Unmanned aerial vehicle cluster system based on real-time communication and collaborative decision and distributed task planning method

Abstract

The invention discloses an unmanned aerial vehicle cluster system and a distributed task planning method based on real-time communication and collaborative decision, and the method carries out the grouping and network simplification of unmanned aerial vehicle cluster distributed task planning based on conflict resolution. Comprising task conflict resolution based on a pre-estimation strategy, grouping based on task selection tendency and distributed network simplification based on bidding. According to the invention, each unmanned aerial vehicle selectively receives messages according to a necessary network constructed by the unmanned aerial vehicle, and redundant communication in internal circulation is further reduced; by comparing compiler output results, inconsistency is detected to discover defect examples. By adopting the multi-unmanned-aerial-vehicle cooperative task allocation method designed by the invention, the task allocation efficiency of a large-scale multi-unmanned-aerial-vehicle system can be remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of collaborative control and task allocation technology for multi-UAV systems, specifically relating to a UAV swarm system and a distributed task planning method based on real-time communication and collaborative decision-making. Background Technology

[0002] In complex scenarios such as wilderness search and rescue, disaster monitoring, and logistics delivery, multi-UAV systems are widely used due to their efficiency, flexibility, and robustness. Task allocation is one of the core issues in the collaborative control of multi-UAV systems. Currently, market-based distributed task allocation algorithms, such as Consensus-Based Bundle Algorithm (CBBA) and its derivatives, have become a research hotspot due to their good scalability and distributed characteristics.

[0003] However, existing distributed task allocation algorithms still face the following challenges when dealing with large-scale drone swarms and complex task scenarios: 1. High communication overhead: With the increase in the number of drones, the amount of communication interaction between nodes grows exponentially. Especially in the consensus phase, a large amount of redundant communication seriously affects the real-time performance and energy efficiency of the system.

[0004] 2. Slow convergence speed: Frequent task conflicts require the algorithm to perform more external loop iterations to reach a final consensus, reducing the efficiency of task allocation.

[0005] 3. Insufficient task redistribution capability: Traditional algorithms such as PI-MinAvg have limited ability to redistribute unassigned tasks after task allocation, which may lead to a low task completion rate.

[0006] To address these issues, researchers have proposed improved algorithms such as PI-MaxAss and PI-Apex, aiming to enhance task allocation efficiency and completion rates. PI-Apex, by introducing exchange strategies and local sequence reallocation, improves the flexibility of task allocation to some extent. However, these improvements do not fundamentally solve the communication efficiency problem in large-scale clusters, and key technologies such as prediction strategies and network simplification still need to be deeply integrated and optimized in distributed task allocation algorithms. Summary of the Invention

[0007] Purpose of the invention: This invention aims to provide a drone swarm system and a distributed task planning method based on real-time communication and collaborative decision-making, which significantly reduces the amount of communication interaction in the task allocation process of multiple drone systems, improves the convergence speed of the algorithm, and maintains or improves the quantity and efficiency of task allocation.

[0008] Technical Solution: A distributed task planning method for UAV swarms based on real-time communication and collaborative decision-making. This method, for distributed task planning of UAV swarms, is based on conflict resolution through grouping and network simplification, and includes the following steps: (1) Task conflict resolution based on prediction strategy: Before the task inclusion phase begins, each UAV identifies a set of UAVs that have task conflicts with itself, and uses an estimation function to predict the task results that the UAVs in the conflicting UAV set may add; and resolves the conflicting tasks. Performance impact value Conflict drones Performance impact on this task Perform a comparison, if < Then add the task in the current round. Otherwise, abandon the task; (2) Grouping based on task selection preference: After the UAV generates the initial task list, it is clustered and grouped according to the minimum performance impact value of the UAV on the task. In the task coordination process, the intra-group loop is executed first to achieve consensus on task allocation within the group, and then the inter-group loop is executed to resolve inter-group task allocation conflicts. (3) Bid-based distributed network simplification: The bid-based distributed fragmentation primitive algorithm is adopted to eliminate the closed triangle structure in the communication network to simplify the network topology while ensuring network connectivity and information propagation efficiency; each UAV selectively receives messages according to the necessary network it has built, reducing redundant communication in the internal loop.

[0009] Furthermore, the performance impact value refers to the contribution of the UAV to the global objective function of the entire system after performing a certain task. It is calculated using the minimum average performance impact algorithm for candidate tasks. The performance impact value mentioned above is expressed as follows: ,in Indicates drone The current task sequence, Indicates the task Insertion sequence position , This represents the total time cost of executing the task sequence; the drone selection... The task with the smallest value and its insertion position are included in its own sequence; The performance impact of removing a task is expressed as follows: ,in This indicates removing a task from the sequence. When multiple drones are performing the same task When a conflict occurs, retain The drone with the highest value acquires ownership of the mission, while other drones remove the mission.

[0010] In the task conflict resolution step, the method by which the UAV infers the location of the conflicting UAV is as follows: (11) Based on the inverse function of the performance impact value calculation formula, combined with the bid value of the conflict UAV for at least three different tasks, the task location, the task execution time and the flight speed of the UAV, calculate the distance of the conflict UAV to these tasks; (12) Using the location of each mission as the center and the corresponding distance as the radius, determine the spatial coordinates of the conflicting UAVs through the intersection of the three circles; (13) Based on the inferred location information, the current mission sequence and performance impact value of the conflicting UAV, simulate its mission decision-making process to complete the mission result prediction.

[0011] Furthermore, the steps of the above method for performing task conflict pre-resolution based on the prediction strategy include: Each drone identifies the set of drones that have mission conflicts with it based on the bidding information of neighboring drones received in the previous round of communication consensus phase. For each drone in the conflict drone set, estimate the new tasks that may be added in the next iteration and their corresponding performance impact values; When making mission inclusion decisions locally, if a mission is found to have a lower estimated performance impact on a conflict drone, the drone will abandon adding that mission to avoid potential conflict.

[0012] In the above method, the grouping step based on task selection tendency is defined by the following rule for grouping UAVs: if two UAVs have the same minimum performance impact value task in their task sequences, they are determined to have similar task selection tendency and are grouped into the same group. Task list to be processed for each group of drones Defined as:

[0013] In the formula, Indicates a group, This represents the union of all drone mission sequences within the group. This represents the set of all tasks that have not been assigned to any drone.

[0014] Furthermore, in the distributed network simplification step, the execution flow of the BDBM algorithm is as follows: (31) Initialization: Each UAV will establish its own necessary communication network. Initialize to a zero matrix; (32) Constructing initial communication edges: Each UAV maintains a direct communication edge with its neighboring UAVs that are competing for the same task, thus establishing a communication network. The corresponding position in the middle is set to 1; (33) Extended network information: Each UAV receives the communication side information that its neighboring UAVs need to maintain and merges it into its own communication network. ; (34) Eliminate redundant communication edges: Check the communication links with neighbors. If there are intermediate drones... Can transfer to another neighbor The message will be removed. The direct communication edge, only through Relay messages to eliminate the closed triangle communication structure.

[0015] Furthermore, this method simplifies the network through functions. Determine the triangular communication structure; the function is used to detect drones. and Are there any intermediate drones nearby that can relay messages between the two? .

[0016] The optimization objective of this method in multi-UAV cooperative task allocation is to minimize the waiting time of rescued targets while maximizing the number of rescued targets. The mathematical expression for the optimization objective is:

[0017] In the formula, It is a task In drones task sequence Time cost in the process; Task assignment must meet the following constraints: Task capacity constraints: , For drones Maximum number of executable tasks; Task uniqueness constraint: , It is an indicator function; Task set constraints: ; Time deadline constraints: , , For the task The corresponding latest deadline; Functional adaptability constraints: A value of 1 indicates a drone. Functionally capable of undertaking tasks ; Task sequence functional constraints: .

[0018] Based on the implementation of the above scheme, this invention provides a drone swarm system for executing the aforementioned distributed task planning method for drone swarms based on real-time communication and collaborative decision-making. The system comprises multiple heterogeneous drones, each drone being equipped with: The processing unit is used for local execution of logical operations related to task inclusion, conflict detection and resolution, task result prediction, UAV grouping and network simplification. The communication unit is used to conduct local communication with neighboring drones within the communication network and exchange mission bidding-related information. The navigation and execution unit is used to fly to the task location in sequence and execute the corresponding task after the task sequence is determined.

[0019] On the other hand, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor in a drone swarm, enables the drone swarm to implement the multi-drone collaborative task allocation method described above.

[0020] Beneficial effects: Compared with the prior art, the substantive features and significant advantages of this invention are: Communication efficiency is greatly improved: By using grouping strategies and BDBM network simplification methods, the communication range of UAVs is reduced from the entire cluster to intra-group and necessary connections, achieving a reduction in total communication volume of 65.7% to 80%.

[0021] The convergence speed is significantly accelerated: the task conflict resolution based on the prediction strategy reduces the number of external loop iterations by about 73% compared with existing algorithms, thus accelerating the convergence speed of task allocation.

[0022] Task allocation performance is maintained: While significantly reducing communication volume, this method maintains the same number of task allocations as the PI-Apex algorithm, and the increase in task waiting time is controllable (not exceeding 8%), achieving a balance between efficiency and performance.

[0023] Distributed and robust: The BDBM algorithm is a distributed network simplification method that is suitable for dynamically changing network topologies and enhances the robustness of the system in complex environments. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the method described in this invention; Figure 2 This is a schematic diagram illustrating the process of inferring the location of the conflicting drones; Figure 3 This is a schematic diagram of the task conflict resolution process based on the prediction strategy; Figure 4 This is a simplified diagram of the network process of the Distributed Breaking Primitives (BDBM) algorithm based on bidding. Detailed Implementation

[0025] In the above technical solution of the present invention, the method, based on the distributed task allocation algorithm, introduces three core components: task conflict resolution based on prediction strategy, grouping strategy based on task selection preference, and distributed network simplification method (BDBM) based on bidding. Specifically, it includes the following steps: Before the task inclusion phase begins, the UAV analyzes the set of conflicting UAVs, uses an estimation function h to predict the possible task outcomes that the conflicting UAVs might add, and compares this prediction with its own current iteration results. If it has conflicting task outcomes... Performance impact value Smaller than conflict drones Performance impact value Then add the task in the current round. Otherwise, abandon the addition to avoid task conflicts, thereby reducing the number of outer loop iterations.

[0026] Task selection preference-based grouping: After the UAVs generate the initial task list, they are clustered and grouped according to the UAVs' minimum performance impact on the tasks (i.e., task selection preference). During the task coordination phase, intra-group loops are first used to achieve task consensus within the group, and then inter-group loops are used to resolve inter-group allocation conflicts. This limits the communication scope to a necessary subset of UAVs, significantly reducing the amount of internal loop communication interaction.

[0027] Distributed network simplification employs a bid-based distributed fragmentation primitive (BDBM) algorithm. This algorithm simplifies the network topology by eliminating closed triangle structures in the communication network, while ensuring network connectivity and information propagation efficiency. Each UAV constructs its own necessary network. Selectively receive messages to further reduce redundant communication in the internal loop.

[0028] Each drone, based on its own necessary communication network Selective message reception further reduces redundant communication in the internal loop; inconsistencies are detected by comparing compiler output to identify defective samples. The multi-UAV cooperative task allocation method designed using this invention can significantly improve the task allocation efficiency of large-scale multi-UAV systems.

[0029] Combination Figure 1 Specifically, the implementation steps of this method include: S1. Task Inclusion: Based on local information, each UAV aims to minimize time costs by calculating the performance impact value of the included tasks and using the minimum average performance impact algorithm to select tasks that meet the constraints, greedily including them into its own task sequence. S2. Communication consensus: Each UAV exchanges task bidding information with neighboring UAVs in the communication network. The bidding information includes at least a list of winning tasks and the corresponding performance impact values. S3. Conflict Detection and Resolution: Each UAV detects whether there are conflicts in its own task sequence based on the bidding information received from neighboring UAVs; when a conflict is detected, the conflicting task is removed based on preset rules, and the local task sequence and performance impact value are updated. S4. Iterative Execution: Repeat the task inclusion phase, communication consensus phase, and conflict detection and resolution phase until the task sequence of all UAVs no longer changes within the predetermined number of iterations, thereby completing the distributed task planning.

[0030] The above methods can be widely applied to the following scenarios: Wilderness Search and Rescue: In post-disaster or remote area search and rescue missions, multiple heterogeneous drones (such as those carrying medicine, food, thermal imagers, etc.) need to be quickly and efficiently distributed to survivors with different needs. This method can significantly reduce communication latency between drones and accelerate rescue response speed.

[0031] Environmental monitoring: In large-scale environmental monitoring missions, multiple drones need to collaborate to collect data. This method, through grouping and network simplification, ensures efficient communication and task allocation even with a large number of drones, achieving rapid coverage of the monitoring area.

[0032] Logistics and Delivery: In multi-drone logistics and delivery systems within cities or industrial parks, it is necessary to efficiently distribute packages to different delivery points. This method can optimize task sequences, reduce drone waiting time, and improve overall delivery efficiency.

[0033] This embodiment sets up a scenario: addressing the problem of multi-drone collaborative task allocation in a field rescue scenario, defining... Let T be a set of n heterogeneous unmanned aerial vehicles (UAVs). The set of tasks (survivors) is represented as drones It can be represented as a tuple ,in For the initial three-dimensional coordinates, For travel speed. Task The features are represented as ,in In three-dimensional coordinates, It is a rescue mission Required time. The mission sequence for each UAV is represented as follows: , For drones The task sequence.

[0034] The optimization objective of this scenario is to minimize the waiting time C for rescued survivors while maximizing the number N of rescued survivors, expressed as:

[0035] in It is a task In drones task sequence The time cost in the process; that is, drones Arrival of the survivors Location and time point:

[0036] In the formula, dis(.) represents the Euclidean distance.

[0037] Based on the above schemes, the design of the RCFPI-Apex algorithm is introduced below. The RCFPI-Apex algorithm significantly improves the task allocation efficiency of large-scale multi-UAV systems by improving the communication and conflict resolution mechanisms of the PI-Apex algorithm. That is, the core three parts of the UAV swarm distributed task planning method with real-time communication and collaborative decision-making described in this invention are: (1) Combination Figure 2 As shown, the present invention resolves task conflicts based on a prediction strategy. This step avoids task conflicts by predicting the scheduling scheme of conflicting UAVs, thereby reducing the number of external loop iterations.

[0038] Conflict Identification: Drones From the received winning vector Identify its own task list Drones with mission conflicts This forms a conflicting set of drones, W. A conflict occurs when two or more drones are simultaneously assigned the task. Conflicts are identified through the task assignment list and determined based on the winning vector, which indicates that the drone has won the corresponding task during the iteration.

[0039] Forecasting and Avoidance: During the task addition process, drones Computational Conflict Drones Possible task results to be added. If the task... Both appeared in the mission sequences of both parties, and the drone The performance impact on this task is less than The performance impact value, then Add to Otherwise, drones Give up adding To avoid conflict.

[0040] Conflict drone location inference: to achieve prediction Need to infer conflict drones Location By utilizing the received historical bidding data, the inverse function of the performance impact function is used to derive the drone's... With the task Distance between .

[0041] This invention achieves this through at least three different tasks , , The bidding information can be used to calculate the conflict drones. The spatial coordinates, that is, three circles centered on the task location and corresponding distances. , , The intersection of circles with radius .

[0042] (2) Combination Figure 3 As shown, this invention uses a task selection preference-based grouping strategy to reduce communication interaction by clustering drones with similar task selection preferences into the same group.

[0043] Group construction: Generate an initial task list for each drone. Subsequently, if two drones have the same minimum performance impact value in their mission sequences, they are considered to have similar mission selection tendencies and are assigned to the same group. .

[0044] Task list: for each group Task list to be processed Defined as:

[0045] in Includes all unassigned tasks.

[0046] Conflict Coordination: Conflict coordination is divided into intra-group and inter-group rounds. Intra-group rounds ensure that there are no conflicts in task allocation within a group. Inter-group rounds resolve inter-group allocation conflicts through message exchange, and the process is consistent with the conflict coordination phase in the PI-Apex algorithm. If conflicting tasks exist, the group with the higher performance impact value for that task retains the task, while drones in the other group must remove the task.

[0047] (3) This invention uses the BDBM algorithm to simplify the distributed communication network, thereby minimizing the amount of internal loop communication interaction, such as Figure 4 As shown.

[0048] Necessary network construction: for each drone Build a necessary network ,in Indicates in In the network, from drones The message can be transmitted to .

[0049] BDBM algorithm steps: Initial task list acquisition: Each UAV uses the PI-MinAvg algorithm to calculate the initial task list.

[0050] Conflict edge maintenance: Each drone maintains communication edges with its neighboring drones competing for the same task, thus building the initial network. .

[0051] Neighbor information extension: Each drone receives information from its connected neighboring drones regarding the communication edges that need to be maintained, and extends its own information accordingly. .

[0052] Connectivity checks and extensions: Drones Inspect neighboring drones The communication link ensures connectivity can be maintained through direct connection or bridging. This process involves determining the presence of an intermediate drone. Forming a triangular network topology To achieve this.

[0053] After running the BDBM algorithm, each UAV, based on its constructed network... Selective message reception eliminates unnecessary communication edges and reduces network load while maintaining conflict resolution capabilities.

Claims

1. A method for real-time communication and collaborative decision based UAV swarm distributed mission planning, characterized in that, The method for distributed task planning in UAV swarms is based on conflict resolution for grouping and network simplification, and includes the following steps: (1) Task conflict resolution based on prediction strategy: Before the task inclusion phase begins, each UAV identifies a set of UAVs that have task conflicts with itself, and uses an estimation function to predict the task results that the UAVs in the conflicting UAV set may add; and resolves the conflicting tasks. Performance impact value Conflict drones Performance impact on this task Perform a comparison, if < Then add the task in the current round. Otherwise, abandon the task; (2) Grouping based on task selection preference: After the UAV generates the initial task list, it is clustered and grouped according to the minimum performance impact value of the UAV on the task. In the task coordination process, the intra-group loop is executed first to achieve consensus on task allocation within the group, and then the inter-group loop is executed to resolve inter-group task allocation conflicts. (3) Bid-based distributed network simplification: The bid-based distributed fragmentation primitive algorithm is adopted to eliminate the closed triangle structure in the communication network to simplify the network topology while ensuring network connectivity and information propagation efficiency; each UAV selectively receives messages according to the necessary network it has built, reducing redundant communication in the internal loop.

2. The distributed task planning method for UAV swarms based on real-time communication and collaborative decision-making according to claim 1, characterized in that, The performance impact value refers to the contribution of a UAV to the global objective function of the entire system after performing a certain task. It is calculated using the minimum average performance impact algorithm for candidate tasks. The performance impact value mentioned above is expressed as follows: ,in Indicates drone The current task sequence, Indicates the task Insertion sequence position , This represents the total time cost of executing the task sequence; the drone selection... The task with the smallest value and its insertion position are included in its own sequence; The performance impact of removing a task is expressed as follows: ,in This indicates removing a task from the sequence. When multiple drones are performing the same task When a conflict occurs, retain The drone with the highest value acquires ownership of the mission, while other drones remove the mission.

3. The distributed task planning method for UAV swarms based on real-time communication and collaborative decision-making as described in claim 1, characterized in that, In the task conflict resolution step, the method by which the UAV infers the location of the conflicting UAV is as follows: (11) Based on the inverse function of the performance impact value calculation formula, combined with the bid value of the conflict UAV for at least three different tasks, the task location, the task execution time and the flight speed of the UAV, calculate the distance of the conflict UAV to these tasks; (12) Using the location of each mission as the center and the corresponding distance as the radius, determine the spatial coordinates of the conflicting UAVs through the intersection of the three circles; (13) Based on the inferred location information, the current mission sequence and performance impact value of the conflicting UAV, simulate its mission decision-making process to complete the mission result prediction.

4. The distributed task planning method for UAV swarms based on real-time communication and collaborative decision-making according to claim 1, characterized in that, The steps for performing task conflict pre-resolution based on the prediction strategy include: Each drone identifies the set of drones that have mission conflicts with it based on the bidding information of neighboring drones received in the previous round of communication consensus phase. For each drone in the conflict drone set, estimate the new tasks that may be added in the next iteration and their corresponding performance impact values; When making mission inclusion decisions locally, if a mission is found to have a lower estimated performance impact on a conflict drone, the drone will abandon adding that mission to avoid potential conflict.

5. The distributed task planning method for UAV swarms based on real-time communication and collaborative decision-making according to claim 1, characterized in that, In the task selection tendency-based grouping step, the grouping rule for UAVs is: if two UAVs have the same minimum performance impact value task in their task sequences, then they are determined to have similar task selection tendencies and are classified into the same group. Task list to be processed for each group of drones Defined as: In the formula, Indicates a group, This represents the union of all drone mission sequences within the group. This represents the set of all tasks that have not been assigned to any drone.

6. The distributed task planning method for UAV swarms based on real-time communication and collaborative decision-making according to claim 1, characterized in that, In the simplified steps of the distributed network, the execution flow of the BDBM algorithm is as follows: (31) Initialization: Each UAV will establish its own necessary communication network. Initialize to a zero matrix; (32) Constructing initial communication edges: Each UAV maintains a direct communication edge with its neighboring UAVs that are competing for the same task, thus establishing a communication network. The corresponding position in the middle is set to 1; (33) Extended network information: Each UAV receives the communication side information that its neighboring UAVs need to maintain and merges it into its own communication network. ; (34) Eliminate redundant communication edges: Check the communication links with neighbors. If there are intermediate drones... Can transfer to another neighbor The message will be removed. The direct communication edge, only through Relay messages to eliminate the closed triangle communication structure.

7. The distributed task planning method for UAV swarms based on real-time communication and collaborative decision-making according to claim 6, characterized in that, Through function Determine the triangular communication structure; the function is used to detect drones. and Are there any intermediate drones nearby that can relay messages between the two? .

8. The distributed mission planning method for UAV swarms based on real-time communication and collaborative decision-making according to any one of claims 1-7, characterized in that, The optimization objective of this method in multi-UAV cooperative task allocation is to minimize the waiting time of rescued targets while maximizing the number of rescued targets. The mathematical expression for the optimization objective is: In the formula, It is a task In drones task sequence Time cost in the process; Task assignment must meet the following constraints: Task capacity constraints: , For drones Maximum number of executable tasks; Task uniqueness constraint: , It is an indicator function; Task set constraints: ; Time deadline constraints: , , For the task The corresponding latest deadline; Functional adaptability constraints: A value of 1 indicates a drone. Functionally capable of undertaking tasks ; Task sequence functional constraints: .

9. A drone swarm system, characterized in that, For executing the distributed task planning method for UAV swarms based on real-time communication and collaborative decision-making as described in any one of claims 1-8, the system comprises multiple heterogeneous UAVs, each UAV being equipped with: The processing unit is used for local execution of logical operations related to task inclusion, conflict detection and resolution, task result prediction, UAV grouping and network simplification. The communication unit is used to conduct local communication with neighboring drones within the communication network and exchange mission bidding-related information. The navigation and execution unit is used to fly to the task location in sequence and execute the corresponding task after the task sequence is determined.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor in the drone cluster, enables the drone cluster to implement the multi-drone collaborative task allocation method as described in any one of claims 1-8.

Images