which then execute the distributed CBBA assignment algorithm to reassign both the new and released tasks. Only participating UAVs communicate within their cluster, keeping the process of task reassignment fully decentralized and efficient, as shown in Figure 8.
- Communication Among Clusters: Beyond the decentralized task execution within clusters, the system remains centralized for mission supervision and inter-cluster coordination, where the central controller oversees operations and triggers reassignment when necessary. This hybrid approach ensures scalability, efficiency, and adaptability by combining decentralized UAV autonomy with centralized decision-making. By restricting inter-cluster communication to the central controller, the system reduces unnecessary UAV-to-UAV communication overhead, enhancing coordination, resource utilization, and overall operational efficiency.
- Handling Dynamic Events: Efficiently managing dynamic events is a critical aspect of our proposed hybrid solution, which is designed to handle the complexities of real-time changes within multi-UAV operations. Upon the detection of a dynamic event, the proposed solution does not require the participation of all UAVs. Instead, it selectively engages a subteam of UAVs that are geographically closest to the new event. This partial participation strategy reduces the disruption to the fleet and maintains ongoing task performance, while swiftly incorporating the new task into the allocation. Furthermore, the proposed solution incorporates a task release mechanism whereby only a portion of the previously assigned tasks are released during reallocation. This approach is particularly useful when integrating new tasks or redistributing the load due to a UAV failure. By releasing only tasks that are less critical or farther from the UAVs’ current positions, we minimize the impact on overall mission performance and focus resources on time-sensitive tasks. Overall, this strategy enhances the responsiveness of the UAV fleet to dynamic events, strengthens the robustness of the solution, and adapts rapidly and effectively to changing operational conditions, without significant disruptions (Figure 9).
- Large-scale Problem Instances: The proposed solution excels in managing large-scale problem instances, which are particularly challenging due to the extensive coordination and computation needed. By employing a sophisticated clustering approach, the algorithm segments the operational area into smaller manageable units, each handled by a subset of UAVs. Clustering ensures a more organized deployment and optimizes the coverage and responsiveness of the UAV network, which will be highly effective in extensive and complex scenarios. By focusing on local clusters, while maintaining a global overview through centralized control, our solution achieves a balanced distribution of tasks, optimal resource usage, and a significant reduction in operational delays, which are critical for scalability and efficiency in large-scale applications.
5. Empirical Evaluation
5.1. Simulation Setup
5.2. Evaluating the Performance of the HCPR Algorithm
- Maximize Responsiveness: This objective ensures that the proposed solution reacts quickly and effectively to start and complete mission tasks. Metrics such as task waiting time and the number of performed tasks were used to assess this responsiveness.
- Increase Scalability: Scalability involves the solution’s ability to manage an increasing number of tasks or to operate effectively across broader geographic areas, without performance degradation. This was measured using UAV throughput and mission completion time under varying instance sizes.
- Improve Solution Robustness: This objective assesses the solution’s ability to adapt to unexpected operational changes, such as new tasks or UAV failures, while maintaining consistent performance. Robustness was quantified by the number of detected and covered dynamic tasks.
- Waiting Time: Time elapsed before a UAV begins a task.
- UAV Throughput: Total reward for tasks completed by UAVs, adjusted for time discounting, where the rewards decrease with late task execution times.
- Number of Performance Tasks: Total number of tasks completed by the UAV fleet.
- Number of Covered New Tasks: Number of new tasks detected and performed during the mission.
- Completion Time: Total time taken to complete all allocated tasks.
Result Analysis
- Solution Responsiveness: We evaluated the solution’s responsiveness using the waiting time and the number of performed tasks metrics, where the waiting time reflects the speed at which UAVs begin performing tasks and the number of performed tasks indicates the solution’s efficiency in handling tasks within specific time frames and underscores its reactive capabilities.
- (a)
- Average Task Waiting Time:As shown in Figure 10, on the small dataset, the CBBA generally exhibited lower average waiting times when the number of UAVs was on the lower end (5 to 7 UAVs). However, as the number of UAVs increased, the performance of HCPR relative to the CBBA improved, particularly at higher UAV counts (12 and 15). This indicates that HCPR’s response improved with a larger number of UAVs. An increase in the map size did not have a consistent impact on the performance of either algorithm, since the UAV velocity is proportional to the map size, as described above. For the large dataset (Figure 11), HCPR consistently showed lower average waiting times across almost all scenarios, particularly as the number of UAVs increased. HCPR demonstrated remarkable responsiveness in large-scale scenarios, and improved its performance advantage as the scale of operations increased. To justify and analyze how HCPR outperformed the CBBA, several reasons can explain the enhanced efficiency and reduced waiting times observed in Figure 10 and Figure 11:
- Geographical Distribution of UAVs: Distributing UAVs among different clusters allows tasks throughout the area to have an equal opportunity for earlier execution, thereby reducing their waiting times. In contrast, with the CBBA, where UAVs are concentrated around the base station, tasks located in distant areas experience longer waiting times and may expire before being reached by any UAV.
- Handling Dynamic Events with Nearby UAVs: Assigning a small team of nearby UAVs to address dynamic events can significantly reduce task waiting times. This approach speeds up the reassignment process by involving only a limited number of UAVs and selecting those closest to the event and minimizing travel distances. Furthermore, by partially releasing only a small set of the farthest tasks for the reassignment process, this strategy shortens the reassignment duration and allows new tasks, which are nearer to the detected UAVs, to be executed sooner, with reduced waiting times. It also gives these distant tasks an opportunity to be reassigned to UAVs that are closer, potentially also accelerating their execution. This also ensures that closer original tasks maintain their scheduled earlier execution times. Conversely, the CBBA’s method of reallocating all tasks and involving all UAVs leads to work interruptions and extended delays in task execution, thereby increasing waiting times.
- Utilization of Idle UAVs: Employing idle UAVs to handle new tasks or assist others within the same or neighboring clusters enhances the speed of task execution and reduces waiting times, ensuring a more efficient use of available resources.
- Communication Restriction: Limiting communication to only involved UAVs and central control within and between clusters helps maintain task execution, without delays caused by dynamic events. This approach allows most UAVs to continue their tasks uninterrupted and leads to earlier completion and reduced waiting times. In the CBBA, however, the requirement for all UAVs to communicate after detecting dynamic events disrupts ongoing tasks and results in prolonged waiting times.In the case of 55 UAVs, where HCPR exhibited longer waiting times than the CBBA, it is important to note that HCPR covered more tasks than the CBBA. The shorter waiting times reported by the CBBA were due to its inability to address tasks before they expired, resulting in fewer completed tasks overall.
- (b)
- Average Number of Performed Tasks: For small instances, as shown in Figure 12, both algorithms demonstrated an increased number of performed tasks as the number of UAVs increased, but HCPR’s growth rate and overall numbers remained superior. This suggests that HCPR’s task allocation strategies are more responsive, particularly as resources scale up. For the large dataset, shown in Figure 13, the advantage of HCPR over the CBBA became even more pronounced on the large dataset. As the map size and UAV count increased on the large dataset, HCPR maintained its lead in the number of tasks completed across both the small and large datasets. This consistency reinforces the responsiveness and adaptability of HCPR in dynamic and potentially unpredictable environments.The following outlines why HCPR outperformed the CBBA in terms of the number of tasks performed (as shown in Figure 12 and Figure 13):
- Balanced Task Distribution: HCPR distributes UAVs evenly across clusters and ensures that even tasks in remote areas are completed before expiration, which leads to more tasks being performed. In contrast, the CBBA’s focus near the base station often leads to distant tasks expiring.
- Efficient Task Reassignment: HCPR utilizes a small team of UAVs for partial reassignment processes and releases only a minor portion of their original tasks. This enables these UAVs to complete the reassignment quickly and resume their primary tasks sooner, thus executing more tasks overall. In contrast, the CBBA’s comprehensive reassignment approach involves all UAVs and their tasks, resulting in delays and fewer completed tasks.
- Utilization of Idle UAVs: HCPR actively engages idle UAVs in handling new tasks or assisting current or nearby clusters and, as a result, the task completion rates are enhanced. The CBBA, however, restricts UAVs to their assigned task lists and neglects idle UAVs, which can lead to the expiration of many tasks.
- Communication Efficiency: HCPR restricts communication to selected participant UAVs and the central controller, minimizes task interruptions, and allows more tasks to be completed in a timely manner before the end time. The CBBA’s extensive communication requirements for all UAVs during dynamic events cause task execution delays and higher expiration rates.
- Solution Scalability: Scalability was evaluated by measuring the UAV throughput and completion time. Throughput, which accounts for time discounting, reflects the solution’s capacity to complete tasks efficiently, even as the problem scale increases, while completion time reveals the solution’s ability to finalize the mission swiftly and indicates its ability to efficiently scale up to larger or more complex operations.
- (a)
- Average UAV Throughput: For small instances, as shown in Figure 14, the throughput advantage of HCPR over the CBBA was clear across all tested scenarios. As the number of UAVs increased, the throughput for both algorithms increased. However, HCPR maintained a higher margin over the CBBA, which indicates that HCPR’s task allocation and execution strategies scale more effectively with increased resources. In the context of large instances (Figure 15), HCPR showed consistent performance improvements over the CBBA across different map sizes and UAV numbers. HCPR also not only demonstrated a higher efficiency in task execution, but also superior scalability. As the number of UAVs increased, HCPR was able to leverage the additional resources more effectively than the CBBA, leading to a significantly higher throughput in both small and large operational scenarios and suggesting that HCPR is particularly well-suited for large-scale, dynamic environments.Many factors explain why HCPR outperforms the CBBA in the various scenarios (Figure 14 and Figure 15):
- Balanced Distribution and UAV Throughput: Distributing UAVs across different areas in a balanced manner positively impacts throughput. This is because tasks located further away receive timely attention before they expire, which ensures maximum reward collection. On the other hand, the CBBA concentrates UAV activity near the base station, which leads to delayed execution and reduced rewards for distant tasks, ultimately resulting in lower throughput, due to tasks expiring unattended.
- Efficiency in Dynamic Event Handling: HCPR consistently outperformed the CBBA by selecting groups of nearby UAVs for reassignment processes, which leads to quicker negotiations and reduced travel times. This efficient management allows these UAVs to resume their tasks sooner and enhances the overall throughput. Furthermore, UAVs not involved in the event continue their tasks uninterrupted and avoid delays that negatively impact throughput. Additionally, by releasing only the farthest portion of the original assignments, the reassignment process is accomplished more quickly, further increasing the UAV throughput.
- Utilization of Idle UAVs: Implementing an idle UAV handler that fully utilizes all available UAVs accelerates the completion of missions, increases the number of tasks performed, and consequently boosts UAV throughput.
- Communication Restriction Benefits: Limiting communication to a few UAVs and the central controller enables most UAVs to complete their tasks with minimal interruptions and leads to greater throughput. This strategy also reduces waiting times that adversely decrease task rewards, thereby enhancing overall operational efficiency.
- (b)
- Average Completion Time: Across all map sizes and numbers of UAVs on the small datasets, shown in Figure 16, HCPR consistently achieved much lower average completion times than the CBBA. As the number of UAVs increased, the completion time for both algorithms increased, due to the complexity of coordinating more agents. However, the increase in the completion time of HCPR was much less pronounced than that of the CBBA, indicating that HCPR has superior scalability and task management efficiency. On large datasets, as shown in Figure 17, the efficiency gap between HCPR and the CBBA widened significantly. HCPR significantly enhanced the operational efficiency and enabled much faster mission completions across the various scales and complexities. This efficiency is crucial for dynamic environments, where timely task execution is essential.To explain why HCPR outperformed the CBBA in terms of mission completion time, as shown in Figure 16 and Figure 17, the following streamlined justifications are used:
- Localized Management through Clustering: HCPR divides the environment into manageable clusters for localized control and faster completion times. In contrast, the CBBA treats the environment as a single unit, which is less efficient and more time consuming.
- Efficient Task Reassignment: HCPR uses a small subgroup of UAVs for quick partial reassignments and releases only a minor portion of their tasks, which greatly speeds up the completion process. The CBBA, however, involves all UAVs and their entire task list in reassignments when new events are detected that lead to prolonged completion times.
- Utilization of Idle UAVs: HCPR actively employs idle UAVs to assist in ongoing missions, significantly reducing completion times. The CBBA lacks a mechanism to utilize idle UAVs effectively; each UAV adheres rigidly to its assigned tasks, which can delay overall mission completion.
- Communication Efficiency: HCPR limits communication to the necessary UAVs within clusters and to a central controller, restricts operations, and reduces mission completion times. The CBBA’s extensive communication among all UAVs leads to delays, requiring more time to complete missions.
- Solution Robustness: Robustness can be measured by the number of new tasks covered, which helps estimate the solution’s robustness and adaptability to unexpected changes and additional demands, reflecting its capacity to integrate and manage new tasks effectively.
- Average Number of Covered New Tasks: For small instances, as shown in Figure 18, both algorithms demonstrated the ability to cover new tasks, with HCPR generally outperforming the CBBA as the number of UAVs increased. This suggests that HCPR has a more effective mechanism for integrating new tasks, due to its dynamic reallocation strategies. For large instances, as shown in Figure 19, the robustness of HCPR became evident, with a significant increase in the number of new tasks covered across all UAV numbers and map sizes, which demonstrates the superior adaptability and robustness of HCPR to dynamic changes.To explain why HCPR outperformed the CBBA in covering new tasks, as shown in Figure 18 and Figure 19, concise explanations are presented as follows:
- (a)
- Even Distribution of UAVs: HCPR’s strategy of evenly distributing UAVs throughout the environment enhances the detection of new tasks across a broader area and prevents task expiration by ensuring timely discovery. In contrast, the CBBA focuses UAV activity near the base station, often resulting in delayed arrival and potential expiration of tasks in distant regions.
- (b)
- Task Reallocation Strategy: Upon detecting new tasks, HCPR instructs selected participating UAVs to release a portion of the farthest tasks. This prioritization ensures that newly detected tasks, being closer to the involved UAVs, are more likely to be allocated quickly, which enhances coverage efficiency.
- (c)
- Utilization of Idle UAVs: HCPR proactively checks for new task lists when UAVs become idle and immediately reallocates these UAVs to cover new tasks. Conversely, the CBBA lacks a specific strategy for utilizing idle UAVs, often leading to new tasks expiring before they can be addressed.
- (d)
- Communication Efficiency: HCPR’s restricted communication protocol allows UAVs to concentrate on covering new tasks with fewer interruptions, leading to more efficient task coverage compared to the CBBA’s approach, which involves extensive communication, which can disrupt task coverage.
5.3. Comparison of the Results with State-of-the-Art Solutions
- Appearance of a new task: First, a new search task, denoted as , was assumed to have emerged in the time-sensitive and dynamic search and rescue (S&R) scenario. The time window (TW) for this urgent new task was set to s. Table 4 shows a comparative analysis of different reassignment strategies against the HCPR algorithm in terms of their respective scores. Notably, the HCPR algorithm outperformed all listed strategies, with a score of 422.9202, which highlights its superior adaptability and optimization capabilities in dynamic environments. The superior score achieved by the HCPR algorithm compared to the other benchmarks can be attributed to several distinct features of its design:
- While the HCPR algorithm includes a clustering mechanism for efficient area division and task management, this feature was not utilized because the dataset, which comprises four UAVs and 14 tasks, was too small, and the area was treated as a single region. Therefore, clustering did not contribute to the higher score in this instance.
- The HCPR strategy allows only a subset of the UAVs nearest to a new task to engage in the reassignment process, which enhances efficiency. However, given that the new task was a search task and that only two UAVs in the dataset were designated searchers, both participated in the process, which made the participant selection aspect irrelevant in this scenario.
- The HCPR algorithm optimizes task reassignment by releasing only those original tasks that are farthest from the new task, thereby minimizing the negative impact on the score function from the travel distance. In contrast to the methodology described in the paper, which selects tasks for reassignment based on overlapping time windows (TWs) with the new task, without considering their proximity, the HCPR algorithm prioritizes the release of tasks that are geographically further away from the new task, thereby ensuring that tasks with greater potential to reduce the score are prioritized for reassignment, significantly contributing to the improved score outcome.
- An additional strength of HCPR is its management of idle UAVs; the algorithm effectively reallocates UAVs that have completed their assignments, to assist others of the same type. This proactive utilization of resources is another key element that boosted the algorithm’s score.
- Different TWs of a New Task : Then, we assumed that the new search task could have varying time windows (TWs), and the score increments were compared for the HCPR algorithm and the other algorithms. As shown in Table 5, the HCPR algorithm, with scores of 10.5163 for TWs and , 20.2231 for TWs and , and 32.5820 for TWs and , significantly outperformed the strategies proposed in other papers. The HCPR showed a progressive improvement as the TW increased, since the new task was identical in location, duration, and reward, differing only in its time window (TW), thereby uniquely influencing the overall performance metric. This divergence is attributed to the algorithm’s sensitivity to task execution timings. The original tasks, which commenced at time zero, experienced delays upon the introduction of an earlier-starting new task, which detrimentally affected this performance metric. Conversely, integrating a new task with a delayed start, while preceding tasks initiated at zero, minimized the impact, preserving or potentially enhancing the performance metric. This dynamic underscores the critical role of task schedule timing in optimizing performance within the proposed framework.
- Continuous Appearance of : One hundred Monte Carlo simulations were conducted to analyze the appearance of five new tasks. The score improvements of HCPR and the other benchmarks are shown in Table 6. The consistent performance of the HCPR algorithm in these simulations highlights its robustness in dynamic scenarios, where multiple new tasks appear continuously. The adaptability of HCPR to efficiently integrate multiple new tasks into the ongoing mission, without significant disruptions to existing operations, contributed to its higher score outcomes. In essence, HCPR’s higher score was largely due to two key factors:
- It employs a task release mechanism that considers the spatial positioning of tasks and favors the reassignment of tasks that are located further away. This reduces the travel distance and enhances operational efficiency.
- The algorithm adopts the reallocation of idle UAVs that have completed their tasks, ensuring full resource utilization and enhancing the overall score value.
- Different Scales of S&R Scenarios: The simulations described previously demonstrated both the feasibility and superiority of the HCPR algorithm compared to the algorithm proposed in the other papers. Subsequently, various scales of S&R scenarios were implemented to assess the scalability of the algorithms. The authors created scenarios with randomly generated original tasks, where is the number of UAVs and is the number of tasks within a 1 km × 1 km 2D mission area. The time windows () for these tasks were randomly selected within seconds. The score improvements from 100 Monte Carlo simulations are illustrated in Table 7. The superior performance of the HCPR algorithm compared to the other algorithms can be attributed to several optimization factors that enhance efficiency and effectiveness:
- (a)
- It incorporates a clustering algorithm that segments the dataset into smaller, more manageable units, which facilitates more effective task reassignment.
- (b)
- In contrast to the CBBA-LR algorithm, which involves all searcher UAVs in reassignment, HCPR selectively engages only a subset of UAVs—those in closest proximity to the new task. This targeted approach reduces the travel distance, thereby enhancing the overall efficiency of the score function.
- (c)
- HCPR strategically releases tasks that are farthest from critical zones and minimizes their impact on the algorithm’s score, unlike other algorithms that may not prioritize tasks based on task proximity.
- (d)
- The algorithm effectively utilizes idle UAVs by redeploying them to assist with tasks in their current or adjacent clusters. This approach optimizes the number of tasks completed, which leads to greater rewards and, consequently, a higher score.
These factors collectively contributed to the elevated scoring metrics of HCPR and reflect a methodical and resource-efficient allocation of tasks within dynamic operational environments.
- UAV Failure: The UAVs depart from the base stations at . At , UAV was compromised by an enemy attack and can no longer perform its tasks. The TS-DTA score () after UAV failure was 2051.28, whereas the HCPR score was 1018.
- The Appearance of a New Task: At , the early warning UAV detected three new tasks. The TS-DTA score () after UAV failure was 2166.98, while the corresponding HCPR score was 1288.7.
6. Conclusions
Author Contributions
Funding
Data Availability Statement
Conflicts of Interest
References
- Fang, B.; Zhang, Q.; Wang, H.; Yuan, X. Personality driven task allocation for emotional robot team. Int. J. Mach. Learn. Cybern. 2018, 9, 1955–1962. [Google Scholar] [CrossRef]
- Palmer, A.W.; Hill, A.J.; Scheding, S.J. Modelling resource contention in multi-robot task allocation problems with uncertain timing. In Proceedings of the 2018 IEEE International Conference on Robotics and Automation (ICRA), Brisbane, Australia, 21–25 May 2018; IEEE: New York, NY, USA, 2018; pp. 3693–3700. [Google Scholar]
- Khamis, A.; Hussein, A.; Elmogy, A. Multi-robot task allocation: A review of the state-of-the-art. In Cooperative Robots and Sensor Networks 2015; Springer: Cham, Switzerland, 2015; pp. 31–51. [Google Scholar]
- Li, Z.; Li, X. Research on model and Algorithm of task allocation and path planning for multi-robot. Open J. Appl. Sci. 2017, 7, 511. [Google Scholar] [CrossRef]
- Korsah, G.A.; Stentz, A.; Dias, M.B. A comprehensive taxonomy for multi-robot task allocation. Int. J. Robot. Res. 2013, 32, 1495–1512. [Google Scholar] [CrossRef]
- Qin, B.; Zhang, D.; Tang, S.; Wang, M. Distributed grouping cooperative dynamic task assignment method of UAV swarm. Appl. Sci. 2022, 12, 2865. [Google Scholar] [CrossRef]
- Yang, M.; Bi, W.; Zhang, A.; Gao, F. A distributed task reassignment method in dynamic environment for multi-UAV system. Appl. Intell. 2022, 52, 1582–1601. [Google Scholar] [CrossRef]
- Saif, A.; Dimyati, K.; Noordin, K.A.; Alsamhi, S.H.; Hawbani, A. Multi-UAV and SAR collaboration model for disaster management in B5G networks. Internet Technol. Lett. 2024, 7, e310. [Google Scholar] [CrossRef]
- Otte, M.; Kuhlman, M.J.; Sofge, D. Auctions for multi-robot task allocation in communication limited environments. Auton. Robot. 2020, 44, 547–584. [Google Scholar] [CrossRef]
- Robin, C.; Lacroix, S. Multi-robot target detection and tracking: Taxonomy and survey. Auton. Robot. 2016, 40, 729–760. [Google Scholar] [CrossRef]
- Kapoutsis, A.C.; Chatzichristofis, S.A.; Doitsidis, L.; De Sousa, J.B.; Pinto, J.; Braga, J.; Kosmatopoulos, E.B. Real-time adaptive multi-robot exploration with application to underwater map construction. Auton. Robot. 2016, 40, 987–1015. [Google Scholar] [CrossRef]
- Torreno, A.; Onaindia, E.; Komenda, A.; Štolba, M. Cooperative multi-agent planning: A survey. ACM Comput. Surv. (CSUR) 2017, 50, 1–32. [Google Scholar] [CrossRef]
- Cao, Y.; Yu, W.; Ren, W.; Chen, G. An overview of recent progress in the study of distributed multi-agent coordination. IEEE Trans. Ind. Inform. 2012, 9, 427–438. [Google Scholar] [CrossRef]
- Chopra, S.; Notarstefano, G.; Rice, M.; Egerstedt, M. A distributed version of the hungarian method for multirobot assignment. IEEE Trans. Robot. 2017, 33, 932–947. [Google Scholar] [CrossRef]
- Jang, I.; Shin, H.S.; Tsourdos, A. Anonymous hedonic game for task allocation in a large-scale multiple agent system. IEEE Trans. Robot. 2018, 34, 1534–1548. [Google Scholar] [CrossRef]
- Schillinger, P.; Bürger, M.; Dimarogonas, D.V. Simultaneous task allocation and planning for temporal logic goals in heterogeneous multi-robot systems. Int. J. Robot. Res. 2018, 37, 818–838. [Google Scholar] [CrossRef]
- Xie, S.; Zhang, A.; Bi, W.; Tang, Y. Multi-UAV mission allocation under constraint. Appl. Sci. 2019, 9, 2184. [Google Scholar] [CrossRef]
- Whitbrook, A.; Meng, Q.; Chung, P.W. Reliable, distributed scheduling and rescheduling for time-critical, multiagent systems. IEEE Trans. Autom. Sci. Eng. 2017, 15, 732–747. [Google Scholar] [CrossRef]
- Gerkey, B.P.; Matarić, M.J. A formal analysis and taxonomy of task allocation in multi-robot systems. Int. J. Robot. Res. 2004, 23, 939–954. [Google Scholar] [CrossRef]
- Nunes, E.; Manner, M.; Mitiche, H.; Gini, M. A taxonomy for task allocation problems with temporal and ordering constraints. Robot. Auton. Syst. 2017, 90, 55–70. [Google Scholar] [CrossRef]
- Chen, X.; Liu, Y.; Yin, L.; Qi, L. Cooperative task assignment and track planning for multi-UAV attack mobile targets. J. Intell. Robot. Syst. 2020, 100, 1383–1400. [Google Scholar]
- Li, M.; Liu, C.; Li, K.; Liao, X.; Li, K. Multi-task allocation with an optimized quantum particle swarm method. Appl. Soft Comput. 2020, 96, 106603. [Google Scholar] [CrossRef]
- Kim, J.; Oh, H.; Yu, B.; Kim, S. Optimal task assignment for UAV swarm operations in hostile environments. Int. J. Aeronaut. Space Sci. 2021, 22, 456–467. [Google Scholar] [CrossRef]
- Fan, C.; Han, S.; Li, X.; Zhang, T.; Yuan, Y. A modified nature-inspired meta-heuristic methodology for heterogeneous unmanned aerial vehicle system task assignment problem. Soft Comput. 2021, 25, 14227–14243. [Google Scholar] [CrossRef]
- Ye, F.; Chen, J.; Sun, Q.; Tian, Y.; Jiang, T. Decentralized task allocation for heterogeneous multi-UAV system with task coupling constraints. J. Supercomput. 2021, 77, 111–132. [Google Scholar] [CrossRef]
- Xu, L.; Qiao, J.; Lin, S.; Wang, X. Research on the task assignment problem with maximum benefits in volunteer computing platforms. Symmetry 2020, 12, 862. [Google Scholar] [CrossRef]
- Duan, X.; Liu, H.; Tang, H.; Cai, Q.; Zhang, F.; Han, X. A novel hybrid auction Algorithm for multi-UAVs dynamic task assignment. IEEE Access 2019, 8, 86207–86222. [Google Scholar] [CrossRef]
- Luo, L.; Chakraborty, N.; Sycara, K. Distributed algorithms for multirobot task assignment with task deadline constraints. IEEE Trans. Autom. Sci. Eng. 2015, 12, 876–888. [Google Scholar] [CrossRef]
- Schneider, E.; Sklar, E.I.; Parsons, S. Mechanism selection for multi-robot task allocation. In Proceedings of the Towards Autonomous Robotic Systems: 18th Annual Conference, TAROS 2017, Guildford, UK, 19–21 July 2017; Proceedings 18. Springer: Cham, Switzerland, 2017; pp. 421–435. [Google Scholar]
- Kim, K.S.; Kim, H.Y.; Choi, H.L. A bid-based grouping method for communication-efficient decentralized multi-UAV task allocation. Int. J. Aeronaut. Space Sci. 2020, 21, 290–302. [Google Scholar] [CrossRef]
- Choi, H.L.; Brunet, L.; How, J.P. Consensus-based decentralized auctions for robust task allocation. IEEE Trans. Robot. 2009, 25, 912–926. [Google Scholar] [CrossRef]
- Fu, X.; Feng, P.; Gao, X. Swarm UAVs task and resource dynamic assignment Algorithm based on task sequence mechanism. IEEE Access 2019, 7, 41090–41100. [Google Scholar] [CrossRef]
- Chen, J.; Qing, X.; Ye, F.; Xiao, K.; You, K.; Sun, Q. Consensus-based bundle Algorithm with local replanning for heterogeneous multi-UAV system in the time-sensitive and dynamic environment. J. Supercomput. 2022, 78, 1712–1740. [Google Scholar] [CrossRef]
- Johnson, L.; Choi, H.L.; Ponda, S.; How, J.P. Allowing non-submodular score functions in distributed task allocation. In Proceedings of the 2012 IEEE 51st IEEE Conference on Decision and Control (CDC), Maui, HI, USA, 10–13 December 2012; IEEE: New York, NY, USA, 2012; pp. 4702–4708. [Google Scholar]
- Buckman, N.; Choi, H.L.; How, J.P. Partial replanning for decentralized dynamic task allocation. In Proceedings of the AIAA Scitech 2019 Forum, San Diego, CA, USA, 7–11 January 2019; p. 0915. [Google Scholar]
- Wu, W.; Xu, J.; Sun, Y. Integrate Assignment of Multiple Heterogeneous Unmanned Aerial Vehicles Performing Dynamic Disaster Inspection and Validation Task with Dubins Path. IEEE Trans. Aerosp. Electron. Syst. 2023, 59, 4018–4032. [Google Scholar] [CrossRef]
- Zhao, W.; Meng, Q.; Chung, P.W. A heuristic distributed task allocation method for multivehicle multitask problems and its application to search and rescue scenario. IEEE Trans. Cybern. 2015, 46, 902–915. [Google Scholar] [CrossRef]
- Turner, J.; Meng, Q.; Schaefer, G.; Whitbrook, A.; Soltoggio, A. Distributed task rescheduling with time constraints for the optimization of total task allocations in a multirobot system. IEEE Trans. Cybern. 2017, 48, 2583–2597. [Google Scholar] [CrossRef] [PubMed]
- Cui, W.; Li, R.; Feng, Y.; Yang, Y. Distributed task allocation for a multi-UAV system with time window constraints. Drones 2022, 6, 226. [Google Scholar] [CrossRef]
- Qamar, R.A.; Sarfraz, M.; Rahman, A.; Ghauri, S.A. Multi-criterion multi-UAV task allocation under dynamic conditions. J. King Saud Univ.-Comput. Inf. Sci. 2023, 35, 101734. [Google Scholar] [CrossRef]
- Wang, M.; Cao, S. Task assignment based on improved contract net in dynamic environment. Glob. J. Eng. Technol. Adv. 2020, 2, 045–053. [Google Scholar] [CrossRef]
- Wang, G.; Lv, X.; Yan, X. A Two-Stage Distributed Task Assignment Algorithm Based on Contract Net Protocol for Multi-UAV Cooperative Reconnaissance Task Reassignment in Dynamic Environments. Sensors 2023, 23, 7980. [Google Scholar] [CrossRef]
- Gao, X.; Wang, L.; Su, X.; Lu, C.; Ding, Y.; Wang, C.; Peng, H.; Wang, X. A Unified Multi-Objective Optimization Framework for UAV Cooperative Task Assignment and Re-Assignment. Mathematics 2022, 10, 4241. [Google Scholar] [CrossRef]
- Wu, Y.; Gou, J.; Ji, H.; Deng, J. Hierarchical mission replanning for multiple UAV formations performing tasks in dynamic situation. Comput. Commun. 2023, 200, 132–148. [Google Scholar] [CrossRef]
- Holland, J.H. Genetic algorithms. Sci. Am. 1992, 267, 66–73. [Google Scholar] [CrossRef]
- Kennedy, J.; Eberhart, R. Particle swarm optimization. In Proceedings of the ICNN’95-International Conference on Neural Networks, Perth, Australia, 27 November–1 December 1995; IEEE: New York, NY, USA, 1995; Volume 4, pp. 1942–1948. [Google Scholar]
- Zhang, J.; Chen, Y.; Yang, Q.; Lu, Y.; Shi, G.; Wang, S.; Hu, J. Dynamic task allocation of multiple UAVs based on improved A-QCDPSO. Electronics 2022, 11, 1028. [Google Scholar] [CrossRef]
- Liu, W.; Wang, Z.; Zeng, N.; Yuan, Y.; Alsaadi, F.E.; Liu, X. A novel randomised particle swarm optimizer. Int. J. Mach. Learn. Cybern. 2021, 12, 529–540. [Google Scholar] [CrossRef]
- Lv, X.; Wang, G.; Chen, J. Multi-UAV Cooperative Reconnaissance Task Allocation Based on IEPPSO Algorithm. In Proceedings of the CCF Conference on Computer Supported Cooperative Work and Social Computing, Harbin, China, 18–20 August 2023; Springer: Singapore, 2023; pp. 119–129. [Google Scholar]
- Xu, S.; Li, L.; Zhou, Z.; Mao, Y.; Huang, J. A task allocation strategy of the UAV swarm based on multi-discrete wolf pack algorithm. Appl. Sci. 2022, 12, 1331. [Google Scholar] [CrossRef]
- Peng, Q.; Wu, H.; Li, N. Modeling and solving the dynamic task allocation problem of heterogeneous UAV swarm in unknown environment. Complexity 2022, 2022, 9219805. [Google Scholar] [CrossRef]
- Shi, J.; Tan, L.; Lian, X.; Xu, T.; Zhang, H.; Zhang, Y. A multi-unmanned aerial vehicle dynamic task assignment method based on bionic algorithms. Comput. Electr. Eng. 2022, 99, 107820. [Google Scholar] [CrossRef]
- Liu, Z.; Qiu, C.; Zhang, Z. Sequence-to-sequence multi-agent reinforcement learning for multi-UAV task planning in 3D dynamic environment. Appl. Sci. 2022, 12, 12181. [Google Scholar] [CrossRef]
- Liu, B.; Wang, S.; Li, Q.; Zhao, X.; Pan, Y.; Wang, C. Task assignment of UAV swarms based on deep reinforcement learning. Drones 2023, 7, 297. [Google Scholar] [CrossRef]
- Liu, D.; Dou, L.; Zhang, R.; Zhang, X.; Zong, Q. Multi-agent reinforcement learning-based coordinated dynamic task allocation for heterogenous UAVs. IEEE Trans. Veh. Technol. 2022, 72, 4372–4383. [Google Scholar] [CrossRef]
- Seenu, N.; RM, K.C.; Ramya, M.; Janardhanan, M.N. Review on state-of-the-art dynamic task allocation strategies for multiple-robot systems. Ind. Robot. Int. J. Robot. Res. Appl. 2020, 47, 929–942. [Google Scholar]
- Peng, Q.; Wu, H.; Xue, R. Review of dynamic task allocation methods for UAV swarms oriented to ground targets. Complex Syst. Model. Simul. 2021, 1, 163–175. [Google Scholar] [CrossRef]
- Biswas, S.; Anavatti, S.G.; Garratt, M.A. Path planning and task assignment for multiple UAVs in dynamic environments. In Unmanned Aerial Systems; Elsevier: Amsterdam, The Netherlands, 2021; pp. 81–102. [Google Scholar]
- Chen, X.; Zhang, P.; Du, G.; Li, F. A distributed method for dynamic multi-robot task allocation problems with critical time constraints. Robot. Auton. Syst. 2019, 118, 31–46. [Google Scholar] [CrossRef]
- Fei, B.; Liu, D.; Zhou, J.; Bao, W.; Chen, F.; Zhang, H. A spectral clustering enabled dynamic task allocation approach of multiple uavs. In Proceedings of the 2022 8th International Conference on Big Data and Information Analytics (BigDIA), Guiyang, China, 24–25 August 2022; IEEE: New York, NY, USA, 2022; pp. 408–415. [Google Scholar]
Symbol | Definition |
---|---|
T | Set of tasks |
Set of unknown new tasks | |
U | Set of UAVs |
The assigned tasks list of UAV , where | |
M | Number of tasks |
N | Number of UAVs |
Q | Number of new tasks |
Number of Clusters | |
i | The UAVs index with the range |
j | The tasks index with the range |
q | The new tasks index with the range |
The score value of UAV performing task | |
The start time of task | |
The end time of task | |
The execution time of task by | |
The maximum number of tasks that can be performed by | |
The task reward of | |
The distance from the base station to task | |
The distance from task to task | |
The velocity of | |
The duration of task | |
Time discount factor | |
The waiting time of task until executed by | |
The gained throughput of UAV after performing task | |
Number of participated UAVs in reassignment process | |
Number of released tasks during reassignment process | |
Width of the simulation map | |
Task distribution, which can be random, concentrated, or predetermined |
Instances Size | Map Size | #Tasks (M) | #UAVs (N) |
---|---|---|---|
Small | 1000-1250-1500-1750-2000 | 100-150-200-250-300 | 5-7-10-12-15 |
Large | 6000-7000-8000-9000-10,000 | 600-700-800-900-1000 | 35-40-45-50-55 |
Parameter | Value (Small and Large Instances) |
---|---|
#Clusters | Small: 2, 3, 4, 5, 6 | Large: 12, 13, 14, 15, 16 |
UAVs Max Load | |
UAVs Velocity | 20% MapSize |
#New Tasks | 5% M |
Sensor Detection Range | 7% MapSize |
Task Time Window |
Algorithms | Score |
---|---|
No reset [31] | 380.7479 |
Full reset [34] | 382.0065 |
Single reset [35] | 381.3196 |
PR with fixed reset [35] | 382.0065 |
PR with team reset [35] | 382.0065 |
CBBA-LR [33] | 382.0065 |
HCPR | 422.9202 |
Algorithms | Time Window (TW) | |||||
---|---|---|---|---|---|---|
[5, 25] | [5, 50] | [10, 25] | [10, 50] | [15, 25] | [15, 50] | |
No reset [31] | 0 | 5.7042 | 0 | 5.7042 | 0 | 5.7042 |
Full reset [34] | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 |
Single reset [35] | 0 | 5.7042 | 0 | 5.7042 | 0 | 5.7042 |
Heuristic full reset [35] | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 |
PR with fixed reset [35] | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 |
PR with team reset [35] | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 |
CBBA-LR [33] | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 | 6.9627 |
HCPR | 10.5163 | 10.5163 | 20.2231 | 20.2231 | 32.5820 | 32.5820 |
Algorithms | One New Task | Two New Tasks | Three New Tasks | Four New Tasks | Five New Tasks |
---|---|---|---|---|---|
No reset [31] | 5.1589 | 6.8019 | 10.2410 | 12.5276 | 14.384 |
Full reset [34] | 8.5182 | 10.2016 | 16.6892 | 18.846 | 22.7084 |
Single reset [35] | 5.2262 | 6.8019 | 10.2410 | 12.5276 | 14.384 |
Heuristic full reset [35] | 8.5182 | 10.2016 | 15.2255 | 17.5712 | 19.9235 |
PR with fixed reset [35] | 8.5182 | 10.2016 | 15.2255 | 17.820 | 20.1234 |
PR with team reset [35] | 8.5182 | 10.2016 | 15.2255 | 17.820 | 20.1234 |
CBBA-LR [33] | 8.5182 | 10.2016 | 16.6892 | 18.846 | 22.7084 |
HCPR | 21.575518 | 41.780054 | 72.875293 | 87.511316 | 103.221674 |
Scenario | No [31] | Full [34] | Single [35] | Heuristic [35] | Fixed [35] | Team [35] | CBBA-LR [33] | HCPR |
---|---|---|---|---|---|---|---|---|
NU = 5, NT = 10 | 6.8650 | 9.1492 | 6.8650 | 8.6004 | 8.3589 | 8.3589 | 9.1492 | 12.456548 |
NU = 5, NT = 20 | 3.8654 | 5.1698 | 3.8654 | 5.1698 | 4.3984 | 4.3984 | 5.1698 | 10.703387 |
NU = 5, NT = 30 | 2.0211 | 4.8586 | 2.0211 | 4.7583 | 2.8516 | 2.6798 | 4.8586 | 11.454415 |
NU = 10, NT = 20 | 9.3835 | 10.9559 | 9.3835 | 10.4736 | 10.7176 | 10.3445 | 10.9559 | 18.517290 |
NU = 10, NT = 30 | 6.4906 | 8.9158 | 6.4906 | 8.3174 | 8.1911 | 8.2538 | 8.9158 | 11.712061 |
NU = 10, NT = 40 | 4.6215 | 7.7695 | 4.6215 | 7.5584 | 5.9541 | 5.0388 | 7.7695 | 10.413622 |
NU = 15, NT = 30 | 9.0957 | 11.3661 | 9.0957 | 10.2369 | 11.0944 | 10.8700 | 11.3661 | 17.486393 |
NU = 15, NT = 40 | 7.5692 | 10.0984 | 7.5692 | 9.2497 | 9.1497 | 8.6179 | 10.0984 | 17.123623 |
NU = 15, NT = 50 | 6.2113 | 7.4322 | 6.2113 | 7.3528 | 7.1030 | 6.3295 | 7.4322 | 11.635331 |
NU = 15, NT = 60 | 4.9995 | 6.8825 | 4.9995 | 6.8344 | 5.6678 | 4.8690 | 6.8825 | 9.987903 |
NU = 20, NT = 40 | 9.3634 | 12.3817 | 9.3634 | 10.5540 | 12.1520 | 10.7517 | 12.3817 | 23.528094 |
NU = 20, NT = 50 | 8.3122 | 10.4728 | 8.3122 | 9.6811 | 10.1067 | 8.7328 | 10.4728 | 19.281718 |
NU = 20, NT = 60 | 7.2814 | 11.1660 | 7.2814 | 9.7506 | 10.0227 | 7.2916 | 11.1660 | 18.284739 |
NU = 20, NT = 80 | 5.4008 | 8.0004 | 5.4008 | 7.7693 | 6.5903 | 5.4262 | 8.0004 | 16.604281 |
NU = 20, NT = 100 | 4.6560 | 6.1548 | 4.6560 | 6.0738 | 4.9219 | 4.6684 | 6.1548 | 11.543217 |
NU = 25, NT = 50 | 9.9358 | 12.9114 | 9.9358 | 12.8417 | 12.8963 | 12.6168 | 12.9114 | 28.909494 |
NU = 25, NT = 60 | 9.1387 | 11.7833 | 9.1387 | 10.2684 | 11.6273 | 11.6594 | 11.7833 | 23.188391 |
NU = 25, NT = 80 | 8.3721 | 9.7330 | 8.3721 | 8.3692 | 9.2030 | 8.6084 | 9.7330 | 18.165586 |
NU = 25, NT = 100 | 6.2311 | 8.5242 | 6.2311 | 8.1836 | 7.4738 | 6.9631 | 8.5242 | 15.904294 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Alqefari, S.; Menai, M.E.B. A Hybrid Method to Solve the Multi-UAV Dynamic Task Assignment Problem. Sensors 2025, 25, 2502. https://doi.org/10.3390/s25082502
Alqefari S, Menai MEB. A Hybrid Method to Solve the Multi-UAV Dynamic Task Assignment Problem. Sensors. 2025; 25(8):2502. https://doi.org/10.3390/s25082502
Chicago/Turabian StyleAlqefari, Shahad, and Mohamed El Bachir Menai. 2025. "A Hybrid Method to Solve the Multi-UAV Dynamic Task Assignment Problem" Sensors 25, no. 8: 2502. https://doi.org/10.3390/s25082502
APA StyleAlqefari, S., & Menai, M. E. B. (2025). A Hybrid Method to Solve the Multi-UAV Dynamic Task Assignment Problem. Sensors, 25(8), 2502. https://doi.org/10.3390/s25082502