A relay network topology generation method for drone swarm applications

Abstract

The present invention discloses a relay network topology generation method for drone cluster applications. The method first initializes the size of the multi-machine collaborative task area, the initial positions of all relay machines, the initial positions of all task machines, the communication range, the safety distance between drones, the business volume, and the business volume storage capacity in the multi-machine collaborative area. Then, based on the PID control algorithm and the load balancing adaptive adjustment mechanism, the load of the relay machine is dynamically adjusted. The business volume attraction is achieved by calculating the attraction of the upstream task machine to the relay machine. The Kalman filter algorithm is used to predict the position of the task machine at the next moment, and a prediction force is introduced to achieve the pulling of the relay machine. By comprehensively considering the business volume attraction, prediction force, anti-collision force between drones, and obstacle repulsion force, a complete virtual force algorithm system is constructed to ensure that the relay machine effectively covers the task machine and maintains communication with the central control, thereby improving the performance and reliability of the drone group when performing complex tasks.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communication technology, and in particular relates to a relay network topology generation method for unmanned aerial vehicle (UAV) cluster applications. Background Art

[0002] With the rapid development of unmanned swarms, the collaboration and topology construction between them are playing an increasingly important role in fields such as smart agriculture, disaster relief, and logistics. Generating an effective network topology that adapts to multi-service transmission is one of the key technologies to ensure the efficient operation of unmanned swarm networks. For example, in logistics and transportation scenarios, unmanned swarms play a vital role in logistics distribution, especially in last-mile delivery. Through collaborative work, they improve delivery efficiency and service quality. In modern precision agriculture, unmanned swarms are widely used for crop health monitoring, soil analysis, and pest control. This requires drone networks to cover vast farmlands and be capable of processing and transmitting large amounts of data. Unmanned swarms can monitor soil moisture, pest and disease conditions, and crop growth in real time, providing timely data support to help farmers make informed decisions.

[0003] Network topology affects the communication efficiency and stability between unmanned swarms and is crucial for successful mission execution. Unmanned swarms require an efficient network topology to support communication and coordination. Virtual force algorithms are a crucial tool in this regard. They optimize network layout by simulating the attractive and repulsive forces between drones, thereby improving the collaborative efficiency and overall performance of the swarm. However, traditional virtual force methods face several key challenges when executing missions in large, highly dynamic, multi-drone collaborative scenarios with large numbers of drones. This is primarily due to the fact that traditional approaches are often centralized. As the number of drones in a swarm increases, the computational complexity of centralized virtual force algorithms increases exponentially. Force calculations and position updates between each drone require significant computing resources, making it difficult for the system to respond and adjust in real time. Centralized control requires all drones to transmit information back to a central control unit for processing, which then feeds the results back to each drone. This approach increases communication latency in highly dynamic, multi-drone scenarios, impacting the collaborative efficiency and real-time performance of the entire swarm. Centralized virtual force algorithms fail to account for drone load. If a single drone becomes overloaded, mission execution for the entire swarm is severely impacted.

[0004] In multi-machine collaborative applications, dynamically adjusting UAV loads and predicting future topology can effectively improve the stability of UAV systems. Current research proposes an improved virtual potential field approach to coordinate UAV swarms and provide local topology information. Such studies typically only consider the attractive and repulsive forces between UAV swarms. In the paper "Yan, Mengfan, et al., "AVirtual-Potential-Field-Based Cooperative Opportunistic Routing Protocol for UAV Swarms," ​​presented at the International Conference on Wireless Algorithms, Systems, and Applications, Springer, Cham, 2021," a virtual potential field-based cooperative opportunistic routing protocol (VPFCOR) was proposed. VPFCOR combines three key metrics for candidate ranking: forwarding distance, virtual force, and residual energy. Leveraging these metrics, VPFCOR can effectively adapt to different tasks, reduce routing overhead, and meet real-time requirements. However, this approach only considers connectivity and generates a new relay network topology based solely on the current topology. Furthermore, it falls under the traditional centralized approach and does not consider the future. This can cause future topology changes to conflict with the maximum flight speed of UAVs, making it difficult to accommodate topology changes. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a relay network topology generation method for drone cluster applications.

[0006] The technical solution adopted by the present invention is: a method for generating relay network topology for drone cluster applications, which specifically includes the following steps:

[0007] S1. In the multi-machine collaborative application scenario, initialize the size of the multi-machine collaborative task area, the initial positions of all relay machines, the initial positions of all task machines, the communication range, the safe distance between drones, the business volume, and the business volume storage capacity. The relay machine communicates with the central control by covering all task machines.

[0008] S2. Model the data transmission requirements of the upstream mission machine to the relay machine as the attraction between UAVs. Combined with the relay machine load adaptive adjustment strategy, this strategy works together to adjust the relay machine's position. This adapts the network topology to the mission machine's transmission requirements while avoiding local overload of the relay machine.

[0009] S3. Build topology prediction capabilities by predicting the flight trajectory of task nodes at subsequent moments, guide the future topology movement of relay nodes, and trigger the early movement and deployment of relay nodes, reducing topology generation latency.

[0010] The Kalman filter algorithm is used to predict the position information of the task aircraft at the next moment. The position and velocity of the task aircraft are estimated through the state vector and linear state transition model. In the prediction and update steps of the Kalman filter, the process noise and observation noise covariance matrix are used for state estimation and correction. The predicted force vector is calculated based on the predicted position of the task aircraft and the current position of the relay aircraft.

[0011] S4. Based on the results of the two forces in steps S2 and S3, a complete improved virtual force algorithm system is constructed by comprehensively considering the topological attraction, prediction force, and the repulsive force between nodes and obstacles due to flight safety distance or obstacle avoidance. Through the combined action of different forces, the position of the relay drone is changed to optimize the network topology configuration;

[0012] After the relay UAV has the business volume attraction and prediction force, during the flight, when the distance between the UAVs is less than the safe distance, the repulsive force begins to act, and the magnitude of the force decays nonlinearly according to the distance. At the same time, the repulsive force of obstacles on the relay machine during the flight of the UAV must also be considered. By integrating these four force vectors, the final relay UAV topology structure is generated.

[0013] Furthermore, in step S2, a dynamic load balancing algorithm based on feedback control is introduced. First, the target load and feedback control are set. The target load L target,i :Each relay i sets a target load, which indicates the load level at which the UAV can operate stably; Feedback controller structure: Introduce PID controller to adjust the load balancing factor β i , the typical feedback controller includes three control links: ratio P, integration I and differentiation D;

[0014] A1, proportional term P: based on the current error e i (t) Adjust the load balancing factor β i (t), if the current load L i (t) is lower than the target load L target,i , that is, e i (t)>0, then increase β i (t) Increase the attractiveness of drones and attract more business;

[0015] On the contrary, if L i (t) is higher than L target,i , that is, e i (t)<0, then reduce β i (t) Reduce the attractiveness of drones and divert business volume;

[0016] A2, Integral term I: Integral control term I adjusts β by accumulating error i(t), used to eliminate steady-state errors;

[0017] A3. Differential term D: The differential control term D adjusts β by the error change rate. i (t), used to predict and adjust future error changes;

[0018] The feedback controller works as follows:

[0019] Error calculation e i (t): Calculate the current UAV load L i (t) and target load L target，i The error between .

[0020] e i (t) = L target，i -L i (t)

[0021] Controller output β i (t): According to the error e i (t) and controller parameters (proportional K p , integral K i and differential K d Coefficient) to calculate the adjustment amount of the load balancing factor.

[0022]

[0023] Among them, K p , K i and K d are the proportional, integral and differential gains, which are used to adjust the response characteristics of the controller, e i (t) is the error at time t, ∫e i (t)dt is the integral of the error, which is used to eliminate the steady-state error. is the rate of change of the error, which is used to predict the future error change trend;

[0024] The dynamic load balancing factor β i (t) Introduce the business volume attraction formula to adjust the attraction of drones to business traffic. The upstream drone is the source of the data flow and the matrix W is defined ni represents the traffic volume from upstream UAV n to relay i. If W ni ≠0, it means that there is a data flow from drone n to drone i, and the value represents the size of the business volume. Its attractiveness formula is:

[0025]

[0026] Among them, β i (t) is the load balancing factor adjustment of relay i; k biz,up: Business volume attraction coefficient, indicating the attraction intensity of upstream drones; d ni : The distance from the mission machine to the relay machine; : The unit direction vector from the task machine to the relay machine.

[0027] Total business volume attractiveness For the attraction of upstream drones:

[0028]

[0029] in, is the total business volume attractiveness, is the traffic attraction exerted by the upstream UAV n to the relay machine i.

[0030] Furthermore, in step S3, the Kalman filter algorithm is used to predict the position information of the task machine at the next moment as follows:

[0031] B1. Definition of state vector and observation vector. The state vector of task machine n Include Location and speed

[0032]

[0033] B2. State transition model. The state transition model of task machine n is a linear motion model: in, It describes how the state of the drone changes from time t-1 to time t without external control input. Is the control matrix, which describes the effect of control input (such as acceleration) on the state. In the motion of the UAV, acceleration is a key control input, which affects the position and velocity of the UAV. Here is the control vector (acceleration).

[0034] B3. Use the Kalman filter algorithm to estimate the state of the task machine n and covariance P n (t):

[0035] B3.1. Prediction steps:

[0036]

[0037]

[0038] in, is the process noise covariance matrix.

[0039] B3.2, Update steps:

[0040]

[0041]

[0042]

[0043] in, is the Kalman gain, is the observation matrix, is the observation noise covariance matrix.

[0044] B4, the next position of the task machine n estimated by Kalman filter Predictive power Expressed as the predicted speed direction vector of relay i:

[0045]

[0046] Among them, k predict is the predictive power coefficient, is the predicted position estimate of the task machine n, is the current location of relay i.

[0047] Furthermore, in step S4, the quantified model of the anti-collision force between UAVs and the repulsive force of obstacles is constructed as follows:

[0048] Repulsive force between relays Safety distance d safe When the distance between drones is less than this safe distance, the repulsive force begins to work, and the magnitude of the force decays nonlinearly according to the distance to avoid collisions between drones. The formula is as follows:

[0049]

[0050] Among them, k repel is the repulsive force coefficient, d ij is the distance between relays i and j, is the unit direction vector of UAV i and UAV j, is the indicator function, when d ij <d safe The value is 1 when , otherwise it is 0.

[0051] The repulsive force of obstacle o on relay i The formula is: Among them, k obs is the obstacle repulsion coefficient, A oi is the area or volume affected by obstacle o on relay i, or the effective area or volume of the obstacle, d oi is the distance from obstacle o to relay i, is a unit direction vector, indicating the direction of action of obstacle o on relay i.

[0052] Furthermore, in step S4, the resultant force expression of the improved virtual force algorithm in the entire UAV topology generation algorithm is specifically as follows:

[0053] Beneficial effects of the present invention: The relay network topology generation method of the present invention first initializes the size of the multi-machine collaborative task area, the initial positions of all relay machines, the initial positions of all task machines, the communication range, the safety distance between drones, the business volume size, and the business volume storage capacity size in the multi-machine collaborative area; then based on the PID control algorithm and the load balancing adaptive adjustment mechanism, the load of the relay machine is dynamically adjusted to avoid drone overload, and the business volume attraction is achieved by calculating the attraction of the upstream task machine to the relay machine; the Kalman filter algorithm is used to predict the position of the task machine at the next moment, and the prediction force is introduced to achieve the pulling of the relay machine. Taking into account the business volume attraction, prediction force, anti-collision force between drones and obstacle repulsion force, a complete virtual force algorithm system is constructed to ensure that the relay machine effectively covers the task machine and maintains communication with the central control. By introducing prediction force and business volume load balancing, this method optimizes the coverage effect and communication efficiency of the relay machine, and improves the performance and reliability of the drone group when performing complex tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flow chart of a relay network topology generation method based on an improved virtual force algorithm for drone cluster applications according to the present invention.

[0055] Figure 2 This is a schematic diagram of relay drone topology generation in a multi-machine collaboration scenario in an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In the multi-machine collaborative application scenario, the method of the present invention models the data transmission demand of the upstream task machine to the relay machine as the attraction between drones, and combines the relay machine load adaptive adjustment strategy to jointly act on the position adjustment of the relay machine, and realizes the adaptation of the network topology and the transmission demand of the task machine under the premise of avoiding local overload of the relay machine. Secondly, the present invention also constructs a topological prediction force by predicting the flight trajectory of the task machine at subsequent moments, realizes the guidance of the future topological movement of the relay machine, and then triggers the early movement and deployment of the relay machine to reduce the topology generation delay. Finally, the present invention comprehensively considers the topological attraction, prediction force, and the repulsive force between drones and between drones and obstacles due to flight safety distance or obstacle avoidance, and constructs a complete improved virtual force algorithm system. Through the joint action of different forces, the position of the relay drone is changed and the network topology configuration is optimized. The method of the present invention is further explained below with reference to the accompanying drawings and examples.

[0057] The process of the relay network topology generation method for drone cluster application of the present invention is as follows: Figure 1 The specific steps are as follows:

[0058] S1. In a multi-machine collaborative scenario, initialize the size of the multi-machine collaborative task area, the initial positions of all relay machines, the initial positions of all task machines, the communication range, the safe distance between drones, the business volume, and the business volume storage capacity. The relay machine communicates with the central control by covering all task machines.

[0059] S2. Model the data transmission requirements of the upstream mission machine to the relay machine as the attraction between UAVs. Combined with the relay machine load adaptive adjustment strategy, this strategy works together to adjust the relay machine's position. This adapts the network topology to the mission machine's transmission requirements while avoiding local overload of the relay machine.

[0060] Define the traffic volume of the upstream mission machine and calculate its attraction to the relay machine. Upstream drones can be distinguished based on the direction of data flow: upstream drones are data sources. The main reason for using it as a nonlinear form is that it can better reflect the effect of distance on force. A dynamic load balancing algorithm based on feedback control is introduced to adjust the load balancing factor β i ,For the attraction of the mission aircraft to the relay aircraft, it is hoped that the attraction is greater when the distance is close and rapidly decays as the distance increases.,This ensures that the relay aircraft is more inclined to approach the mission aircraft or convergence drone with a larger traffic volume;

[0061] S3. Build topology prediction capabilities by predicting the flight trajectory of task nodes at subsequent moments, guide the future topology movement of relay nodes, and trigger the early movement and deployment of relay nodes, reducing topology generation latency.

[0062] The Kalman filter algorithm is used to predict the next mission aircraft's position. The position and velocity of the mission aircraft are estimated using the state vector and linear state transition model. The covariance matrix of process noise and observation noise is then used for state estimation and correction during the Kalman filter's prediction and update steps. The predicted force vector is calculated based on the predicted mission aircraft position and the relay aircraft's current position.

[0063] S4, based on the results of the two forces in S2 and S3, comprehensively considering the topological attraction, prediction force, and the repulsive force between nodes and between nodes and obstacles due to flight safety distance or obstacle avoidance, a complete improved virtual force algorithm system was constructed. Through the combined action of different forces, the position of the relay drone was changed to optimize the network topology configuration;

[0064] After the relay UAV has the business volume attraction and prediction force, during the flight, when the distance between the UAVs is less than the safe distance, the repulsive force begins to act, and the magnitude of the force decays nonlinearly according to the distance. At the same time, the repulsive force of obstacles on the relay machine during the flight of the UAV must also be considered. By integrating these four force vectors, the final relay UAV topology structure is generated.

[0065] like Figure 2 As shown, in this embodiment, in step S1, there are a total of N mission drones and M relay drones, and their communication radius is r and their safety distance is d. safe , the maximum data storage capacity is d max .

[0066] In this embodiment, in step S2, a dynamic load balancing algorithm based on feedback control is introduced.

[0067] First set the target load and feedback control, the target load L target,i :Each relay i sets a target load, which represents the load level at which the UAV can operate stably. Feedback controller structure: Too high a load will lead to problems such as increased latency and packet loss, while too low a load may lead to resource waste. In order to dynamically adjust the load of each UAV, we can introduce a PID controller to adjust the load balancing factor β i A typical feedback controller includes three control links: proportional P, integral I and differential D.

[0068] A1, proportional term P: based on the current error e i (t) Adjust the load balancing factor β i (t), if the current load L i (t) is lower than the target load L target,i , that is, e i (t)>0, then increase β i (t) Increase the attractiveness of drones and attract more business;

[0069] On the contrary, if L i (t) is higher than L target,i , that is, e i (t)<0, then reduce β i (t), reduce the attractiveness of drones and divert business volume.

[0070] A2, Integral term I: Integral control term I adjusts β by accumulating error i (t), used to eliminate steady-state errors, if K i Increase means that the accumulation of historical errors is given more attention. When the error accumulation is large, β i As (t) increases, the repeater attracts more traffic, and vice versa.

[0071] A3. Differential term D: The differential control term D adjusts β by the error change rate. i (t), used to predict and adjust future error changes, if K d Increase means that the response to the error change rate is more sensitive. When the error change rate is large, β i (t) increases or decreases, and the attraction of the relay is adjusted quickly.

[0072] The feedback controller works as follows:

[0073] Error calculation e i (t): Calculate the current UAV load L i (t) and target load L target，i The error between .

[0074] e i (t) = L target，i -L i (t)

[0075] Controller output β i (t): According to the error e i (t) and controller parameters (proportional K p , integral K i and differential K d Coefficient) to calculate the adjustment amount of the load balancing factor.

[0076]

[0077] Among them, K p , K i and K d are the proportional, integral and differential gains, which are used to adjust the response characteristics of the controller, e i (t) is the error at time t, ∫e i (t)dt is the integral of the error, which is used to eliminate the steady-state error. It is the rate of change of error, which is used to predict future error change trends.

[0078] The dynamic load balancing factor β i The business volume attraction formula is introduced to adjust the attraction of drones to business traffic. The upstream drone is the source of the data flow, that is, the definition matrix W ni represents the traffic volume from upstream UAV n to relay i. If W ni ≠0, it means that there is a data flow from UAV n to UAV i, and the value represents the size of the business volume.

[0079] The attractiveness formula is:

[0080] Among them, β i (t) is the load balancing factor adjustment of relay i;

[0081] k biz,up : Business volume attraction coefficient, indicating the attraction strength of upstream UAVs;

[0082] d ni : The distance from the mission machine to the relay machine;

[0083] : The unit direction vector from the task machine to the relay machine.

[0084] Total business volume attractiveness For the attraction of upstream drones:

[0085]

[0086] in, is the total business volume attractiveness, is the traffic attraction exerted by the upstream UAV n to the relay machine i.

[0087] like Figure 2 As shown, in this embodiment, in step S3, the Kalman filter algorithm is used to predict the position information of the task machine at the next moment as follows:

[0088] B1. Definition of state vector and observation vector. The state vector of task machine n Include Location and speed

[0089]

[0090] B2. State transition model. The state transition model of task machine n is a linear motion model: in, It mainly describes how the state of the drone changes from time t-1 to time t without external control input. Is the control matrix, which describes the effect of control input (such as acceleration) on the state. In the motion of the UAV, acceleration is a key control input, which affects the position and velocity of the UAV. Here is the control vector (acceleration).

[0091] B3. Use the Kalman filter algorithm to estimate the state of the task machine n and covariance P n (t):

[0092] B3.1. Prediction steps:

[0093]

[0094]

[0095] in, is the process noise covariance matrix.

[0096] B3.2, Update steps:

[0097]

[0098]

[0099]

[0100] in, is the Kalman gain, is the observation matrix, is the observation noise covariance matrix.

[0101] B4, the next position of the task machine n estimated by Kalman filter Predictive power It can be expressed as the predicted speed direction vector of relay i:

[0102]

[0103] Among them, k predict is the predictive power coefficient, is the predicted position estimate of the task machine n, is the current location of relay i.

[0104] like Figure 2 As shown, in this embodiment, in step S4, the quantified model of the anti-collision force between drones and the repulsive force of obstacles is constructed as follows:

[0105] Repulsive force between relays Safety distance d safe When the distance between drones is less than this safe distance, the repulsive force begins to work, and the magnitude of the force decays nonlinearly according to the distance to avoid collisions between drones. The formula is as follows:

[0106]

[0107] Among them, k repel is the repulsive force coefficient, d ij is the distance between relays i and j, is the unit direction vector of UAV i and UAV j, d safe For a safe distance, is the indicator function, when d ij <dsafe The value is 1 when , otherwise it is 0.

[0108] The repulsive force of obstacle o on relay i for: Among them, k obs is the obstacle repulsion coefficient, A oi is the area or volume affected by obstacle o on relay i, or the effective area or volume of the obstacle; d oi is the distance from obstacle o to relay i, is a unit direction vector, indicating the direction of action of obstacle o on relay i.

[0109] In step S4, the resultant force expression of the improved virtual force algorithm in the entire UAV topology generation algorithm is as follows:

[0110] Based on steps S1-S4, different from the virtual force algorithms studied in previous studies, this embodiment, based on the original virtual force, introduces relay nodes to provide communication relay services in multi-machine collaborative application scenarios to ensure that the task nodes always maintain communication with the remote command and control center and support the transmission of data generated by the task nodes; by dynamically adjusting the positions of relay nodes, the communication topology configuration between drones is optimized to ensure network connectivity and data transmission performance; by introducing load balancing mechanism and prediction force, comprehensively considering topological attraction, prediction force, and the repulsive force between drones and between drones and obstacles due to flight safety distance or obstacle avoidance, a complete improved virtual force algorithm system is constructed. Through the combined action of different forces, the position of relay drones is changed to optimize the network topology configuration.

[0111] In summary, to ensure effective relay coverage of task aircraft and communication with the central control, the method of the present invention first initializes the size of the multi-machine collaborative task area, the initial positions of all relays, the initial positions of all task aircraft, the communication range, the safe distance between drones, the traffic volume, and the traffic storage capacity within the multi-machine collaborative area. The upstream task aircraft's data transmission requirements for the relay are then modeled as the attractive force between drones. This, combined with the relay aircraft's load adaptive adjustment strategy, acts on the relay aircraft's position adjustment, achieving adaptation of the network topology to the task aircraft's transmission requirements while avoiding local overload on the relay aircraft. A Kalman filter algorithm is then used to construct a topology prediction force by predicting the flight trajectory of task nodes at subsequent moments. This guides the future topological movement of relay nodes, thereby triggering their early movement and deployment, reducing topology generation latency. By comprehensively considering topological attraction, prediction force, and the repulsive forces between nodes and between nodes and obstacles due to flight safety distances or obstacle avoidance, a complete improved virtual force algorithm system is constructed. Through the combined action of different forces, the position of relay drones is altered to optimize the network topology.

[0112] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. A method for generating a relay network topology for drone swarm applications, comprising the following steps: S1. In a multi-machine collaborative application scenario, initialize the size of the multi-machine collaborative task area, the initial positions of all relay machines, the initial positions of all task machines, the communication range, the safe distance between drones, the business volume, and the business volume storage capacity. The relay machine communicates with the central control by covering all task machines; S2. Model the data transmission requirements of the upstream mission machine to the relay machine as the attraction between UAVs. Combined with the relay machine load adaptive adjustment strategy, this strategy works together to adjust the relay machine's position. This adapts the network topology to the mission machine's transmission requirements while avoiding local overload of the relay machine. S3. Build topology prediction capabilities by predicting the flight trajectory of task nodes at subsequent moments, guide the future topology movement of relay nodes, and trigger the early movement and deployment of relay nodes, reducing topology generation latency. The Kalman filter algorithm is used to predict the position information of the task aircraft at the next moment. The position and velocity of the task aircraft are estimated through the state vector and linear state transition model. In the prediction and update steps of the Kalman filter, the process noise and observation noise covariance matrix are used for state estimation and correction. The predicted force vector is calculated based on the predicted position of the task aircraft and the current position of the relay aircraft. S4. Based on the results of the two forces in steps S2 and S3, a complete improved virtual force algorithm system is constructed by comprehensively considering the topological attraction, prediction force, and the repulsive force between nodes and obstacles due to flight safety distance or obstacle avoidance. Through the combined action of different forces, the position of the relay drone is changed to optimize the network topology configuration; After the relay UAV has the business volume attraction and prediction force, during the flight, when the distance between the UAVs is less than the safe distance, the repulsive force begins to act, and the magnitude of the force decays nonlinearly according to the distance. At the same time, the repulsive force of obstacles on the relay machine during the flight of the UAV must also be considered. By integrating these four force vectors, the final relay UAV topology structure is generated.

2. A method for generating a relay network topology for a UAV cluster application according to claim 1, characterized in that: In step S2, a dynamic load balancing algorithm based on feedback control is introduced. First, the target load and feedback control are set. The target load L target,i :Each relay i sets a target load, which indicates the load level at which the UAV can operate stably; Feedback controller structure: Introduce PID controller to adjust the load balancing factor β i , the typical feedback controller includes three control links: ratio P, integration I and differentiation D; A1, proportional term P: based on the current error e i (t) Adjust the load balancing factor β i (t), if the current load L i (t) is lower than the target load L target,i , that is, e i (t)>0, then increase β i (t) Increase the attractiveness of drones and attract more business; On the contrary, if L i (t) is higher than L target,i , that is, e i (t)<0, then reduce β i (t) Reduce the attractiveness of drones and divert business volume; A2, Integral term I: Integral control term I adjusts β by accumulating error i (t), used to eliminate steady-state errors; A3. Differential term D: The differential control term D adjusts β by the error change rate. i (t), used to predict and adjust future error changes; The feedback controller works as follows: Error calculation e i (t): Calculate the current UAV load L i (t) and target load L target，i The error between e i (t)=L target，i -L i (t) Controller output β i (t): According to the error e i (t) and the controller parameters to calculate the adjustment of the load balancing factor, Among them, K p , K i and K s are the proportional, integral and differential gains, which are used to adjust the response characteristics of the controller; e i (t) is the error at time t; ∫e i (t)dt is the integral of the error, which is used to eliminate the steady-state error; is the rate of change of the error, which is used to predict the future error change trend; The dynamic load balancing factor β i (t) Introduce the business volume attraction formula to adjust the attraction of drones to business traffic. The upstream drone is the source of the data flow and the matrix W is defined ni represents the traffic volume from upstream UAV n to relay i. If W ni ≠0, it means that there is a data flow from drone n to drone i, and the value represents the size of the business volume. Its attractiveness formula is: Among them, β i (t) is the load balancing factor adjustment of relay i; k biz,up : Business volume attraction coefficient, indicating the attraction intensity of upstream drones; d ni : The distance from the mission machine to the relay machine; : The unit direction vector from the mission machine to the relay machine; Total business volume attractiveness For the attraction of upstream drones: in, is the total business volume attractiveness, is the traffic attraction exerted by the upstream UAV n to the relay machine i.

3. The method for generating a relay network topology for a UAV cluster application according to claim 1, characterized in that: In step S3, the Kalman filter algorithm is used to predict the position information of the task machine at the next moment as follows: B1. Definition of state vector and observation vector. The state vector of task machine n Include Location and speed B2. State transition model. The state transition model of task machine n is a linear motion model: in, It describes how the state of the drone changes from time t-1 to time t without external control input. is the control matrix, is the control vector; B3. Use the Kalman filter algorithm to estimate the state of the task machine n and covariance P n (t): B3.

1. Prediction steps: in, is the process noise covariance matrix; B3.2, Update steps: in, is the Kalman gain, is the observation matrix, is the observation noise covariance matrix; B4, the next position of the task machine n estimated by Kalman filter Predictive power Expressed as the predicted speed direction vector of relay i: Among them, k predict is the predictive power coefficient, is the predicted position estimate of the task machine n, is the current location of relay i.

4. The method for generating a relay network topology for a UAV cluster application according to claim 1, wherein: In step S4, the quantified model of the anti-collision force between UAVs and the repulsive force of obstacles is constructed as follows: Repulsive force between relays Safety distance d safe When the distance between drones is less than this safe distance, the repulsive force begins to work, and the magnitude of the force decays nonlinearly according to the distance to avoid collisions between drones. The formula is as follows: Among them, k repel is the repulsive force coefficient, d ij is the distance between relays i and j, is the unit direction vector of UAV i and UAV j, is the indicator function, when d ij <d safe The value is 1 when , otherwise it is 0; The repulsive force of obstacle o on relay i for: Among them, k obs is the obstacle repulsion coefficient, A oi is the area or volume affected by obstacle o on relay i, or the effective area or volume of the obstacle, d oi is the distance from obstacle o to relay i, is a unit direction vector, indicating the direction of action of obstacle o on relay i.

5. The method for generating a relay network topology for a UAV cluster application according to claim 1, wherein: In step S4, the resultant force expression of the improved virtual force algorithm in the entire UAV topology generation algorithm is as follows:

Images