A Communication Method and System for UAV Swarms in Complex Environments

Abstract

The present invention provides a communication method and system for an unmanned aerial vehicle (UAV) cluster in a complex environment. The communication method includes: traversing all node pairs in the UAV cluster, filling the connection relationship between nodes into an adjacency matrix, and establishing an adjacency matrix of the UAV cluster; using an artificial potential field obstacle avoidance function and a connectivity maintenance function to simultaneously constrain the UAV cluster; defining the nodes with a value of 0 in the UAV cluster adjacency matrix as outlier UAVs, and using a distance vector algorithm to plan the return path of the outlier UAVs until the outlier UAVs return to the UAV cluster; repeating the above steps until the UAV cluster reaches a set target point. The present invention enhances the communication ability between UAVs in the cluster while realizing automatic obstacle avoidance of the cluster. The outlier UAVs in the cluster can quickly re-establish a connection with the cluster within a short time, further enhancing the communication ability between cluster UAVs in various complex environments.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle communication technology, and in particular to a communication method and system for a cluster of unmanned aerial vehicles in a complex environment. Background Art

[0002] A drone swarm is a collection of multiple drones that are under unified command to perform a certain task together. It is interconnected through network technology to form a centralized and unified whole, thereby achieving collaborative operations or completing other specific tasks.

[0003] In the prior art, drone cluster communication technology mainly includes reinforcement learning, network topology control methods and distributed control strategies. For example, in controller design, the local perception range and intermittent link failure of the intelligent agent are taken into consideration, and the use of non-smooth navigation functions combined with a common Lyapunov function helps the drone cluster maintain communication. Another example is a distributed group algorithm, in which intelligent agents transmit information through a proximity graph, use artificial potential field functions to simulate natural attraction and repulsion, and consider dynamic network topology to ensure the continuity of communication connections when the intelligent agent moves.

[0004] The above methods play an important role in enhancing the communication of drones in a cluster, but their effects are greatly weakened in complex environments, such as narrow spaces with obstacles and environments with multiple obstacles. This is because when a drone cluster avoids obstacles in a complex environment, the topology of the cluster changes, causing the drone bodies that were originally connected to be disconnected, and they cannot quickly establish new connections in a short period of time, thus leaving the entire cluster. In addition, the existing drone cluster communication technology shows insufficient consideration of large-scale cluster scenarios, and often focuses on the communication mechanism and strategy of small-scale drone formations. Most of its experimental designs are limited to an extremely limited number of drones, generally not exceeding ten individuals, which leads to limitations when a large number of drones need to work together. Summary of the invention

[0005] Purpose of the invention: The first purpose of the present invention is to provide a communication method for a drone cluster in a complex environment that realizes the functions of automatic obstacle avoidance and automatic return of stray drones to the cluster. The second purpose is to provide a communication system for a drone cluster in a complex environment.

[0006] Technical solution: A communication method for a drone cluster in a complex environment, comprising the following steps:

[0007] (1) Traverse all node pairs in the drone cluster and fill the connection relationship between nodes into the adjacency matrix to establish the drone cluster adjacency matrix;

[0008] (2) Using artificial potential field obstacle avoidance function and connectivity maintenance function to constrain the UAV cluster at the same time;

[0009] (3) The nodes with a value of 0 in the adjacency matrix of the drone cluster are defined as outlier drones, and the distance vector algorithm is used to plan the return path of the outlier drone until the outlier drone returns to the drone cluster;

[0010] (4) Repeat steps (1) to (3) until the drone cluster reaches the set target point.

[0011] Specifically, step (1) includes the following sub-steps:

[0012] (1.1) Calculate the Euclidean distance between each pair of nodes in the drone cluster;

[0013] (1.2) Compare the Euclidean distance between a pair of nodes with the perception radius of the drone. If the Euclidean distance is less than or equal to the perception radius, there is a connection between the pair of nodes, and the value of the corresponding position in the adjacency matrix is ​​set to 1. If the Euclidean distance is greater than the perception radius, there is no connection between the pair of nodes, and the value of the corresponding position in the adjacency matrix is ​​set to 0.

[0014] (1.3) Repeat step (1.2) to fill the connection relationships between all nodes into the adjacency matrix.

[0015] Specifically, in step (2), the formula of the artificial potential field obstacle avoidance function is as follows:

[0016] ,

[0017] Where: is the artificial potential field obstacle avoidance function, is the repulsion function, is the gravitational function, k is the set of other drones and obstacles within the sensing range of drone i, g is the target point of the drone cluster movement, is the distance between drone i and obstacles or other drones within its sensing range, is the distance between UAV i and the target point of the UAV cluster movement, is the repulsion constant, is the gravitational constant, The sensing range of obstacles or other drones is set.

[0018] Specifically, in step (2), the connectivity maintenance function formula is as follows:

[0019] ,

[0020] Where: is the connectivity maintenance function, is the distance between UAV i and its neighboring UAV j, is the sensing radius of the drone, It is the buffer distance used to maintain network connectivity.

[0021] Specifically, step (3) includes the following sub-steps:

[0022] (3.1) Traverse the adjacent nodes of the outlier drones and construct the outlier drone adjacency matrix. The initial node value of the outlier drone adjacency matrix is ​​0;

[0023] (3.2) Establish a routing table for each drone in the drone cluster. The routing table records the physical distance between each drone and its neighboring drones. Randomly select the drone node with a value of 1 in the drone cluster adjacency matrix and set it as the target drone node.

[0024] (3.3) Select a neighboring drone node from the outlier drone’s routing table as the starting point, and calculate the physical distance from the outlier drone node to other drone nodes through the neighboring drone node, and traverse all the neighboring drone nodes in the outlier drone’s routing table until the shortest path from the outlier drone node to the target drone node is found;

[0025] (3.4) Sending navigation instructions to the outlier drone based on the shortest path from the outlier drone node to the target drone node;

[0026] (3.5) Update the adjacency matrix of the outlier drone in real time. If the node value of the adjacency matrix of the outlier drone is 1, the outlier drone has returned to the drone cluster. If the node value of the adjacency matrix of the outlier drone is 0, return to step (3.3) until the node value of the adjacency matrix of the outlier drone is 1.

[0027] Specifically, the routing table contains the following fields: node ID, which is used to uniquely identify each drone in the drone cluster; neighbor node ID, which is used to record the IDs of all neighboring drones that can directly communicate with a drone; distance set, which is used to record the physical distance between a drone and all of its neighboring drones.

[0028] Specifically, in step (3.3), traversing all adjacent drone nodes in the routing table of the outlier drone specifically includes: traversing the adjacent drone nodes, and calculating the physical distance from each adjacent drone node to the target drone node, and comparing the obtained physical distance with the physical distance calculated last time after each calculation. If the physical distance from a certain adjacent drone node to the target drone node is smaller than the physical distance calculated last time, then the shortest physical distance to the target drone node is updated, and the adjacent drone node corresponding to the shortest physical distance is recorded as the optimal adjacent drone node. After traversing all adjacent drone nodes, the shortest path from the outlier drone node to the target drone node is obtained.

[0029] Preferably, the above communication method further comprises the following steps:

[0030] (5) Calculate and record the average degree of the network of the drone cluster adjacency matrix in real time, compare the calculated average degree of the network with the set threshold, and evaluate whether it meets the set requirements.

[0031] The average degree of the network is calculated as follows:

[0032] ,

[0033] Where: is the average degree of the network, is the total degree of nodes in the network, is the adjacency matrix of the drone cluster, is the total number of nodes in the network.

[0034] The present invention also provides a communication system for a drone cluster in a complex environment, comprising:

[0035] Adjacency matrix construction module: used to traverse all node pairs in the drone cluster and fill the connection relationship between nodes into the adjacency matrix to establish the drone cluster adjacency matrix;

[0036] Cluster constraint module: used to constrain the drone cluster simultaneously using the artificial potential field obstacle avoidance function and connectivity maintenance function;

[0037] Outlier drone path planning module: It is used to define the nodes with a value of 0 in the drone cluster adjacency matrix as outlier drones, and use the distance vector algorithm to plan the return path of the outlier drone until the outlier drone returns to the drone cluster;

[0038] Iteration module: used to repeatedly execute the adjacency matrix construction module, cluster constraint module and outlier drone path planning module until the drone cluster reaches the set target point.

[0039] Beneficial effects: Compared with the prior art, the significant effect of the present invention is that the present invention constrains the drone cluster through the artificial potential field obstacle avoidance function and the connectivity maintenance function, and enhances the communication capability between drones in the cluster while realizing automatic obstacle avoidance of the cluster. At the same time, by constructing the adjacency matrix of the drone cluster and applying the distance vector algorithm based on graph theory, the out-of-group drone can quickly reconnect with the cluster in a short time, further enhancing the communication capability between cluster drones in a variety of complex environments. The present invention is suitable for automatic obstacle avoidance and automatic pathfinding of a large number (more than 10) of drone clusters in a variety of complex environments, and has a wider range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart of the method of the present invention.

[0041] Figure 2 It is a plane simulation diagram of the first half of the obstacle avoidance process of a drone cluster passing through multiple obstacles in Example 1 of the present invention.

[0042] Figure 3 It is a plane simulation diagram of the second half of the obstacle avoidance process of the drone cluster passing through multiple obstacles in Example 1 of the present invention.

[0043] Figure 4 It is a plane simulation diagram of the first half of the obstacle avoidance process of the drone cluster passing through a narrow space in Example 1 of the present invention.

[0044] Figure 5 It is a plane simulation diagram of the second half of the obstacle avoidance process of the drone cluster passing through a narrow space in Example 1 of the present invention. DETAILED DESCRIPTION

[0045] A preferred embodiment of the present invention is further described below in conjunction with the accompanying drawings.

[0046] Example 1

[0047] See also Figure 1 As shown, this embodiment provides a communication method for a drone cluster in a complex environment, comprising the following steps:

[0048] (1) Traverse all the node pairs in the drone cluster, calculate the distance between each pair of nodes, and compare the distance between each pair of nodes with the perception radius of the drone to obtain the connection relationship between the node pairs. Fill the connection relationship information into a two-dimensional adjacency matrix of size N×N to establish the drone cluster adjacency matrix A, which includes the following sub-steps:

[0049] (1.1) For each pair of nodes in the drone cluster , calculate the Euclidean distance between them , obtained by taking the modulus of the difference between the position vectors of two nodes i and j:

[0050] ,

[0051] Where: It is an N×M matrix that stores the spatial positions of all nodes, where N is the total number of nodes in the network and M is the spatial dimension.

[0052] (1.2) The calculated pair of nodes Euclidean distance Perception radius For comparison, if the Euclidean distance Less than or equal to the perception radius , there is a connection relationship between this pair of nodes, and the value of the corresponding position in the adjacency matrix is ​​set to 1. If the Euclidean distance Larger than the perception radius , there is no connection relationship between this pair of nodes, and the value of the corresponding position in the adjacency matrix is ​​set to 0.

[0053] (1.3) Repeat step (1.2) to fill the connection relationship between all nodes into the adjacency matrix. At this point, the adjacency matrix fully reflects the connection status between the nodes in the drone cluster. By checking the elements in the adjacency matrix, it is possible to quickly understand whether there is a direct connection relationship between any two nodes.

[0054] (2) In order to achieve obstacle avoidance and maintain the connectivity of the internal communication of the UAV cluster, a constraint function is designed. , the constraint function comprehensively considers the relative position relationship between any UAV and its neighboring UAVs and surrounding obstacles, and uses the artificial potential field obstacle avoidance function and connectivity maintenance function to constrain the UAV cluster at the same time, which specifically includes the following sub-steps:

[0055] (2.1) Define the communication and perception range of the drone cluster: Define the set of neighboring drones of drone i as , all drones in this set can communicate with drone i, and the set of all drones within the sensing range is , the set of obstacles in the sensing area is , and The union of ,The above sets all change dynamically with the movement of the drone cluster.

[0056] (2.2) Calculation of the artificial potential field obstacle avoidance function , this function is constructed based on artificial potential field theory and is used to simulate the repulsive force between drone i and obstacles or other drones k, and the attractive force between drone i and the target point g of the drone cluster movement. The function is designed as follows:

[0057] ,

[0058] Where: is the artificial potential field obstacle avoidance function, is the repulsion function, is the gravitational function, k is the set of other drones and obstacles within the sensing range of drone i, g is the target point of the drone cluster movement, The obstacle or other drones within the drone’s sensing range The distance between is the distance between UAV i and the target point g of the UAV cluster movement, is the repulsion constant, is the gravitational constant, The sensing range of obstacles or other drones is set.

[0059] (2.3) Calculation of connectivity maintenance function ,The connectivity maintenance function is used to ensure that the distance between the UAV and its neighboring UAVs remains within the communication range. The design of this function is as follows:

[0060] ,

[0061] Where: is the connectivity maintenance function, is the distance between UAV i and its neighboring UAV j, is the sensing radius of the drone, The buffer distance is used to maintain network connectivity. When the distance between a drone and its neighboring drones is close to the perception radius, the buffer distance can ensure that the drone does not leave the perception range, thereby avoiding network disconnection.

[0062] when hour, , indicating that the distance between UAV i and its neighboring UAV j is within the safe range. hour, The value of This will generate a control force that makes UAV i move towards UAV j to avoid leaving the sensing range. hour, , indicating that UAV i has left the sensing range of UAV j, and the distance vector algorithm will be used to restore connectivity.

[0063] (2.4) The artificial potential field obstacle avoidance function and connectivity maintenance functions Fusion Constraint Function :

[0064] ,

[0065] Where: is the overall constraint function of the UAV, that is, the artificial potential field obstacle avoidance function is applied to the UAV at the same time and connectivity maintenance functions To constrain.

[0066] (3) In the context of autonomous navigation and formation control of drone clusters, when a drone is out of the direct perception and communication range of its formation members or other drones in the communication network, in order to ensure the integrity of the formation and the continuity of task execution, a strategy needs to be adopted to enable the drone to re-establish connection or guide it to return to the group safely. In the present invention, the nodes assigned a value of 0 in the drone cluster adjacency matrix A are defined as outlier drones, and the distance vector algorithm (DVA) is used to plan the return path of the outlier drone. DVA is a dynamic routing selection algorithm that dynamically updates the routing table by continuously exchanging distance information between each node in the network until the outlier drone returns to the drone cluster, including the following sub-steps:

[0067] (3.1) Based on the same method as step (1), traverse the adjacent nodes of the outlier drones and construct the outlier drone adjacency matrix B. The adjacency matrix B is a 1×1 two-dimensional adjacency matrix. The initial node value of the outlier drone adjacency matrix B is 0.

[0068] (3.2) A routing table is established for each drone in the drone cluster. The physical distance between each drone and its adjacent drones is recorded in the routing table. A drone node with a value of 1 in the drone cluster adjacency matrix A is randomly selected and set as the target drone node. Different from the traditional routing table which records the number of hops and network address of information transmission, the routing table in the present invention is mainly used to record the physical distance between the outlier drone and the adjacent drone.

[0069] The above routing table contains the following fields:

[0070] Node ID, used to uniquely identify each drone in the drone cluster;

[0071] Neighbor node ID, used to record the IDs of all neighboring drones that can directly communicate with a drone;

[0072] The distance collection is used to record the physical distance between a drone and all its neighboring drones.

[0073] (3.3) In order to effectively plan the path for the stray drone to return to the team, it is first necessary to access and parse its routing table in detail, select an adjacent drone node from the routing table of the stray drone as the starting point, and calculate the physical distance from the stray drone node to other drone nodes through the neighboring drone node. The physical distance element contained in the distance set in the routing table is used in the physical distance calculation. All adjacent drone nodes in the routing table of the stray drone are traversed until the shortest path from the stray drone node to the target drone node is found. Specifically, the adjacent drone nodes of the stray drone are traversed, and the physical distance from each adjacent drone node to the target drone node is calculated. After each calculation, the obtained physical distance is compared with the physical distance calculated last time. If the physical distance from a certain adjacent drone node to the target drone node is smaller than the physical distance calculated last time, the shortest physical distance to the target drone node is updated, and the adjacent drone node corresponding to the shortest physical distance is recorded as the optimal adjacent drone node. After traversing all adjacent drone nodes, the shortest path from the stray drone node to the target drone node is obtained.

[0074] (3.4) Based on the shortest path from the outlier drone node to the target drone node, a navigation command is sent to the outlier drone to guide the outlier drone to move along the shortest path to the target drone node. After receiving the navigation command, the outlier drone will adjust its flight attitude and power output and perform the reunion action. During the whole process, the state and position of the outlier drone are continuously monitored, and the above steps are repeatedly used to adjust the navigation command and optimize the reunion strategy.

[0075] (3.5) Update the adjacency matrix B of the outlier drone in real time. If the node value of the adjacency matrix B of the outlier drone is 1, the outlier drone has returned to the drone cluster. If the node value of the adjacency matrix B of the outlier drone is 0, return to step (3.3) until the node value of the adjacency matrix of the outlier drone is 1.

[0076] (4) Repeat the above steps (1) to (3) until the drone cluster reaches the set target point.

[0077] (5) In order to evaluate the communication capability of the drone cluster, after each update of the drone cluster adjacency matrix, the average degree of the network of the drone cluster adjacency matrix is ​​calculated and recorded, and the calculated average degree of the network is compared with the set threshold to evaluate whether it meets the set requirements.

[0078] The average degree of the network is calculated as follows:

[0079] ,

[0080] Where: is the average degree of the network, is the total degree of nodes in the network, is the adjacency matrix of the drone cluster, is the total number of nodes in the network.

[0081] If the average degree of the network is high, it means that the connection between nodes is relatively close, and the network may have good connectivity and information transmission efficiency. If the average degree of the network is low, it may mean that there are more isolated nodes in the network or the connection between nodes is relatively sparse, which may affect the connectivity and information transmission efficiency of the network. However, too high a network average degree may cause network congestion and performance degradation, while too low a network average degree may affect the reliability and fault tolerance of the network. Therefore, the threshold is set to a moderate range based on multiple experiments.

[0082] The following tests are performed using the solution of the present invention in two typical complex environments with multiple obstacles.

[0083] Please refer to Figures 2 to 3 As shown in the figure, the obstacle avoidance process of a drone cluster passing through multiple obstacles. The small red dots represent drones, and the large red dots represent obstacles. If there is a communication connection between two drones, a blue line is used to connect them. In this embodiment, there are 80 drones in total, the step size is set to 0.009, and the result is output every 30 steps.

[0084] The obstacle avoidance process is described below. Figure 2 Sub-graph a in the figure is the initial state, where 80 drones are randomly distributed within the area; Figure 2 Sub-figure b in the figure shows the state where the drone cluster begins to approach an obstacle but has not yet started to avoid the obstacle and has communication connection with each other; Figure 2 Sub-graph c in the figure shows the drone cluster starting to pass through a multi-obstacle environment. It can be seen from the figure that most of the drones are still connected, and there are two outliers. Figure 2 Sub-image d in the figure shows a drone cluster passing through a multi-obstacle environment, in which 7 drones have left the cluster. Figure 3 In sub-image a, the drone cluster is passing through a multi-obstacle environment. Figure 2 In sub-graph d, some of the outlier drones have returned to the cluster and established communication connections. There are still 7 outlier drones, but 4 of them have established communication connections with each other. Figure 3 In sub-image b, some drones have crossed the obstacle environment and reached the target area, and the number of outlier drones has been reduced to 3; Figure 3 In sub-image c, most of the drones have crossed the obstacle environment and reached the target area, while a few drones are still in the obstacle environment, and there are 8 outliers. Figure 3In sub-graph d in , all drones successfully cross the obstacle environment, and there are no outlier drones in the drone cluster.

[0085] Please refer to Figures 4 to 5 As shown in the figure, the obstacle avoidance process of a drone cluster passing through a narrow space is shown in the figure. Figures 4 to 5 In the figure, red dots are used to represent drones, and yellow ovals are used to represent obstacles. Obstacles form a narrow space in space. Figure 4 In the figure, a green dot represents the target point. If there is a communication connection between two drones, a blue line is used to connect them. In this embodiment, a total of 50 drones are used for traversing a narrow space simulation. The step size is set to 0.009, and the result is output every 30 steps. Representative output results are selected for illustration.

[0086] Figure 4 Sub-graph a in the figure is the initial state, where 50 drones are randomly distributed within the area; Figure 4 Sub-image b in the figure shows the state where the drone cluster begins to approach the narrow space formed by obstacles, but has not yet started to avoid obstacles and has communication connections with each other; Figure 4 Sub-graph c in the figure shows the drone cluster starting to pass through a narrow space. It can be seen from the figure that there is communication connection between all drone clusters, and no drone is out of the group; Figure 4 Sub-image d in the figure shows that a drone cluster is passing through a narrow space, and some drones have reached the target point and are waiting at the target point. Figure 5 In sub-image a, a drone cluster is passing through a narrow space. Three drones leave the cluster and appear below the obstacle, but they establish communication connections with each other. One drone leaves the cluster and appears above the obstacle. Figure 5 In sub-image b, a drone cluster is passing through a narrow space. Figure 5 In the sub-figure a, the three drones that left the cluster at the bottom have returned to the cluster and established communication connections, but there is still one drone that has left the cluster; Figure 5 In the subgraph c in Figure 5 In sub-figure b, one drone that has left the cluster has returned to the cluster and established a communication connection, and most of the drones have reached the target point; Figure 5 In the sub-graph d in , all drones pass through the narrow space to reach the target point, and there is no outlier drone in the drone cluster. The whole process of this embodiment has a total of 270 steps and takes a total of 2.43 seconds.

[0087] Through the above simulation experiments, it is proved that the communication method provided by the present invention can enable the drone cluster to successfully cross a complex environment with multiple obstacles, and can maintain communication connection before and after crossing. Under specific terrain conditions such as narrow passages, the dense distribution and irregular arrangement of obstacles not only limit the flight path selection of drones, but also greatly increase the possibility of signal attenuation and interference. Therefore, it is difficult for traditional algorithms to effectively deal with the deterioration of the communication environment caused by obstacles, and it is impossible to ensure that the multi-drone cluster can still maintain reliable communication connection when crossing narrow passages; the communication method provided by the present invention uses a distance vector algorithm and a real-time obstacle avoidance mechanism to ensure that the drone cluster can quickly cross the narrow space without physical contact with any obstacles and maintaining communication.

[0088] Example 2

[0089] This embodiment provides a communication system for a drone cluster in a complex environment corresponding to the communication method described in Embodiment 1, including:

[0090] Adjacency matrix construction module: used to traverse all node pairs in the drone cluster and fill the connection relationship between nodes into the adjacency matrix to establish the drone cluster adjacency matrix;

[0091] Cluster constraint module: used to constrain the drone cluster simultaneously using the artificial potential field obstacle avoidance function and connectivity maintenance function;

[0092] Outlier drone path planning module: It is used to define the nodes with a value of 0 in the drone cluster adjacency matrix as outlier drones, and use the distance vector algorithm to plan the return path of the outlier drone until the outlier drone returns to the drone cluster;

[0093] Iteration module: used to repeatedly execute the adjacency matrix construction module, cluster constraint module and outlier drone path planning module until the drone cluster reaches the set target point.

Claims

1. A communication method for a drone cluster in a complex environment, characterized in that: The following steps are involved: S1, traverse all node pairs in the drone cluster, fill the connection relationship between the nodes into the adjacency matrix, and establish the drone cluster adjacency matrix; S2, using the artificial potential field obstacle avoidance function and connectivity maintenance function to constrain the UAV cluster at the same time; The formula of the artificial potential field obstacle avoidance function is as follows: ， Where: is the artificial potential field obstacle avoidance function, is the repulsion function, is the gravitational function, k is the set of other drones and obstacles within the sensing range of drone i, g is the target point of the drone cluster movement, is the distance between drone i and obstacles or other drones within its sensing range, is the distance between UAV i and the target point of the UAV cluster movement, is the repulsion constant, is the gravitational constant, To set obstacles or other drone sensing range; The connectivity maintenance function formula is as follows: ， Where: is the connectivity maintenance function, is the distance between UAV i and its neighboring UAV j, is the sensing radius of the drone, It is the buffer distance used to maintain network connectivity; S3, defining the nodes with a value of 0 in the drone cluster adjacency matrix as outlier drones, and using the distance vector algorithm to plan the return path of the outlier drone until the outlier drone returns to the drone cluster; S4. Repeat steps S1-S3 until the drone cluster reaches the set target point.

2. The communication method according to claim 1, characterized in that: The step S1 comprises the following sub-steps: S11, calculating the Euclidean distance between each pair of nodes in the drone cluster; S12, comparing the Euclidean distance between a pair of nodes with a preset interaction radius, if the Euclidean distance is less than or equal to the preset interaction radius, there is a connection relationship between the pair of nodes, and the value of the corresponding position in the adjacency matrix is ​​set to 1; if the Euclidean distance is greater than the preset interaction radius, there is no connection relationship between the pair of nodes, and the value of the corresponding position in the adjacency matrix is ​​set to 0; S13. Repeat step S12 to fill the connection relationship between all nodes into the adjacency matrix.

3. The communication method according to claim 1, characterized in that: The step S3 comprises the following sub-steps: S31, traverse the adjacent nodes of the outlier drones, and construct an adjacency matrix of the outlier drones, where the initial node value of the adjacency matrix of the outlier drones is 0; S32, establishing a routing table for each drone in the drone cluster, recording the physical distance between each drone and its adjacent drones in the routing table, and randomly selecting a drone node with a value of 1 in the drone cluster adjacency matrix and setting it as the target drone node; S33, selecting an adjacent drone node from the routing table of the outlier drone as a starting point, and calculating the physical distance from the outlier drone node to other drone nodes through the adjacent drone node, traversing all adjacent drone nodes in the routing table of the outlier drone until the shortest path from the outlier drone node to the target drone node is found; S34, sending navigation instructions to the outlier drone based on the shortest path from the outlier drone node to the target drone node; S35. Update the adjacency matrix of the outlier drone in real time. If the node value of the adjacency matrix of the outlier drone is 1, the outlier drone has returned to the drone cluster. If the node value of the adjacency matrix of the outlier drone is 0, return to step S33 until the node value of the adjacency matrix of the outlier drone is 1.

4. The communication method according to claim 3, characterized in that: The routing table contains the following fields: Node ID, used to uniquely identify each drone in the drone cluster; The neighbor node ID is used to record the IDs of all neighboring drones that can directly communicate with a drone. The distance set is used to record the physical distance between a drone and all its neighboring drones.

5. The communication method according to claim 3, characterized in that: In the step S33, traversing all adjacent drone nodes in the routing table of the outlier drone specifically includes: traversing adjacent drone nodes, and calculating the physical distance from each adjacent drone node to the target drone node, and comparing the obtained physical distance with the physical distance calculated last time after each calculation. If the physical distance from a certain adjacent drone node to the target drone node is smaller than the physical distance calculated last time, then the shortest physical distance to the target drone node is updated, and the adjacent drone node corresponding to the shortest physical distance is recorded as the optimal adjacent drone node. After traversing all adjacent drone nodes, the shortest path from the outlier drone node to the target drone node is obtained.

6. The communication method according to claim 1, characterized in that: The following steps are also included: S5. Calculate and record the average degree of the network of the drone cluster adjacency matrix in real time, compare the calculated average degree of the network with the set threshold, and evaluate whether it meets the set requirements.

7. The communication method according to claim 6, characterized in that: In step S5, the average degree of the network is calculated as follows: ， Where: is the average degree of the network, is the total degree of nodes in the network, is the adjacency matrix of the drone cluster, is the total number of nodes in the network.

8. A communication system for drone swarms in complex environments, characterized in that: include: Adjacency matrix construction module: used to traverse all node pairs in the drone cluster and fill the connection relationship between nodes into the adjacency matrix to establish the drone cluster adjacency matrix; Cluster constraint module: used to constrain the drone cluster simultaneously using the artificial potential field obstacle avoidance function and connectivity maintenance function; The formula of the artificial potential field obstacle avoidance function is as follows: ， Where: is the artificial potential field obstacle avoidance function, is the repulsion function, is the gravitational function, k is the set of other drones and obstacles within the sensing range of drone i, g is the target point of the drone cluster movement, is the distance between drone i and obstacles or other drones within its sensing range, is the distance between UAV i and the target point of the UAV cluster movement, is the repulsion constant, is the gravitational constant, To set obstacles or other drone sensing range; The connectivity maintenance function formula is as follows: ， Where: is the connectivity maintenance function, is the distance between UAV i and its neighboring UAV j, is the sensing radius of the drone, It is the buffer distance used to maintain network connectivity; Outlier drone path planning module: It is used to define the nodes with a value of 0 in the drone cluster adjacency matrix as outlier drones, and use the distance vector algorithm to plan the return path of the outlier drone until the outlier drone returns to the drone cluster; Iteration module: used to repeatedly execute the adjacency matrix construction module, cluster constraint module and outlier drone path planning module until the drone cluster reaches the set target point.

Images