Dynamic consensus execution method operating on multiple unmanned aerial vehicles

Abstract

The invention discloses a dynamic consensus execution method operating on multiple unmanned aerial vehicles, which ensures that an unmanned aerial vehicle cluster always has one leader node by using a Raft-based leader election process, ensures the fault tolerance of the unmanned aerial vehicle cluster, and ensures the consistency of the unmanned aerial vehicle cluster by using a Paxos algorithm-based consensus process. The communication traffic of the consensus process is in direct proportion to the number of cluster nodes, the balance of performance and stability is achieved, the communication content between the unmanned aerial vehicles is prevented from being stolen based on AES encryption transmission, the communication safety between the unmanned aerial vehicles is ensured, the unmanned aerial vehicles can actively propose decision opinions through the truth value discovery function, and the communication efficiency is improved. A more reasonable and comprehensive decision is made after the opinions of multiple parties are integrated, the scientificity and effectiveness of task execution are improved, the AI tracking function is based on the two AI models of Yolo and SiamRPN, accurate and real-time tracking of the target is achieved, and the unmanned aerial vehicle cluster can execute more complex tasks.

Description

Technical Field

[0001] The present invention relates to the field of drone cluster technology, and in particular to a dynamic consensus execution method running on multiple drones. Background Art

[0002] With the rapid development of drone technology, the demand for multi-drone collaborative work is growing. However, in the collaborative work of multiple drones, how to ensure the overall consistency and fault tolerance of the cluster, achieve efficient decision initiation and execution, and ensure the security of communication have become technical problems that need to be solved urgently.

[0003] The prior art has the following technical problems:

[0004] 1. The current consensus algorithms usually serve the fields of blockchain, distributed backup and disaster recovery, but are rarely used in drone clusters.

[0005] 2. In the existing semi-centralized consensus algorithms, the central node often unilaterally outputs proposals during the consensus process, without considering that all nodes can make proposals.

[0006] 3. Currently, most drone cluster tracking requires manual target selection Summary of the invention

[0007] In view of the shortcomings of the existing technology, the present invention has made improvements and provided a dynamic consensus execution method running on multiple drones, which uses Paxos's Raft as a consensus algorithm to ensure the overall consistency and fault tolerance of the cluster; in the process of initiating decision-making, the present invention can integrate the decision opinions of each drone through the truth value discovery capability, solving the problem of a single source of proposals for traditional consensus algorithms. The present invention is implemented through the following technical solutions:

[0008] The invention discloses a decision-making method running on multiple drones, comprising:

[0009] Initiate a request to become the master node through one of the drones;

[0010] Send the request to other drones, which generate affirmative votes through the election algorithm;

[0011] The approval votes are sent to the drone. When the number of approval votes received by the drone exceeds half, the drone is determined to be the leader node.

[0012] Proposals are generated in real time by all drones and sent to the leader node in real time;

[0013] The leader node reads the pre-recorded or real-time generated proposal, encrypts it, obtains the encrypted proposal, and transmits it to other drones. The proposal content is the parameters of the command to be executed.

[0014] Through the consensus process, other drones vote in favor of the encrypted proposal and send it to the leader node;

[0015] When the number of votes received by the leader node exceeds half, the leader node generates a resolution based on the encrypted content of the proposal and sends it to other drones;

[0016] The resolution content is analyzed by other drones to carry out specific tasks.

[0017] As a further improvement, the present invention provides that the request is sent to other drones, and other drones generate affirmative votes through an election algorithm as follows:

[0018] The drone initiates a request to become the master node and attaches its proposal_id = promised_id + 1. All drones maintain two values, proposal_id and promised_id.

[0019] After receiving the request, other drones will determine whether the proposal_id is greater than their promised_id. If so, they will return a positive vote and update their promised_id to the corresponding proposal_id.

[0020] The drone that initiates the request will become the leader node if it receives more than half of the votes in favor.

[0021] As a further improvement, the method of the present invention of using other drones to vote in favor of the encrypted proposal through a consensus process and sending it to the leader node is specifically as follows:

[0022] The leader node continuously takes out proposals to be agreed upon from its own consensus queue and broadcasts them to other drones;

[0023] Other drones decide whether to return a positive vote for the message based on their own pre-set programs;

[0024] When the leader receives more than half of the votes in favor, it generates a resolution message based on the proposal content and broadcasts it to all slave nodes.

[0025] As a further improvement, the real-time generation of proposals and sending them to the leader node in real time described in the present invention is specifically as follows:

[0026] All drone nodes participating in the consensus can generate proposals, and then package the proposals into a truth value discovery message and send it to the leader node;

[0027] The leader node will continue to collect for up to 30 seconds after receiving the truth value discovery message for the first time. If it collects more than 1 / 3 of the total proposals, it will put the proposal corresponding to the truth value discovery into its own consensus queue.

[0028] The beneficial effects of the present invention are as follows:

[0029] 1. The Raft-based leader election process ensures that the drone cluster always has a leader node, thereby ensuring the fault tolerance of the drone cluster;

[0030] 2. The consensus process based on the Paxos algorithm is used to ensure the consistency of the drone cluster and avoid mission errors caused by inconsistent data or node failures. The communication volume of the consensus process is proportional to the number of cluster nodes, achieving a balance between performance and stability.

[0031] 3. AES-based encrypted transmission prevents the communication content between drones from being stolen, ensuring the security of communication between drones;

[0032] 4. The true value discovery function enables each drone to proactively propose decision-making opinions, and make more reasonable and comprehensive decisions after integrating opinions from multiple parties, thus improving the scientificity and effectiveness of task execution;

[0033] 5. The AI ​​tracking function is based on two AI models, Yolo and SiamRPN, which achieves accurate and real-time tracking of targets, allowing drone clusters to perform more complex tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is the main flow chart of the scheme of the present invention;

[0035] Figure 2 It is a consensus flow chart of the scheme of the present invention;

[0036] Figure 3 This is the AI ​​tracking flow chart of the solution of the present invention. DETAILED DESCRIPTION

[0037] The technical solution of the present invention is further described below by means of specific embodiments in conjunction with the accompanying drawings:

[0038] Figure 1This is the main flow chart of the solution of the present invention; in a cluster consisting of three drones A, B, and C, A initiates a request to become the master node and attaches its own proposal_id, which has a value of promised_id+1=1 (A, B, and C will maintain two values, proposal_id and promised_id, which are both 0 at the beginning);

[0039] After receiving the request, B and C will determine whether the proposal_id is greater than their own promised_id. If so, they will return a vote of approval and update their own promised_id to 1. Since the received proposal_id = 1 > their own promised_id = 0, they will return a vote of approval.

[0040] If A receives more than half of the votes in favor, it becomes the leader node;

[0041] A, B, and C can all generate proposals, and then package the proposals into a truth-finding message and send it to A;

[0042] A will continue to collect for up to 30 seconds after receiving the truth value discovery message for the first time. If it collects more than 1 / 3 of the total number of proposals, it will put the proposal corresponding to the truth value discovery into its own consensus queue.

[0043] Figure 2 This is the consensus flow chart of the solution of the present invention. A reads a pre-recorded or real-time generated proposal, attaches its own proposal_id=1 to encrypt it, obtains the encrypted proposal, and transmits it to B and C. The proposal content is the parameters of the command to be executed.

[0044] B and C will determine whether proposal_id is equal to their own promised_id. If so, they will return a vote of approval. Since proposal_id = 1 = their own promised_id, B and C will return a vote of approval.

[0045] When more than half of the votes A receives are in favor, A generates a resolution based on the encrypted content of the proposal and sends it to B and C.

[0046] Analyze the resolution content through B and C and perform specific tasks;

[0047] After the execution is completed, A continues to read the proposal and starts the next round of consensus;

[0048] If A loses contact with B and C, or A sends an unreasonable resolution, B and C will initiate a re-vote after their respective timeouts and try to become the new leader node. Since B's timeout is shorter than C's, B will initiate the vote first. At this time, the election process described above will be repeated, and B will eventually become the new leader.

[0049] Figure 3 This is the AI ​​tracking flow chart of the solution of the present invention. During the execution of the task, B and C identify the target person. After 2 seconds of confirmation, if the target continues to be in the field of view and the credibility given by the Yolo target recognition model continues to be greater than 85%, a true value discovery message for tracking the target is sent to A. If A collects more than 1 / 3 of the true value discovery messages for tracking the target within 30 seconds, it will read the proposal to track the target and reach a consensus with B and C. After the consensus is successful, the resolution is issued, and cluster tracking is started when this command is executed.

[0050] The selection box recognized by the Yolo target recognition model is passed as input to the SiamRPN target tracking model;

[0051] SiamRPN takes the box given by Yolo as input, and then continuously outputs the position of the target's latest selection box. The tracking modules A, B, and C continuously obtain the relative position of the target and the drone by parsing the position of the target selection box, and then drive the flight control program and gimbal camera to track the target.

[0052] The above description of the embodiments is to facilitate the understanding and application of the present invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments. Improvements and modifications made by those skilled in the art to the present invention based on the disclosure of the present invention should be within the protection scope of the present invention.

Claims

1. A decision-making method running on multiple drones, characterized in that: include Initiate a request to become the master node through one of the drones; Send the request to other drones, which generate affirmative votes through the election algorithm; The approval votes are sent to the drone. When the number of approval votes received by the drone exceeds half, the drone is determined to be the leader node. Proposals are generated in real time by all drones and sent to the leader node in real time; The leader node reads the pre-recorded or real-time generated proposal, encrypts it, obtains the encrypted proposal, and transmits it to other drones. The proposal content is the parameters of the command to be executed; Through the consensus process, other drones vote in favor of the encrypted proposal and send it to the leader node; When the number of votes received by the leader node exceeds half, the leader node generates a resolution based on the encrypted content of the proposal and sends it to other drones; The resolution content is analyzed by other drones to carry out specific tasks.

2. The decision-making method running on multiple drones according to claim 1, characterized in that: The request is sent to other drones, and other drones generate affirmative votes through the election algorithm as follows: The drone initiates a request to become the master node and attaches its proposal_id = promised_id + 1. All drones maintain two values, proposal_id and promised_id. After receiving the request, other drones will determine whether the proposal_id is greater than their promised_id. If so, they will return a positive vote and update their promised_id to the corresponding proposal_id. The drone that initiates the request will become the leader node if it receives more than half of the votes in favor.

3. The decision-making method running on multiple drones according to claim 1, characterized in that: The specific steps of other drones voting in favor of the encrypted proposal through the consensus process and sending it to the leader node are as follows: The leader node continuously takes out proposals to be agreed upon from its own consensus queue and broadcasts them to other drones; Other drones decide whether to return a positive vote for the message based on their own pre-set programs; When the leader receives more than half of the votes in favor, it generates a resolution message based on the proposal content and broadcasts it to all slave nodes.

4. The decision-making method running on multiple drones according to claim 1, characterized in that: The real-time generation of proposals and sending them to the leader node in real time is specifically as follows: All drone nodes participating in the consensus can generate proposals, and then package the proposals into a truth value discovery message and send it to the leader node; The leader node will continue to collect for up to 30 seconds after receiving the truth value discovery message for the first time. If it collects more than 1 / 3 of the total proposals, it will put the proposal corresponding to the truth value discovery into its own consensus queue.

Images