A multi-machine collaborative planning method under incompletely known conditions

Abstract

The present invention aims at the problem of how to cover as many targets to be supported as possible while ensuring the network connectivity between drones in the communication network coverage support mission, and how to achieve local optimal coverage through distributed negotiation based on the known local information of each drone when the network between drones is not fully connected. A multi-machine collaborative planning method under incompletely known conditions is proposed, which belongs to the field of multi-drone collaboration. Under global information conditions, this method transforms the coverage planning problem of large-scale non-uniform dynamic targets into a multi-objective combinatorial optimization problem; under local information conditions, the local strategy adjustment of dynamic targets is transformed into a distributed negotiation planning problem of drone nodes, which improves the representativeness and authenticity of typical mission scenarios and has stronger scenario adaptability.

Description

Technical Field

[0001] The present invention belongs to the field of multi-UAV collaboration, and in particular relates to a multi-UAV collaboration planning method under non-completely known conditions. Technical Background

[0002] With the increasing diversification of demands in economic and social development, national and national defense security, people are increasingly deepening their application of mobile communication networks in special scenarios, including communication support for large-scale gatherings, civil defense for major events, emergency response, and disaster relief in remote or dangerous areas. Traditional wireless communications mostly rely on pre-established basic network facilities, such as base stations or APs. This fixed, centralized network is limited by infrastructure, terminal location, etc., and has poor flexibility. It cannot flexibly deploy base stations in the best location according to application requirements, and it cannot change the network topology or optimize communication quality as user traffic changes. Drone swarms can quickly form regional coverage in a networked, distributed, and collaborative manner, optimize the scheduling of swarm resources, and improve task completion rate and response speed. The idea of ​​using drone swarms to carry out mobile communication coverage tasks has emerged. Compared with fixed base stations, drones have the characteristics of low cost, mobility, flexibility, ability to carry diverse payloads, and rapid deployment. They can flexibly change the node communication topology structure according to task requirements, and can quickly achieve network self-healing in the event of communication network paralysis and partial node failure. It can be foreseen that relying on drone swarms will bring disruptive capabilities to mobile communication dynamic coverage. Summary of the invention

[0003] In the communication network coverage support mission of UAV swarm, how to cover as many targets as possible while ensuring the network connectivity between UAVs, and how to achieve local optimal coverage through distributed negotiation based on the local information known by each UAV when the network connectivity between UAVs is incomplete. A multi-machine collaborative planning method under incompletely known conditions is proposed.

[0004] The technical solution adopted by the present invention is:

[0005] A multi-machine collaborative planning method under non-completely known conditions specifically includes the following steps:

[0006] Step 1: Each drone in the entire drone group broadcasts and shares its real-time location, and determines whether the global information is known or the local information is known based on the received drone location information; when the global information is known, the current location and connectivity status of the drone group is represented as an undirected graph, and the connectivity of the network is determined by calculating the zero eigenvalue multiplicity of the undirected graph Laplacian matrix; when the local information is known, the connectivity of the network is determined by node traversal;

[0007] Step 2: If the network is in a fully connected state, the target is covered by an undirected graph and the combined optimization method based on the minimum spanning tree is used to calculate the next position of the drone. If the network is not in a fully connected state, the distributed negotiation rule is used to determine the coverage mission target point position obtained by each drone, and clustering is performed to calculate the next position of the drone.

[0008] Step 3: Use the trajectory prediction method based on least squares curve fitting to predict the target situation change as the position of the target point of the next round of step 2 coverage mission.

[0009] Among them, in step 1, the connectivity of the network is determined by calculating the zero eigenvalue multiplicity of the undirected graph Laplacian matrix. The specific process is:

[0010] The current position and connectivity status of the drone swarm are represented as an undirected graph G = (V, E), where V is the drone node and E represents all communication connection edges between drone nodes. Using the Laplace matrix L = DA of the undirected graph G, where D is the degree matrix and A is the spatial adjacency matrix, the multiplicity of the 0 eigenvalue of L is further calculated. If the multiplicity is 1, it means that the network is fully connected, otherwise it is not fully connected.

[0011] In step 2, a distributed negotiation rule is used to determine the location of the coverage mission target point obtained by each drone, and clustering is performed to calculate the next location of the drone. The specific process is as follows:

[0012] The auction algorithm is used, which is divided into two stages: bidding and consensus. In the bidding stage, each drone evaluates the known number and location of covered targets and bids. In the consensus stage, each drone conducts multi-drone negotiation with the connected drones based on the bidding situation. The target tasks covered by each drone are determined according to the order of bids and the total number of covered targets. The next moving position of the drone is calculated by clustering the locations of the target points to be covered by each drone.

[0013] The rules for multi-machine negotiation during the consensus phase are as follows:

[0014] a) Connectivity maintenance rule: The drone maintains connectivity according to the coverage range R c When making decisions based on the changes in the internal targets, first determine whether the movement of the drone will destroy the connectivity of the network, that is, make the distance between the drone and all drones greater than R c ; If yes, the drone maintains the current state and requests support from surrounding neighbors;

[0015] b) Neighbor support rule: When the drone receives a request for help from a neighbor node, it will conduct a value assessment. If the number of target points covered after the movement is less than or equal to the current number of target points covered, the current state will remain unchanged, otherwise it will provide support;

[0016] c) Coverage movement rule: When a target in the current coverage area moves toward an uncovered area, the nearest node between the drone itself and all its neighbors and the target is calculated, and the nearest node responds to assist; if the nearest node is itself, the drone stops asking for help;

[0017] d) Target notification rule: When the target moves towards other drones, notify other drones to detect and adjust strategies.

[0018] The beneficial effects of the method of the present invention are:

[0019] Regardless of the multi-machine collaborative coverage scenario under global information or local information conditions, there are corresponding response strategies. Under global information conditions, the coverage planning problem of large-scale non-uniform dynamic targets is transformed into a multi-objective combinatorial optimization problem. Under local information conditions, the local strategy adjustment of dynamic targets is transformed into a distributed negotiation planning problem of drone nodes, which improves the representativeness and authenticity of typical mission scenarios and has stronger scenario adaptability.

[0020] Instruction Manual

[0021] Figure 1 Flowchart of the multi-machine collaborative planning method under non-completely known conditions of the present invention. DETAILED DESCRIPTION

[0022] Reference Figure 1 The present invention provides a multi-machine collaborative planning method under non-completely known conditions, which specifically includes the following steps:

[0023] Step 1: Connectivity judgment considering global / local information

[0024] 1.1 Judgment of known global / local information

[0025] Each drone in the entire drone group maintains a position matrix P{P 1 ,P 2 ,…,P N}, where P i represents the position of the drone numbered i, and N represents the number of drones in the drone group. Each drone is set to share its real-time position and number by broadcasting at a fixed time interval Δt. If each drone receives broadcast information from all drones except itself within the Δt time period, it will update the position matrix P{P 1 ,P 2 ,…,P N}All N position data including its own position, then it is determined that the global information is known. If one or several drones do not update the position matrix P{P 1 ,P2 ,…,P N}, if there is ≥1 position data within the local information, it is determined that the local information is known.

[0026] 1.2 When global information is known, the Laplace matrix method based on undirected graph is used to determine connectivity

[0027] The current positions of all drones are known to each other, and each drone knows the current positions of all other drones. The current positions and connectivity of the drone group are represented as an undirected graph G = (V, E), where V is a drone node and the distance between any two drones is less than the drone communication range R. c , there is a communication connection edge between the two drone nodes, and E represents all the communication connection edges between drone nodes. Using the Laplace matrix L = DA of graph G (D is the degree matrix, A is the spatial adjacency matrix), the zero eigenvalue multiplicity of L is further calculated. If the multiplicity is 1, it means that the network is fully connected, otherwise it is not fully connected.

[0028] 1.3 When only local information is known, the connectivity judgment is made based on the node broadcast information traversal method

[0029] That is, when not all drones know the current location of other drones, the system connectivity is determined by broadcasting node information. Specifically, any drone node is selected as the starting node, and the depth-first or breadth-first algorithm is used to traverse the nodes starting from this node. The traversal process is to broadcast a message to the surrounding area. The message contains a linked list in the form of (num,i)->(num,i). Among them, num is a counting flag with an initial value of 1; i is the drone number. All nodes connected to the broadcast initiating node can receive the message. The node that receives the message will increase its counting flag by 1 and continue to broadcast to its neighbors in a depth-first or breadth-first manner until no new message receiving node can be found. The algorithm stops. At this time, if num is equal to the total number of nodes, the network is judged to be fully connected, otherwise the network is judged to be not fully connected.

[0030] Step 2: Target coverage optimization based on network connectivity status

[0031] 2.1 Based on the output result of the network full connectivity status in step 2, if the network is in a full connectivity state, a combination optimization method based on the minimum spanning tree is used for coverage

[0032] Then obtain the communication topology graph G = (V, E) of the entire drone group, consider the edge relationships contained in the node degree matrix D and the spatial adjacency matrix A, and calculate all edge weights w of G = (V, E) ij =a ij ·d i ·d j , where a ijis the Euclidean distance between node i and node j, d i and d j Represents the node degree of node i and node j. The minimum spanning tree of graph G is constructed using Prim's algorithm, thereby generating the minimum necessary edge set E to maintain communication connectivity N ∈E, that is, the edge set E N All nodes in the graph G can be connected, and the sum of the weights of all edges can be minimized. With the basic constraint of maintaining this connection relationship, the next position of each drone is calculated based on the combined optimization method of connectivity and coverage (specific reference is "A combined optimization method for connectivity and coverage of drone swarms", ZL202110875181.2).

[0033] 2.2 Based on the output result of the full network connectivity status in step 2, if the network is not in a full connectivity state, a distributed negotiation rule is used to achieve local connectivity optimization coverage.

[0034] The CBAA auction algorithm is used, which is divided into two stages: bidding and consensus. In the bidding stage, each drone evaluates the number and location of known coverage targets and bids; in the consensus stage, each drone negotiates with the connected drones based on the bidding situation, and determines which drones get the coverage mission of which targets based on the order of bids and the total number of coverage targets. By clustering the locations of the target points to be covered by each drone, the next moving position of the drone is calculated.

[0035] The important rules during the consensus phase are established as follows:

[0036] a) Connectivity maintenance rule: The drone maintains connectivity according to the coverage range R c When making decisions based on the changes in the internal targets, first determine whether the movement of the drone will destroy the connectivity of the network, even if the distance between the drone and all drones is greater than R c If so, the drone maintains its current state and requests support from its neighbors.

[0037] b) Neighbor support rule: When a drone receives a request for help from a neighbor node, it first performs a value assessment. If the number of target points covered after movement is less than or equal to the current number of target points covered, the current state remains unchanged, otherwise support is provided.

[0038] c) Coverage movement rule: If a target in the current coverage area moves toward an uncovered area, the nearest node to the target is calculated for the drone itself and all its neighbors, and the nearest node responds to assist; if the nearest node is itself, the drone stops asking for help.

[0039] d) Target notification rule: When the target moves towards other drones, notify other drones in time to conduct detection and strategy adjustment.

[0040] Step 3: Follow-up planning based on trajectory prediction of covered moving targets

[0041] The moving target shows diverse, irregular and uncertain changes during the maneuvering process, and it is difficult to establish a suitable model equation to describe the trajectory curve. Considering that the target position is gradually changing, the target position in the short term in the future is only related to the current position and the most recent position point, but not to the earlier position. The trajectory prediction algorithm based on least squares curve fitting is used to predict the target situation change, and the array {(x i ,y i ),i=1,2,...,M}, where M is the number of historical trajectory points considered, (x i ,y i ) is the position of the i-th historical trajectory point. The array is described by a curve model as: y(x) = f(a, x) = a 0 +a 1 x+a 2 x 2 +...+a n x n (n<M-1), where n is the order of the curve model, and the parameter a is selected based on the least squares method so that the residual δ between the fitted model and the actual observation value at each point i =f(x i )-y i The weighted sum of squares reaches the minimum, that is:

[0042]

[0043] To find the extreme value of the above multivariate function, the function has a partial derivative at a certain point and the partial derivative is zero, that is:

[0044]

[0045] Its matrix form is:

[0046]

[0047] The above equation can be used to obtain the unique solution of the coefficient matrix. Substituting the coefficient matrix into the model can obtain the optimal solution for curve fitting. The above method is used to perform curve fitting on each historical trajectory point of the covered target, and the curve fitting equation of all targets can be constructed.

[0048] The above is only a specific example of the present invention to facilitate the understanding of the present invention by those skilled in the art, but the present invention is not limited to the scope of the specific example. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected by the present invention.

Claims

1. A multi-machine collaborative planning method under non-completely known conditions, characterized in that: The specific steps include: Step 1: Each drone in the entire drone group broadcasts and shares its real-time location, and determines whether the global information is known or the local information is known based on the received drone location information; when the global information is known, the current location and connectivity status of the drone group is represented as an undirected graph, and the connectivity of the network is determined by calculating the zero eigenvalue multiplicity of the undirected graph Laplacian matrix; when the local information is known, the connectivity of the network is determined by node traversal; Step 2: If the network is in a fully connected state, the target is covered by an undirected graph and the combined optimization method based on the minimum spanning tree is used to calculate the next position of the drone. If the network is not in a fully connected state, the distributed negotiation rule is used to determine the coverage mission target point position obtained by each drone, and clustering is performed to calculate the next position of the drone. Step 3: Use the trajectory prediction method based on least squares curve fitting to predict the target situation change as the position of the target point of the next round of step 2 coverage mission.

2. The method for multi-machine collaborative planning under incompletely known conditions according to claim 1, characterized in that: In step 1, the connectivity of the network is determined by calculating the zero eigenvalue multiplicity of the undirected graph Laplacian matrix. The specific process is as follows: The current position and connectivity status of the drone swarm are represented as an undirected graph G = (V, E), where V is the drone node and E represents all communication connection edges between drone nodes. Using the Laplace matrix L = DA of the undirected graph G, where D is the degree matrix and A is the spatial adjacency matrix, the multiplicity of the 0 eigenvalue of L is further calculated. If the multiplicity is 1, it means that the network is fully connected, otherwise it is not fully connected.

3. The method for multi-machine collaborative planning under incompletely known conditions according to claim 1, characterized in that: In step 2, the distributed negotiation rule is used to determine the coverage mission target point location obtained by each drone, and clustering is performed to calculate the next location of the drone. The specific process is as follows: The auction algorithm is used, which is divided into two stages: bidding and consensus. In the bidding stage, each drone evaluates the number and location of known coverage targets and bids. In the consensus stage, each drone conducts multi-drone negotiation with the connected drones based on the bidding situation. The target tasks covered by each drone are determined according to the order of bids and the total number of covered targets. The next moving position of the drone is calculated by clustering the positions of the target points to be covered by each drone.

4. The method for multi-machine collaborative planning under incompletely known conditions according to claim 3, characterized in that: The rules for multi-machine negotiation during the consensus phase are as follows: a) Connectivity maintenance rule: The drone maintains connectivity according to the coverage range R c When making decisions based on the changes in the internal targets, first determine whether the movement of the drone will destroy the connectivity of the network, that is, make the distance between the drone and all drones greater than R c ; If yes, the drone maintains the current state and requests support from surrounding neighbors; b) Neighbor support rule: When the drone receives a request for help from a neighbor node, it will conduct a value assessment. If the number of target points covered after the movement is less than or equal to the current number of target points covered, the current state will remain unchanged, otherwise it will provide support; c) Coverage movement rule: When a target in the current coverage area moves toward an uncovered area, the nearest node to the target is calculated, including the distance between the drone itself and all its neighbors, and the nearest node responds to assist; If the nearest node is itself, stop asking for help; d) Target notification rule: When the target moves towards other drones, notify other drones to detect and adjust strategies.

Images