Distributed mission planning method and device for drone swarm based on local consensus

Abstract

The present application relates to a method and device for distributed task planning of drone clusters based on local consensus. The method proposes the concept of local consensus, that is, the drones and task scope of the consensus are local; unlike the traditional global consensus, the local consensus does not require the task planning information of all drones to be consistent, so the task information of other drones in the cluster can be further screened and integrated, and at the same time, it is only necessary to propagate the processed task information to the local area related to the task; based on the cluster local consensus mechanism, the drone inter-machine communication protocol, task conflict resolution rules and cluster information propagation mechanism are designed to reduce the amount and frequency of communication between drones, which greatly improves the scalability and real-time performance of the algorithm. This method reduces the amount and frequency of communication between drones required by the algorithm, adapts to the large-scale drone cluster self-organization rapid task planning under actual networking conditions, and improves the scalability of the algorithm.

Description

Technical Field

[0001] The present application relates to the technical field of drone swarm mission planning, and in particular to a method and device for distributed mission planning of drone swarms based on local consensus. Background Art

[0002] In the application of drone clusters, the computing pressure of drones in the cluster has been greatly reduced through distributed mechanisms and heuristic algorithms. The greater difficulty lies in the communication pressure of the cluster. With the increase in the number of drones and the scale of tasks, the communication volume and frequency between drones have increased sharply with the increase in the demand for task planning information relay and forwarding, which has brought great challenges to the limited cluster communication resources. Under the network conditions where the actual communication bandwidth is limited, if the communication volume and frequency between machines cannot be reduced to a certain limit, the drone cluster task planning algorithm will be difficult to ensure the real-time performance of large-scale cluster planning, and may even be impossible to apply in practice.

[0003] At present, most UAV swarm mission planning methods use global planning consensus, which means that all UAVs in the swarm must be completely consistent with the swarm mission plan, which means that the mission planning information of all UAVs needs to be propagated globally in the swarm. This requires UAVs to forward all mission planning information multiple times, and the communication volume and frequency increase sharply with the increase of swarm size. Large-scale swarms will bring about communication explosion problems, which seriously affects the performance and application of planning algorithms and causes poor scalability of planning algorithms. Summary of the invention

[0004] Based on this, it is necessary to provide a method and device for distributed task planning of drone clusters based on local consensus to address the above technical problems.

[0005] A distributed task planning method for drone swarm based on local consensus, the method comprising:

[0006] Drone i selects drones that are directly connected or within multi-hop communication range as its neighboring drones; the set of all neighboring drones is regarded as the local swarm of drone i ; Among them, the local cluster All drones within the selected mission are conflict-free.

[0007] Set the drone set and task set that need to be broadcast to empty sets, and set the initial value of counter c to 0.

[0008] In the mission plan calculation phase, UAV i calculates the best mission selection to form its own mission plan, and updates the UAV i and its mission to the set of UAVs and missions that need to be broadcast.

[0009] In the task planning consensus phase, drone i uses the preset conflict resolution rules to resolve conflicts when receiving task messages from its neighboring drones, and then updates the set of drones and tasks that need to be broadcast according to the preset information dissemination mechanism, and the counter c increases by 1.

[0010] When the drone is in the counter When the value of reaches the preset threshold, UAV i stops broadcasting, and if the preset iteration end condition is met, the final mission planning scheme is obtained; otherwise, UAV i continues to broadcast to the local swarm. The drones in the system broadcast their local information for the next iteration.

[0011] The preset conflict resolution rules are:

[0012] Drone i acts as the receiver, and other drones in the local swarm act as senders.

[0013] When the receiver receives the message about task j from the sender, if the receiver can determine that the sender's task message is the latest, then according to the basic criterion of the highest bidder wins, the drone with the highest bid will be regarded as the winner of the task.

[0014] If the receiver cannot be sure that the sender's mission message is the latest, it will determine whether the relevant drone timestamp is the latest based on the time criterion.

[0015] When the conflict between the basic criterion and the time criterion cannot be determined, the drone resets the local information about the corresponding task to facilitate subsequent re-judgment.

[0016] The preset information dissemination mechanism includes dissemination methods and dissemination rules; the dissemination method is used to determine the dissemination object; the dissemination rules are used to limit the dissemination scope.

[0017] The transmission methods include: unicast and broadcast; unicast is point-to-point information transmission, and only the designated receiver can receive the sender's message; broadcast is point-to-multipoint information transmission, and the information can be received as long as it is within the communication range of the sender.

[0018] The propagation rules include:

[0019] Drones i acts as a receiver, and other drones in the local swarm act as senders.

[0020] Before the receiver receives updated mission information from neighboring drones, the previously acquired mission information remains in the sender’s local bidding space; if the receiver does not receive an update message within a period of time, it is assumed that the other party has updated or disconnected.

[0021] When the receiver resolves the task conflict based on the sender's message, it still needs to broadcast so that the sender can get feedback in time and then stop broadcasting the conflicting task information, reducing the number of times the sender sends messages.

[0022] In one embodiment, the mission set of a single communication of the drone is:

[0023] ;

[0024] in, It is a set of tasks for a single communication of the UAV. For the task set, is a set of drones, i is the i-th drone, j is the j mission, k is the kth drone, For local clusters, , For drone k, it is the drone that obtains task j. The drone i considers to have obtained the task j.

[0025] In one embodiment, the communication protocol between drones is:

[0026] {task ID, task winner, task winning price} ∪ {drone ID, drone timestamp}.

[0027] The drone timestamp is:

[0028] ;

[0029] in, Timestamp for all drones that need to communicate, including the plan generation time of drone i itself , the plan generation time of the two machines with task conflicts and , For drone k, it is the drone that obtains task j. The drone i considers to have obtained the task j.

[0030] In one embodiment, the steps of improving the communication protocol between drones include:

[0031] The same amount of information is compressed by sorting and integrating to obtain the compressed communication protocol between drones; the compressed communication protocol between drones contains 4 vectors; the 4 vectors are: task ID vector, task winning price vector, drone composite information vector and drone timestamp vector; among them, the task ID vector needs to be sorted according to the planned tasks under the drone ID in the drone composite information, the task winning price corresponds to the task ID, and the drone timestamp corresponds to the drone ID; the drone composite information vector is obtained by integrating the drone ID and the task winner with low precision range requirements. The drone composite information is a float variable, the integer part of which is the drone ID, and the decimal part is the number of tasks in the drone plan.

[0032] A distributed task planning device for drone swarm based on local consensus, comprising:

[0033] The mission planning setting module is used by drone i to select drones within its direct communication or multi-hop range as its neighboring drones; the set of all neighboring drones is used as the local fleet of drone i ; Among them, the local cluster All drones within the selected mission are conflict-free.

[0034] The iterative initial condition setting module is used to set the drone set and task set that need to be broadcast to be empty sets, and set the initial value of the counter c to 0.

[0035] The task calculation module is used to calculate the best task selection for UAV i to form the local task plan during the task plan calculation phase, and update the local UAV i and its tasks to the set of UAVs and tasks that need to be broadcast.

[0036] The task plan consensus module is used in the task plan consensus link. UAV i uses the preset conflict resolution rules to resolve the task message received from its neighboring UAVs, and then updates the set of UAVs and task sets that need to be broadcast according to the preset information propagation mechanism. The counter c increases by 1. When UAV i is in the counter When the value of reaches the preset threshold, UAV i stops broadcasting, and if the preset iteration end condition is met, the final mission planning scheme is obtained; otherwise, UAV i continues to broadcast to the local swarm. The drones in the group broadcast their local information and proceed to the next round of iteration; the preset conflict resolution rules are: drone i is the receiver, and other drones in the local group are the senders; when the receiver receives the message about task j sent by the sender, if the receiver can determine that the sender's task message is the latest, then according to the basic criterion of the highest bidder, the drone with the highest bid will be the winner of the task; if the receiver cannot determine that the sender's task message is the latest, then according to the time criterion, it is determined whether the timestamp of the relevant drone is the latest; when the basic criterion and the time criterion conflict and cannot be determined, the drone resets the local information about the corresponding task to facilitate subsequent re-judgment; the preset information dissemination mechanism includes dissemination mode and dissemination rules; the dissemination mode is used to determine the dissemination object; the dissemination rule is used to limit the dissemination range; the dissemination modes include: unicast and broadcast; unicast is point-to-point information transmission, and only the designated receiver can receive the sender's message; broadcast is point-to-multipoint information transmission, as long as it is within the communication range of the sender, the information can be received; the dissemination rules include: drone i As the receiver, other drones in the local swarm act as senders; before the receiver receives the updated task information from the neighboring drone, the previously acquired task information is still in the sender's local bidding space; if the receiver does not receive an update message within a period of time, it is assumed that the other party has updated or disconnected; when the receiver resolves the task conflict based on the sender's message, it still needs to broadcast so that the sender can get feedback in time, and then stop broadcasting the conflicting task information, reducing the number of times the sender sends messages.

[0037] The above-mentioned method and device for distributed task planning of drone clusters based on local consensus proposes the concept of local consensus, that is, the drones and task scope of consensus are local. Unlike traditional global consensus, local consensus does not require the task planning information of all drones to be consistent, so the task information of other drones in the group can be further screened and integrated, and at the same time, it is only necessary to propagate the processed task information to the local area related to the task; based on the cluster local consensus mechanism, the drone inter-machine communication protocol, the inter-machine task conflict resolution rules and the information propagation mechanism within the cluster are designed to reduce the amount and frequency of inter-machine communication between drones, which greatly improves the scalability and real-time performance of the algorithm. This method reduces the amount and frequency of inter-machine communication required by the algorithm, adapts to the large-scale drone cluster self-organization rapid task planning under actual networking conditions, and improves the scalability of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flowchart of a method for distributed task planning of a drone cluster based on local consensus in one embodiment;

[0039] Figure 2This is a visual comparison diagram of global consensus and local consensus in another embodiment;

[0040] Figure 3 A schematic diagram comparing the communication protocol designed by the present application and the traditional communication protocol in another embodiment;

[0041] Figure 4 A schematic diagram of an improved communication protocol in another embodiment;

[0042] Figure 5 is a comparison diagram of the number of messages sent between the GHAC method and the traditional ACBBA method in another embodiment, wherein (a) is a comparison diagram of the number of messages sent under different n when p=3, (b) is a comparison diagram of the number of messages sent under different n when p=4, and (c) is a comparison diagram of the number of messages sent under different n when p=5;

[0043] Figure 6 is a comparison diagram of planning time between the GHAC method and the ACBBA method in another embodiment, wherein (a) is a comparison diagram of planning time under different n when p=3, (b) is a comparison diagram of planning time under different n when p=4, and (c) is a comparison diagram of planning time under different n when p=5;

[0044] Figure 7 is a comparison diagram of the required average communication bandwidth of the GHAC method and the ACBBA method in another embodiment, wherein (a) is a comparison diagram of the required average communication bandwidth under different n when p=3, (b) is a comparison diagram of the required average communication bandwidth under different n when p=4, and (c) is a comparison diagram of the required average communication bandwidth under different n when p=5;

[0045] Figure 8 is a comparison diagram of the required peak communication bandwidth of the GHAC method and the traditional ACBBA method in another embodiment, wherein (a) is a comparison diagram of the required peak communication bandwidth under different n when p=3, (b) is a comparison diagram of the required peak communication bandwidth under different n when p=4, and (c) is a comparison diagram of the required peak communication bandwidth under different n when p=5;

[0046] Fig. 9 The present invention is a structural block diagram of a distributed task planning device for a drone cluster based on local consensus in one embodiment. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0048] This application proposes a distributed task planning method for drone clusters based on local consensus, Global Harmonious Asynchronous Consensus, referred to as the GHAC method.

[0049] In one embodiment, Figure 1 As shown, a method for distributed task planning of drone clusters based on local consensus is provided, and the method includes the following steps:

[0050] Step 100: UAV i selects UAVs within its direct communication range or multi-hop range as its neighboring UAVs; the set of all neighboring UAVs is regarded as the local group of UAV i ; Among them, the local cluster All drones within the selected mission are conflict-free.

[0051] Specifically, cluster consensus requires drones to transmit information that they rely on to reach consensus, and the transmission of this information requires communication between drones. In a drone cluster, unless a drone is directly connected to all other drones, the information of the drone cannot be directly transmitted to all other drones through local broadcast, and can only be forwarded through relay drones. The amount of communication and frequency required for information forwarding depends on the communication topology of the drone cluster on the one hand, and on the information forwarding rules of the drones on the other.

[0052] When the communication topology of the drone cluster is a fully connected topology, that is, any two drones in the cluster are directly connected, the drone only needs to broadcast its own information, and all other drones can receive it. When the communication topology of the drone cluster is a row topology, that is, any drone in the cluster has at most two drones directly connected to it, and the drone broadcasts its own message, only its directly connected drone can receive it. Other drones need to forward it through relay drones to receive their own information. For example, in an N-machine row topology, the information of the first-end drone needs to be forwarded N-2 times before it can be delivered to the tail-end drone. In other communication topologies, such as ring topology and star topology, drone messages still need to be forwarded multiple times. In practical applications, the communication topology may be a combination of one or more of the above topologies, or it may be a variant slightly different from the above topologies.

[0053] It is feasible for drones to obtain global topology in small-scale clusters, but it is very difficult in large-scale clusters. The main reason is that the communication cost of drones to maintain cluster topology increases sharply with the number of drones, and usually only local routes within a few hops are obtained when the communication bandwidth is limited. Therefore, in large-scale clusters, it is difficult to set information forwarding rules through communication topology. In fact, the current distributed task planning method for drone clusters does not set information forwarding rules based on communication topology, but adapts the forwarding rules to all different communication topologies. Therefore, even in a fully connected topology, the current information forwarding rules will still require all drones to forward to ensure that all drones can receive updated information. In this way, in a cluster consisting of N drones, a message will be broadcast N times, which is no different from the number of broadcasts under the row topology. The messages of N drones will be broadcast N times. 2 The communication volume and number of communications of the drone cluster increase sharply with the increase of the number of drones, which is unacceptable in large-scale clusters.

[0054] Although the current distributed task planning method for drone swarms has reduced the amount and number of inter-machine communications through information fusion (i.e. fusing the local information with other drone information before broadcasting), the fused information still needs to be disseminated globally within the drone swarm, and all drones need to forward it. The communication frequency and volume still increase dramatically with the number of drones.

[0055] Consensus is the common understanding sought by all drones in the swarm. Most traditional drone swarm distributed task planning methods seek a consensus on the cluster plan, that is, all drones agree on the cluster task plan. This requires that the scope of the consensus drones and tasks is global, and drones need to reach a consensus on all task information within the global scope of the cluster, that is, a global consensus.

[0056] However, global consensus requires drones to transmit all task information to all drones in the cluster, which requires drones to forward all task information multiple times. The communication volume and frequency increase dramatically as the size of the cluster increases. In fact, drone clusters do not require all drones to agree on the cluster plan, as long as there is no conflict in task allocation between drones. In actual applications, as long as there is no task conflict in the repeated execution of the same task by drones in the cluster, whether all drones have consistent task information about the cluster does not affect the completion of the cluster task.

[0057] The contradiction between the limited communication resources of drone clusters and the communication volume required for the cluster's global planning consensus makes the current drone cluster task planning method have expansion bottlenecks in terms of drone and task scale, making it difficult to adapt to the self-organized rapid task planning of large-scale drone clusters under actual networking conditions.

[0058] This application proposes the concept of local consensus in a cluster, that is, the scope of the consensus drones and tasks is local. Unlike traditional global consensus, local consensus does not require the mission information of all drones to be consistent, but only expects that there are no mission conflicts between drones. Therefore, when the drone processes the received mission information of other drones in the cluster, it can further filter and integrate the information, and only needs to spread the processed mission information to the local area related to the task. In short, local consensus is to reduce the amount and frequency of communication between drones by resolving conflicts of related tasks within the local scope where the possibility of mission conflict is the greatest.

[0059] The scope of the drones and tasks of the consensus is local, and is achieved through the interaction of related tasks between drones and their neighboring drones. It seeks a reasonable relaxation of the global consensus, so that the scope and content of drone information transmission can be non-global, avoiding the rapid increase of communication volume and communication frequency with the cluster size.

[0060] Local task non-conflict is a reasonable relaxation of the global planning consensus, that is, the drone seeks to have no conflicts with its neighboring drones, that is, to reach a local consensus. The drone selects drones that are directly connected or within the multi-hop communication range as its neighboring drones. The set of neighboring drones k is the local swarm of drone i. , local cluster The tasks selected by all drones are conflict-free, and the mathematical description is as follows:

[0061] ;

[0062] ;

[0063] in, represents the communication hop count between UAV i and UAV k, d is the communication hop count threshold of the local swarm, when hour, That is, drone i and drone k are directly connected in communication (including drone i itself); is the local task set of UAV i, which is the local fleet of UAV i is the set of tasks selected by all drones in the group, and any task j is selected by only one drone.

[0064] Figure 2 By intuitively comparing the difference between global consensus and local consensus, we can find that local consensus ( Figure 2 (b)) and global consensus ( Figure 2Compared with (a), the consensus drones and task scope are local, and the optimization of the content and scope of information transmission between drones is sought to reduce the amount and frequency of inter-machine communication. Under local consensus, the drone cluster no longer seeks global planning consensus. There may be inconsistent task information between drones, but the task selected by the drone does not conflict with its neighboring drones, and the cluster task can still be executed while satisfying the constraints.

[0065] Step 102: Set the drone set and task set that need to be broadcast to be empty sets, and set the initial value of the counter c to 0.

[0066] Step 104: In the mission plan calculation phase, UAV i calculates the best mission selection to form its own mission plan, and updates the UAV i and its mission to the set of UAVs and missions that need to be broadcast.

[0067] Specifically, unlike the traditional global consensus method, the method proposed in this application does not need to communicate the winner information of all tasks, but only communicates the tasks selected by individual drones and the conflicting tasks between neighboring drones to reduce the amount of communication between drones. There are two reasons for choosing to communicate local tasks and conflicting tasks: on the one hand, the tasks selected by the drone itself need to be continuously broadcast to neighboring drones to resolve the task conflicts between them; on the other hand, the drone transmits the conflicting tasks between neighboring drones, which is beneficial for neighboring drones to resolve task conflicts within the local range.

[0068] Therefore, the mission set of a single communication of the UAV is as follows:

[0069] (1)

[0070] in, Contains all the task IDs that need to communicate. represents the set of conflicting tasks between neighboring UAVs, Represents the set of tasks selected by UAV i.

[0071] Step 106: In the task plan consensus phase, UAV i uses the preset conflict resolution rules to resolve conflicts with the task messages received from its neighboring UAVs, and then updates the set of UAVs and task sets that need to be broadcast according to the preset information dissemination mechanism, and the counter c increases by 1.

[0072] Among them, the preset conflict resolution rules are: drone i is the receiver, and other drones in the local fleet are the senders; when the receiver receives the message about task j sent by the sender, if the receiver can determine that the sender's task message is the latest, then according to the basic criterion of the highest bidder, the drone with the highest bid will be regarded as the winner of the task; if the receiver cannot determine that the sender's task message is the latest, then according to the time criterion, it is judged whether the timestamp of the relevant drone is the latest; when the basic criterion and the time criterion conflict and cannot be determined, the drone resets the local information about the corresponding task to facilitate subsequent re-judgment.

[0073] The preset information dissemination mechanism includes dissemination mode and dissemination rules; the dissemination mode is used to determine the dissemination object; the dissemination rules are used to limit the dissemination range; among them, the dissemination modes include: unicast and broadcast; unicast is point-to-point information transmission, and only the designated receiver can receive the sender's message; broadcast is point-to-multipoint information transmission, and the information can be received as long as it is within the communication range of the sender. Dissemination rules include: drones i As the receiver, other drones in the local swarm act as senders; before the receiver receives the updated task information from the neighboring drone, the previously acquired task information is still in the sender's local bidding space; if the receiver does not receive an update message within a period of time, it is assumed that the other party has updated or disconnected; when the receiver resolves the task conflict based on the sender's message, it still needs to broadcast so that the sender can get feedback in time, and then stop broadcasting the conflicting task information, reducing the number of times the sender sends messages.

[0074] Specifically, local consensus aims to resolve conflicts of related tasks in the local area where the possibility of task conflict is the greatest. When the drones in the cluster are distributed in a relatively wide area, drones that are relatively far away in the airspace do not actually need to know each other's task plans, because it is difficult for drones that are far away to have task conflicts. Inter-machine task conflicts usually occur between neighboring drones because their spatiotemporal relationships with nearby tasks are more similar, and the possibility of task competition is greater.

[0075] Therefore, by limiting the mission information content and dissemination range of drones, local consensus of drone clusters can be achieved, reducing the communication volume and frequency of relay drone message forwarding. For example, a group of neighboring drones is only interested in nearby missions within its capabilities, and can only disseminate mission information of interest; distant drones are not interested in the mission information of nearby drones, and the mission information of nearby drones does not need to be disseminated to distant drones, that is, drone mission plans do not need to be disseminated globally but only to drones of interest. The realization of local consensus mainly depends on the drone-to-drone communication protocol, dissemination mechanism, and conflict resolution rules, among which the communication protocol determines the content of inter-machine communication, the dissemination mechanism determines the scope of information dissemination, and the conflict resolution rules are the basis of drone consensus.

[0076] Step 108: When the drone is in the counter When the value of reaches the preset threshold, UAV i stops broadcasting, and if the preset iteration end condition is met, the final mission planning scheme is obtained; otherwise, UAV i continues to broadcast to the local swarm. The drones in the system broadcast their local information for the next iteration.

[0077] In the above-mentioned distributed task planning method for drone clusters based on local consensus, the method proposes the concept of local consensus, that is, the drones and task scope of consensus are local. Unlike traditional global consensus, local consensus does not require the task planning information of all drones to be consistent, so the task information of other drones in the group can be further screened and integrated, and at the same time, it is only necessary to propagate the processed task information to the local area related to the task; based on the cluster local consensus mechanism, the drone inter-machine communication protocol, the inter-machine task conflict resolution rules and the information propagation mechanism within the cluster are designed to reduce the amount and frequency of inter-machine communication between drones, which greatly improves the scalability and real-time performance of the algorithm. This method reduces the amount and frequency of inter-machine communication required by the algorithm, adapts to the large-scale drone cluster self-organization rapid task planning under actual networking conditions, and improves the scalability of the algorithm.

[0078] In one embodiment, the mission set of a single communication of the drone is:

[0079] (2)

[0080] in, It is a set of tasks for a single communication of the UAV. For the task set, is a set of drones, i is the i-th drone, j is the j mission, k is the kth drone, For local clusters, , For drone k, it is the drone that obtains task j. The drone i considers to have obtained the task j.

[0081] In one embodiment, the communication protocol between drones is: {task ID, task winner, task winning price}∪{drone ID, drone timestamp}. The drone timestamp is:

[0082] (3)

[0083] in, Timestamp for all drones that need to communicate, including the plan generation time of drone i itself , the plan generation time of the two machines with task conflicts and , For drone k, it is the drone that obtains task j. The drone i considers to have obtained the task j.

[0084] Specifically, after determining the tasks to be communicated, the drone needs to determine the information content of the communication task. The traditional global consensus method communicates the winner information of all tasks, that is, each drone usually communicates three vectors: task winner , task winning bid , and the task generation time Since these three vectors are in accordance with all m The task IDs are sorted, so there is no need for additional communication task IDs. Since the method proposed in this application does not transmit the information of all tasks, but only transmits the information of individual tasks and adjacent conflicting tasks, the task ID is necessary. In addition, this application adopts an asynchronous consensus method because it has less communication volume, communication frequency and consensus time than the traditional consensus method. Its communication timestamp is not the traditional task generation time, but the drone plan generation time. Therefore, under the local consensus mechanism, the original communication timestamp can be further simplified, and the simplified communication timestamp is shown in formula (3).

[0085] The reason why the timestamps of both conflicting parties are needed is to solve the problem of how the drones with conflicting missions can update after receiving a message with a lower price. For example, if a drone abandons the original mission for its own reasons or joins another mission and lowers the price of the original mission, how can other drones update the original mission information?

[0086] Since the drone timestamp used represents the generation time of the drone plan, and the plan contains all the tasks selected by the drone itself, the above communication task set needs to be converted into the tasks selected by the individual drone and the tasks selected by the neighboring drones with task conflicts. The expression is shown in formula (2).

[0087] Although changing the communication task from a set of conflicting tasks to all tasks of UAVs with task conflicts may increase some communication volume, changing the task timestamp to the UAV timestamp also reduces a certain amount of communication volume. More importantly, the more informative UAV timestamp combined with the agent-related conflict resolution rules can effectively reduce the frequency and volume of inter-machine communication.

[0088] Figure 3 The communication protocol designed for this application is compared with the traditional communication protocol. It can be found that this application requires all the original protocols to be transmitted. m The winners of the tasks and the corresponding winning bids are modified to only pass x in the set In the case of large-scale cluster tasks, the communication volume is 2 m becomes 3x ( m ≫x), which greatly reduces the communication pressure. At the same time, it does not transmit m Instead of using the winning timestamp of a task, the newly designed drone timestamp is used, and the communication volume can be reduced by a large m Become smaller u ( m ≫ u ), and the drone timestamp is richer in information than the mission timestamp, which can resolve conflicts between drones more quickly.

[0089] In one of the embodiments, the improvement step of the communication protocol between drones includes: compressing the same amount of information in a sorting and integrating manner to obtain the compressed communication protocol between drones; the compressed communication protocol between drones includes 4 vectors; the 4 vectors are: task ID vector, task winning price vector, drone composite information vector and drone timestamp vector; wherein, the task ID vector needs to be sorted according to the planned tasks under the drone ID in the drone composite information, the task winning price corresponds to the task ID, and the drone timestamp corresponds to the drone ID; the drone composite information vector is obtained by integrating the drone ID and the task winner with a low accuracy range requirement, and the drone composite information is a float type variable, the integer part of which is the drone ID, and the decimal part is the number of tasks in the drone plan.

[0090] Specifically, in actual engineering applications, drone clusters can further reduce the amount of inter-machine communication by data compression. Here, only the simplest method is proposed. By analyzing the proposed communication protocol, it can be seen that the protocol can be divided into drone-related and task-related. The former includes drone ID and drone timestamp, and the latter includes task ID, task winner and task winning price. It can be found that there may be data redundancy in the task winner, that is, multiple tasks belonging to the same drone will have data duplication in this task winner vector, while other protocol vectors do not have redundancy. Therefore, an attempt is made to compress the task winner vector. Analyzing the communication protocol vector, the drone timestamp is already longlong type with high precision requirements, the task winning price is float type and requires a certain precision, only the task ID, task winner and drone ID are int type. In the programming implementation, longlong type, float type and int type all occupy 4 bytes, but the precision range is different. Based on the above considerations and combined with practical feasibility, the drone ID and task winner vector with low precision range requirements are integrated to define a new variable Q with the same float type occupying 4 bytes. i , which is called UAV composite information, the integer part is the UAV ID, and the decimal part is the number of tasks in the UAV plan. Therefore, the proposed communication protocol is further modified to further reduce the communication volume of a single communication. The communication protocol is improved as follows Figure 4 As shown. The improved communication protocol compresses the same amount of information by sorting and integrating, and only contains 4 vectors, the task ID vector, the task winning price vector, the drone composite information vector and the drone timestamp vector. Among them, the task ID vector needs to be sorted according to the planned tasks under the drone ID in the drone composite information, the task winning price corresponds to the task ID, and the drone timestamp corresponds to the drone ID. It can be found that when it is necessary to transmit u When there are x mission information of a drone, the original protocol requires 3x + 2 u The communication volume of the improved protocol is only 2x + 2 u The communication volume is reduced again.

[0091] When sending a message, drone i needs to traverse the drone set , will be collected Each drone All tasks Traverse to ensure that the information of the tasks to be sent is the latest. This is because the message sending must be placed after the task plan calculation phase and before the task conflicts between machines are resolved, so that the latest local plan can be updated in the broadcast set, but this may also bring about the set Mission updates for other drones in the game.

[0092] When receiving a message, drone i needs to parse the information according to the communication protocol, and also needs to pay attention to resetting some task information in the cache. Mainly, when drone i receives all the task information of drone k, it needs to traverse all the tasks of drone k in the local cache. If a task j is not in the message just received, it is considered that drone k has removed the task, so it needs to update it in the local cache, that is, .

[0093] In a specific embodiment, consensus rules are the basis of drone cluster consensus, and this application decouples them into conflict resolution rules and information propagation mechanisms. Task conflict resolution rules are an important cornerstone of drone cluster consensus, which enables drones to handle conflicts between their own task information and other drone information received. Table 1 is a task conflict resolution rule table, which mainly describes the behavior of drone i (receiver) after receiving a message about task j from drone k (sender), where represents the winner of task j considered by drone k, Indicates the drone that drone k thinks of Update time, The meanings are similar.

[0094] After receiving the message about task j from drone k (sender), drone i (receiver) takes the default action of retaining, i.e. drone i retains its original local information about task j. The next two actions are Action 1 and Action 2, where the former is an update action, i.e. updating the information about task j received from drone k to drone i's local data, and the latter is a reset action, i.e. resetting drone i's local information about task j.

[0095] When the receiver can determine that the sender's task message is the latest, task conflict resolution is mainly based on the basic criterion of the highest bidder, that is, the winner of the task is the drone with the highest bid; otherwise, task conflict resolution also needs to add the time criterion of the task-related drones, that is, whether the timestamp of the relevant drone is the latest. When the basic criterion and the time criterion conflict cannot be determined, the drone resets the local information about the corresponding task to facilitate subsequent re-judgment.

[0096] It is worth mentioning that in the UAV swarm mission planning, the price of the mission will change with the local plan, and the mission price will become lower or even reset. At this time, the UAV needs to broadcast the price change or mission reset. Although the neighboring UAVs can update and broadcast the mission information after receiving the message, the subsequent UAVs cannot update and broadcast due to mission information conflicts.

[0097] Table 1 Conflict resolution rules after UAV i receives a message from UAV k

[0098]

[0099] Note: none indicates other situations. Action 1 is the update behavior, which is to update the information received from drone k about task j to the local drone i. Action 2 is the reset behavior, which is to reset the local information about task j of drone i. Action 3 is the default behavior of drone i (receiver) after receiving the message about task j sent by drone k (sender), which is to retain, that is, drone i retains its original local information about task j.

[0100] In a specific embodiment, the information dissemination mechanism within the drone cluster mainly includes the dissemination method and the dissemination rules. The former mainly determines the dissemination object, and the latter mainly limits the dissemination range. The two together constitute the dissemination of information. There are mainly two types: unicast and broadcast. The former is point-to-point information transmission. Only the designated receiver can receive the message from the sender, which is equivalent to the limitation of the dissemination object; the latter is point-to-multipoint information transmission. As long as it is within the communication range of the sender, the information can be received, which is equivalent to the indefinite dissemination object. The behavioral basis for whether a drone forwards the conflict-free task information processed by the drone after receiving the message from other drones. The dissemination rules are the basis for limiting the dissemination range of task information and resolving task conflicts, and are an important part of achieving consensus in drone clusters.

[0101] In order to reduce the communication volume and frequency caused by task status disputes, the task status is further differentiated, mainly distinguishing whether the task is in the initialization state or the reset state. When the drone initializes the winner of all local tasks, the initial value is set to -1 instead of 0 (0 means that the task is currently considered to be not executed by a drone, not the initial state, but the task reset state). Next, from the perspective of drone i (receiver), the blocking propagation behavior of drone i after receiving information about task j from drone k (sender) will be described. The five principles of blocking propagation are as follows:

[0102] 1) Initial situation changes: UAV i (receiver) updates the local winner of task j from 0 to non-zero , is regarded as a task conflict between non-UAVs, and is only regarded as an update of task information, and the information of task j is not broadcast. In order of sequence number, it is called the first blocking principle, which is similar to the following. The specific mathematical expression is:

[0103] (4)

[0104] in, It means that drone i will not spread the information of task j, the same below; It means that under the three or conditions of the first blocking principle, drone i updates the information of task j to drone k’s information about task j. .

[0105] 2) Task information reset: UAV i (receiver) resets the local information about the winner of task j. At this time, UAV i itself has difficulty in determining the winner of task j and does not broadcast the winner information of task j. The specific mathematical expression is:

[0106] (5)

[0107] in, It means that under the three "or" conditions of the second blocking principle, drone i resets the information of task j. , .

[0108] 3) The three-party dispute has been resolved: When drone i (receiver) or drone k (sender) believes that the other party is the winner of task j, and the other party believes that the third-party drone m is the winner, if the timestamp of the drone that is determined to be the winner is updated, it will prevail and it can be considered that the task conflict with the third-party drone has been resolved, and the winner information of task j will not be broadcast. The specific mathematical expression is:

[0109] (6)

[0110] in, Indicates that the condition Next, drone i retains its original information about mission j; Indicates that the condition Next, UAV i updates the information of task j with the information of UAV k about task j.

[0111] 4) Only different timestamps: When the local winner information of task j of drone k (sender) and drone i (receiver) is consistent, but the timestamps are inconsistent, the latest timestamp shall prevail, but the winner information of task j shall not be broadcast. The specific mathematical expression is:

[0112] (7)

[0113] in, Indicates that the condition Next, drone i updates the information of task j with drone k’s information about task j; Indicates that the condition Next, drone i retains its original information about mission j.

[0114] 5) Task conflict resolved: When drone k (sender) believes that the winner of task j is drone m, and drone i (receiver) believes that the winner of task j is drone When the timestamps of drone k, drone m and drone n are all newer than those of drone i, the task conflict between drone m and drone n is considered to be resolved, and drone i updates but no longer broadcasts the winner information of task j. The specific mathematical expression is:

[0115] (8)

[0116] in, Indicates that the condition Next, UAV i updates the information of task j with the information of UAV k about task j.

[0117] Based on the above five principles of communication blocking, the communication rules based on cluster local consensus are redesigned, as shown in Table 2. It mainly describes the communication behavior of drone i (receiver) after receiving the message about task j sent by drone k (sender), where represents the winner of task j considered by drone k, represents the update time of drone m as considered by drone k, The meanings are similar.

[0118] Table 2 Propagation rules of drone i after receiving the message of task j sent by drone k

[0119]

[0120] In addition, in order to avoid the drone failing to receive the sender's information due to communication reasons such as packet loss, before receiving the updated task information of the neighboring drone, the previously obtained task information is still in the sender's local bidding space, that is, it is still considered that the other party has not resolved the task conflict and needs to continue broadcasting. Of course, if the drone does not receive an update message for a period of time, it is assumed by default that the other party has updated or disconnected. Therefore, when the receiver resolves the task conflict based on the sender's message, it still needs to broadcast so that the sender can get feedback in time, and then stop broadcasting the conflicting task information, reducing the number of times the sender sends messages.

[0121] In one embodiment, the overall process of the cluster asynchronous task planning algorithm under local consensus is shown in Algorithm 1, which mainly involves the iteration of task plan calculation and task plan consensus until the algorithm converges. The difference is that the drone needs to broadcast tasks (determined by the information dissemination mechanism within the cluster) and reconstruct the communication protocol.

[0122] Algorithm 1: Cluster asynchronous task planning algorithm under local consensus:

[0123] 1: Drone i initializes the trigger matrix and , convergence flag ,counter , iteration counter , broadcast logo and counter thresholds and convergence threshold ;

[0124] 2: While algorithm has not converged ;

[0125] 3: Initialize the set of drones that need to be broadcast and task set ;

[0126] 4: Step 1: Task planning calculation;

[0127] 5: Calculate the local task plan and update the local task information and ;

[0128] 6: Update the machine and its mission to the set of drones that need to be broadcast and task set ;

[0129] 7: Stage 2: Consensus on task plan;

[0130] 8: ;

[0131] 9: Send and Task information under scope and to other drones;

[0132] 10: end if;

[0133] 11: Get the mission information of other drone k and , and reset the trigger matrix ;

[0134] 12: for drones directly connected to dronei ;

[0135] 13: for each task ;

[0136] 14: Resolve task conflicts according to the conflict resolution rules (Table 1) and update and ;

[0137] 15: Update UAV set and mission set and ← Information dissemination mechanism (Table 2);

[0138] 16: if ;

[0139] 17: and ;

[0140] 18: end if;

[0141] 19: end for;

[0142] 20: end for;

[0143] 21: end while;

[0144] First, drone i initializes local parameters (line 1), including the trigger matrix and , convergence flag ,counter , iteration counter , broadcast logo and counter thresholds and convergence threshold The trigger matrix Used to record the reason for triggering the sender to broadcast. When the drone needs to send the information of its mission j to drone k, . Then, if the algorithm does not converge (convergence flag When the number of UAVs is 0, the task plan calculation (lines 4-6) and task plan consensus (lines 7-21) are iterated continuously. Different from the traditional global consensus algorithm, the algorithm proposed in this application only broadcasts the task plan information of the local UAV and the task plan information of the conflicting UAVs, so a new UAV cluster needs to be broadcast. and task set , and performs initialization reset before each two-stage loop iteration (line 3). In the mission plan calculation phase, the drone calculates the best mission selection to form the local mission plan (line 5), and updates the local drone and its mission to the set of drones that need to be broadcast. and task set In the task plan consensus phase, the drone will resolve conflicts with the task information received from other drones k (line 14), and then update the set of drones that need to be broadcast according to the information dissemination mechanism. and task set (Line 15). Finally, drone i is in counter The value reaches the threshold When the drone i no longer broadcasts ( ), and iterate the counter Start recording until threshold Confirm that the algorithm converges ( ); otherwise, drone i continues to broadcast its local information, except that the range when sending mission information is from and (Line 9), rather than all the task information of the traditional algorithm.

[0145] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0146] In a verification embodiment, in order to verify whether the distributed task planning method of drone cluster based on local consensus proposed in this application can effectively deal with the communication explosion problem under the expansion of cluster scale, a drone cluster search and rescue experimental scenario is adopted. The main tasks of the drone cluster are to provide medical support and food support, where the duration of the medical task is 300s, the duration of the food task is 350s, the speed of the drone providing medical services is 30m / s, and the speed of the drone providing food is 50m / s. The number of drones n is 32-128 (step length is 32), the number of tasks m is 64-640 (step length is 32), and the ratio p of the number of tasks to the number of drones is still 3-5 (step length is 1).

[0147] Therefore, the experiment randomly distributed the drones in a space of 40km×10km×0m, and the tasks were randomly distributed in a space of 40km×10km×1km. The experiment conducted a large number of tests under various experimental conditions such as drone positions and task conditions, in which 10 Monte Carlo experiments were conducted with randomly distributed drones and tasks under each of the same p and n conditions. The communication range of the drone is still set to 3km. When the drones are randomly distributed, it is ensured that at least one drone is within its communication range, thereby ensuring the possibility of global communication of the drone cluster. The algorithm performance test mainly compares the number of messages sent, the total communication volume, and the planning time. The distributed task planning method for drone clusters based on local consensus proposed in this application (Global Harmonious Asynchronous Consensus, GHAC) and the most representative asynchronous consistent bundling algorithm (ACBBA) are compared and analyzed.

[0148] Figure 5 The difference in the number of messages sent between the proposed GHAC method and the traditional ACBBA method is compared, where (a) is a comparison chart of the number of messages sent under different n when p=3, (b) is a comparison chart of the number of messages sent under different n when p=4, and (c) is a comparison chart of the number of messages sent under different n when p=5. It can be seen that the number of messages sent by the GHAC method is less than that of the ACBBA method under all different drones and mission scales, especially when the number of drones and missions is large, the advantage of the proposed method is more obvious. This is mainly due to the local consensus mechanism of GHAC, which does not require global consistency like ACBBA, allowing GHAC to design an information dissemination mechanism with interrupt diffusion capability, avoiding the global dissemination and global consistency process of many unnecessary information. When p=3 and n=64, the average number of messages required for the ACBBA method is 5438.1 times, and that for the GHAC method is 1792 times, which is a decrease of 67.05%, the highest value in the experiment; when p=5 and n=128, the average number of messages required for the ACBBA method is 17244.8 times, and that for the GHAC method is 9484.2 times, which is a decrease of 45.00%, the lowest value in the experiment.

[0149] Figure 6The comparison of planning time between the GHAC method and the ACBBA method is shown, where (a) is a comparison chart of planning time under different n when p=3, (b) is a comparison chart of planning time under different n when p=4, and (c) is a comparison chart of planning time under different n when p=5. It can be seen that the planning time of the GHAC method is less than that of the ACBBA method under all different drones and task scales, especially when the number of drones and tasks is large, the advantages of the method proposed in this application are more obvious. This is due to the limited local propagation mechanism of the GHAC method, which enables faster conflict resolution in the local scope with task conflicts, without having to wait for global tasks to be consistent in the global scope like ACBBA, so that GHAC can obtain planning results faster. When p=3 and n=32, the average planning time required for the ACBBA method is 6379.9ms, and that for the GHAC method is 3948.1ms, which is a decrease of 38.12%, the lowest value in the experiment. When p=3 and n=64, the planning time required for the ACBBA method is 10343.1ms, and that for the GHAC method is 4587.2ms, which is a decrease of 55.65%, the highest value in the experiment.

[0150] Figure 7 The comparison between the GHAC method and the ACBBA method in terms of the required average communication bandwidth is shown, where (a) is a comparison of the average communication bandwidth required for different n when p=3, (b) is a comparison of the average communication bandwidth required for different n when p=4, and (c) is a comparison of the average communication bandwidth required for different n when p=5. It can be seen that the average communication bandwidth required by the GHAC method is less than that of the ACBBA method under different UAVs and mission scales. When p=3 and n=32, the average communication bandwidth required by the ACBBA method is 0.63Mbps, and that of the GHAC method is 0.31Mbps, a decrease of 51.12%, the highest value in the experiment; when p=4 and n=128, the average communication bandwidth required by the ACBBA method is 5.77Mbps, and that of the GHAC method is 5.35Mbps, a decrease of 7.33%, the lowest value in the experiment. As the number of UAVs increases, the difference in the average communication bandwidth required by the GHAC method and the ACBBA method becomes smaller. Combining the comparison chart of the two methods in terms of planning time, it is easy to see that ( Figure 5 ), it is not that the advantages of the proposed GHAC method become weaker as the cluster size increases, but that the ACBBA method requires a sharp increase in planning time as the cluster size increases, thereby lowering the average communication bandwidth required. The GHAC method proposed in this application requires a short planning time, so the average communication bandwidth required will inevitably increase, but it is still more advantageous than the traditional ACBBA method.

[0151] Figure 8The difference in the required peak communication bandwidth between the GHAC method and the ACBBA method is compared, where (a) is a comparison chart of the required peak communication bandwidth under different n when p=3, (b) is a comparison chart of the required peak communication bandwidth under different n when p=4, and (c) is a comparison chart of the required peak communication bandwidth under different n when p=5. It can be seen that the peak communication bandwidth required by the GHAC method is less than that of the ACBBA method under all different drones and task scales, especially when the number of drones and tasks is large, the advantages of the method proposed in this application are more obvious. This is mainly due to the fact that the GHAC based on local consensus can only send the local task planning information and the task conflict information between neighboring machines, while ACBBA needs to send the global task information of the cluster, which enables GHAC to greatly reduce the single communication volume. In addition, due to the mechanism of local consensus, the number of messages sent by GHAC is also lower, which further reduces the peak communication bandwidth required by GHAC.

[0152] In one embodiment, Fig. 9 As shown, a distributed task planning device for a drone cluster based on local consensus is provided, including: a task planning setting module, an iterative initial condition setting module, a task calculation module and a task plan consensus module, wherein:

[0153] The mission planning setting module is used by drone i to select drones within its direct communication or multi-hop range as its neighboring drones; the set of all neighboring drones is used as the local fleet of drone i ; Among them, the local cluster All drones within the selected mission are conflict-free.

[0154] The iterative initial condition setting module is used to set the drone set and task set that need to be broadcast to be empty sets, and set the initial value of the counter c to 0.

[0155] The task calculation module is used to calculate the best task selection for UAV i to form the local task plan during the task plan calculation phase, and update the local UAV i and its tasks to the set of UAVs and tasks that need to be broadcast.

[0156] The task plan consensus module is used in the task plan consensus link. UAV i uses the preset conflict resolution rules to resolve the task message received from its neighboring UAVs, and then updates the set of UAVs and task sets that need to be broadcast according to the preset information propagation mechanism. The counter c increases by 1. When UAV i is in the counter When the value of reaches the preset threshold, UAV i stops broadcasting, and if the preset iteration end condition is met, the final mission planning scheme is obtained; otherwise, UAV i continues to broadcast to the local swarm. The drones in the group broadcast their local information to carry out the next round of iteration; wherein, the preset conflict resolution rules are: drone i is the receiver, and other drones in the local group are the senders; when the receiver receives the message about task j sent by the sender, if the receiver can determine that the sender's task message is the latest, then according to the basic criterion of the highest bidder, the drone with the highest bid will be the winner of the task; if the receiver cannot determine that the sender's task message is the latest, then according to the time criterion, it is judged whether the timestamp of the relevant drone is the latest; when the basic criterion and the time criterion conflict and cannot be determined, the drone resets the local information about the corresponding task to facilitate subsequent re-judgment; the preset information dissemination mechanism includes dissemination mode and dissemination rules; the dissemination mode is used to determine the dissemination object; the dissemination rules are used to limit the dissemination range; wherein, the dissemination modes include: unicast and broadcast; unicast is point-to-point information transmission, and only the designated receiver can receive the sender's message; broadcast is point-to-multipoint information transmission, as long as it is within the communication range of the sender, the information can be received. Dissemination rules include: drones i As the receiver, other drones in the local swarm act as senders; before the receiver receives the updated task information from the neighboring drone, the previously acquired task information is still in the sender's local bidding space; if the receiver does not receive an update message within a period of time, it is assumed that the other party has updated or disconnected; when the receiver resolves the task conflict based on the sender's message, it still needs to broadcast so that the sender can get feedback in time, and then stop broadcasting the conflicting task information, reducing the number of times the sender sends messages.

[0157] In one embodiment, the task set of a single communication of the UAV is shown in formula (2).

[0158] In one embodiment, the communication protocol between drones is: {task ID, task winner, task winning price}∪{drone ID, drone timestamp}, where the drone timestamp is shown in formula (3).

[0159] In one of the embodiments, the improvement step of the communication protocol between drones includes: compressing the same amount of information in a sorting and integrating manner to obtain the compressed communication protocol between drones; the compressed communication protocol between drones includes 4 vectors; the 4 vectors are: task ID vector, task winning price vector, drone composite information vector and drone timestamp vector; wherein, the task ID vector needs to be sorted according to the planned tasks under the drone ID in the drone composite information, the task winning price corresponds to the task ID, and the drone timestamp corresponds to the drone ID; the drone composite information vector is obtained by integrating the drone ID and the task winner with a low accuracy range requirement, and the drone composite information is a float type variable, the integer part of which is the drone ID, and the decimal part is the number of tasks in the drone plan.

[0160] For the specific limitations of the distributed task planning device for drone swarms based on local consensus, please refer to the limitations of the distributed task planning method for drone swarms based on local consensus above, which will not be repeated here. Each module in the above-mentioned distributed task planning device for drone swarms based on local consensus can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0161] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0162] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A distributed task planning method for drone swarm based on local consensus, characterized in that: The method comprises: Drone i selects drones that are directly connected or within multi-hop communication range as its neighboring drones; the set of all neighboring drones is regarded as the local swarm of drone i ; Among them, the local cluster The tasks selected by all drones are conflict-free; Set the drone set and task set to be broadcasted to be empty sets, and set the initial value of the counter c to 0; In the mission plan calculation phase, UAV i calculates the best mission selection to form its own mission plan, and updates the UAV i and its mission to the set of UAVs and missions that need to be broadcast; In the task plan consensus phase, drone i uses the preset conflict resolution rules to resolve the task message received from its neighboring drones, and then updates the drone set and task set that need to be broadcast according to the preset information dissemination mechanism, and the counter c increases by 1; When the drone is in the counter When the value of reaches the preset threshold, UAV i will no longer broadcast, and if the preset iteration end condition is met, the final mission planning scheme is obtained; otherwise, UAV i continues to broadcast to the local swarm. The drones in the area broadcast their local information for the next iteration; The preset conflict resolution rule is as follows: drone i is used as the receiver, and other drones in the local fleet are used as senders; after receiving the message about task j from the sender, if the receiver can determine that the task message of the sender is the latest, then according to the basic criterion of the highest bidder, the drone with the highest bid is selected as the winner of the task; if the receiver cannot determine that the task message of the sender is the latest, then according to the time criterion, it is determined whether the timestamp of the relevant drone is the latest; when the basic criterion and the time criterion conflict and cannot be determined, the drone resets the local information about the corresponding task to facilitate the subsequent re-judgment; The preset information dissemination mechanism includes a dissemination method and a dissemination rule; the dissemination method is used to determine the dissemination object; the dissemination rule is used to limit the dissemination scope; The transmission modes include: unicast and broadcast; unicast is point-to-point information transmission, and only the designated receiver can receive the message from the sender; broadcast is point-to-multipoint information transmission, and anyone within the communication range of the sender can receive the message; The propagation rules include: Drones i acts as a receiver, and other drones in the local swarm act as senders; Before the receiver receives updated mission information from neighboring drones, the previously acquired mission information is still in the sender’s local bidding space; if the receiver does not receive an update message within a period of time, it is assumed that the other party has updated or disconnected; When the receiver resolves the task conflict based on the sender's message, it still needs to broadcast so that the sender can get feedback in time and then stop broadcasting the conflicting task information, reducing the number of times the sender sends messages.

2. The method for distributed task planning of drone swarm based on local consensus according to claim 1 is characterized in that: The mission set of a single communication of the UAV is: ； in, It is a set of tasks for a single communication of the UAV. For the task set, For drone collection, For the A drone, For the tasks, For the A drone, For local clusters, , For drones Consider the drone that gets task j, For drones Consider the drone that gets task j.

3. The method for distributed task planning of drone swarm based on local consensus according to claim 2 is characterized in that: The communication protocol between drones is: {task ID, task winner, task winning price} ∪ {drone ID, drone timestamp}; The drone timestamp is: ； in, Timestamp for all drones that need to communicate, including drones Own plan generation time , the plan generation time of the two machines with task conflicts and , For drones Consider the drone that gets task j, For drones Think you get the task drone.

4. The method for distributed task planning of drone swarm based on local consensus according to claim 3 is characterized in that: The steps to improve the communication protocol between drones include: The same amount of information is compressed by sorting and integrating to obtain the compressed communication protocol between drones; the compressed communication protocol between drones includes 4 vectors; the 4 vectors are: task ID vector, task winning price vector, drone composite information vector and drone timestamp vector; wherein, the task ID vector needs to be sorted according to the planned tasks under the drone ID in the drone composite information, the task winning price corresponds to the task ID, and the drone timestamp corresponds to the drone ID; the drone composite information vector is obtained by integrating the drone ID and the task winner with low accuracy range requirements, and the drone composite information is a float variable, the integer part of which is the drone ID, and the decimal part is the number of tasks in the drone plan.

5. A distributed task planning device for drone swarm based on local consensus, characterized in that: The device comprises: Mission planning and setup module for drones Select the drones that are directly connected or within the multi-hop communication range as its neighboring drones; the set of all neighboring drones is regarded as the drone Local cluster ; Among them, the local cluster The tasks selected by all drones are conflict-free; Iteration initial condition setting module, used to set the drone set and task set to be broadcasted to be empty sets, and set the initial value of the counter c to 0; The mission calculation module is used to calculate the mission plan of the UAV. Calculate the best mission selection to form the mission plan of the own aircraft, and update the own aircraft and its mission to the set of drones and missions that need to be broadcast; Mission plan consensus module, used in the mission plan consensus link, drone The preset conflict resolution rules are used to resolve the mission messages received from its neighboring drones, and then the drone set and mission set to be broadcast are updated according to the preset information propagation mechanism, and the counter c is increased by 1; when drone i is in the counter When the value reaches the preset threshold, the drone No longer broadcast, and if the preset iteration end condition is met, the final mission planning solution is obtained; otherwise, UAV i continues to send The drones in the local group broadcast their local information and perform the next round of iteration; the preset conflict resolution rule is: drone i is the receiver, and other drones in the local group are the senders; when the receiver receives the message about task j sent by the sender, if the receiver can determine that the sender's task message is the latest, then according to the basic criterion of the highest bidder, the drone with the highest bid for the task winner will be the winner of the task; if the receiver cannot determine that the sender's task message is the latest, then according to the time criterion, it is determined whether the timestamp of the relevant drone is the latest; when the basic criterion and the time criterion conflict and cannot be determined, the drone resets the local information about the corresponding task to facilitate subsequent re-judgment; the preset information propagation mechanism includes propagation mode and propagation rules; the propagation mode is used to determine the propagation object; the propagation rule is used to limit the propagation range; the propagation mode includes: unicast and broadcast; the unicast is a point-to-point information transmission, and only the designated receiver can receive the sender's message; the broadcast is a point-to-multipoint information transmission, as long as it is within the communication range of the sender, the information can be received; the propagation rules include: drone i As the receiver, other drones in the local swarm act as senders; before the receiver receives the updated task information from the neighboring drone, the previously acquired task information is still in the sender's local bidding space; if the receiver does not receive an update message within a period of time, it is assumed that the other party has updated or disconnected; when the receiver resolves the task conflict based on the sender's message, it still needs to broadcast so that the sender can get feedback in time, and then stop broadcasting the conflicting task information, reducing the number of times the sender sends messages.

Images