A brain-like control method and system for unmanned aerial vehicle patrol

Abstract

This invention discloses a brain-like control method and system for unmanned aerial vehicle (UAV) patrol. It performs visual neuron analysis on the three-dimensional structural data of the flight space to obtain all accessible patrol subspaces. Based on the potential interference layout of the accessible patrol subspaces, it calibrates suitable patrol paths and their characteristics. It retrieves historical flight memory data from the UAV swarm to generate performance labels for each UAV within the swarm. By comparing the performance labels with the path characteristics of the patrol paths, it forms patrol task allocation decisions and obtains flight state change plans for each UAV during patrol tasks. Based on the UAV's flight constraints and flight state change plans, it modifies the UAV's motion parameters to adapt to the uncertainty and variability of the external environment during patrol, thereby improving the robustness and accuracy of UAV flight control.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicles (UAVs), and in particular to a brain-like control method and system for UAV patrol. Background Technology

[0002] Unmanned aerial vehicles (UAVs) are widely used in various scenarios due to their high flexibility, strong flight controllability, and compatibility with multiple sensing devices. However, considering that the effective coverage of a single UAV for monitoring and patrolling is insufficient to meet the needs of rapid and comprehensive patrol and monitoring in large-space scenarios, UAV swarms are used for zoned monitoring of target spaces. However, the internal environment of a target space is not uniform, exhibiting a complex and variable three-dimensional structure. If each UAV in the swarm uses the same monitoring mode to patrol its corresponding zone, it cannot accurately patrol based on the unique three-dimensional characteristics of each zone, reducing patrol and monitoring efficiency and reliability. Furthermore, existing UAV swarms lack personalized patrol control for each UAV, and cannot perform brain-like autonomous flight control in response to the uncertainty and variability of the external environment during patrols, thus reducing the robustness and accuracy of UAV flight control. Summary of the Invention

[0003] Considering that existing drone swarms only partition the target space for large-scale target space monitoring, restricting each drone in the swarm to a relatively small area for patrol and monitoring, they cannot accurately patrol based on the three-dimensional characteristics of the partition's environment, nor can they adapt to the uncertainty and variability of the external environment during patrol, thus reducing the robustness and accuracy of drone flight control. In view of the above problems, this invention proposes a brain-like control method for drone patrol to overcome or at least partially solve the above problems, comprising: The system senses the three-dimensional structural data of the space in which the flight is located, performs visual neural network analysis on the three-dimensional structural data, and obtains all accessible patrol subspaces in the space in which the flight is located; based on the potential interference layout of the accessible patrol subspaces, it calibrates the appropriate patrol path and its path characteristics. Historical flight memory data of the drone swarm is retrieved to generate performance tags for each drone in the swarm; the performance tags are compared with the path characteristics of the patrol path to form a patrol task allocation decision. Spatial cognitive learning is performed on the patrol task allocation decision to obtain the flight state change plan for each UAV during the patrol task; based on the flight constraints of the UAV, navigation neural network analysis is performed on the flight state change plan to change the motion parameters of the UAV.

[0004] Optionally, the three-dimensional structural data of the space where the flight is located is perceived, and visual neural network analysis is performed on the three-dimensional structural data to obtain all accessible patrol subspaces in the space where the flight is located; based on the potential interference layout of the accessible patrol subspaces, suitable patrol paths and their path characteristics are calibrated, including: The system senses the image data of objects in the three-dimensional domain of the space where the flight is taking place, and extracts contour feature data and depth feature data from the object image data. Based on the contour feature data and the depth feature data, three-dimensional structural data of each object in the flight space is generated; wherein, the three-dimensional structural data includes the contour-depth fusion data of each object in the three-dimensional domain; Visual neural network analysis is performed on the three-dimensional structural data of all objects to obtain the layout of the empty space between all objects; several connected subspaces are identified from the empty space layout as passable patrol subspaces. Based on the three-dimensional size data of objects within the accessible patrol subspace, a potential interference layout is determined; wherein, the potential interference layout refers to the spatial distribution of UAV communication interference intensity caused by object obstruction within the accessible patrol subspace. Based on the spatial distribution of UAV communication interference intensity included in the potential interference layout, several patrol paths and their path characteristics are identified within the passable patrol subspace; wherein, the patrol path refers to a one-way patrol path within the passable patrol subspace where the average communication interference intensity does not exceed a preset interference intensity threshold; and the path characteristics refer to the average communication interference intensity corresponding to the entire range of the patrol path.

[0005] Optionally, historical flight memory data of the drone swarm is retrieved to generate performance tags for each drone in the swarm; the performance tags are compared with the path characteristics of the patrol path to form a patrol task allocation decision, including: Retrieve historical flight memory data of the drone cluster; wherein, the historical flight memory data includes the anti-communication interference operation data of each drone under the drone cluster during historical flights; based on the historical flight memory data, generate an anti-communication interference performance tag for each drone; wherein, the anti-communication interference performance tag refers to the maximum intensity of communication interference that each drone can withstand; By comparing the anti-communication interference performance label of each UAV with the average communication interference intensity corresponding to the entire range of each patrol path, a patrol task allocation decision is formed; wherein, the patrol task allocation decision includes assigning each UAV a patrol path to complete the patrol.

[0006] Optionally, spatial cognitive learning is performed on the patrol task allocation decision to obtain a flight state change plan for each UAV during its patrol task; based on the flight constraints of the UAV, navigation neural network analysis is performed on the flight state change plan to change the motion parameters of the UAV, including: Spatial trajectory morphology cognitive learning is performed on the patrol path assigned to each UAV within the patrol task allocation decision to obtain flight position change data of each UAV during the patrol task execution along the corresponding patrol path; based on the flight position change data and the external dimensions of the UAV, a flight state change plan is obtained. Extract the maximum change in flight direction angle and the maximum change in flight speed of the UAV per unit time from the flight constraints of the UAV; perform navigation neural network analysis on the maximum change in flight direction angle, the maximum change in flight speed, and the flight state change plan, so as to change the flight direction and / or flight speed of the UAV in a future preset time interval.

[0007] Optionally, based on the flight position change data of each UAV during its patrol mission along the corresponding patrol path, it is determined whether there are any anomalies in the UAV flight, including: Retrieve the actual location coordinates of the drone at each data acquisition moment; Retrieve the expected location coordinates of the drone at each data acquisition moment; The drone flight error parameters are obtained based on the actual and expected position coordinates of the drone at each acquisition time. The UAV flight error parameters are obtained using the following formula: , Where W represents the UAV flight error parameters; n represents the total number of position data acquisitions; x i x represents the actual position coordinates of the UAV at the i-th data acquisition time; ei This represents the expected location coordinates of the UAV at the i-th data acquisition time. The drone flight error parameters are compared with preset error parameter thresholds; If the flight error parameters of the UAV do not exceed the preset error parameter threshold, the flight anomaly determination will not be initiated. When the flight error parameters of the UAV exceed the preset error parameter threshold, the actual change in flight direction angle and the actual change in flight speed per unit time of the UAV are retrieved. Flight anomaly parameters are obtained by combining the actual flight direction angle change and actual flight speed change per unit time of the UAV with the UAV flight error parameters. The flight anomaly parameters are obtained using the following formula: , Where E represents flight anomaly parameters; W represents UAV flight error parameters; Δθ represents the change in actual flight direction angle per unit time; Δθ max Δv represents the maximum permissible change in the drone's heading angle; Δv represents the change in actual flight speed per unit time; Δv max Indicates the maximum permissible speed change for the drone; k 01 and k 02 These represent the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed, respectively, and the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed are obtained by the following formulas: , Where μ represents the preset task priority factor, with a value range of 0-1; η represents the normalized direction control accuracy coefficient, determined by the IMU sensor noise variance, and η=1 / (1+g 2 Meanwhile, g represents the normalized IMU sensor noise variance; θ g This represents the maximum permissible change in the drone's heading angle after normalization; λ represents the near-range attenuation coefficient, which defaults to 0.01 / m; σ represents the standard deviation of the current wind direction angle after normalization; v a The normalized airspeed is represented by d; d represents the distance from the current position to the next waypoint. , Among them, T e The current motor temperature of the drone is indicated by s; the normalized average response time of the drone's flight direction control is indicated by h; and the difference between the current altitude and the preset altitude is indicated by h. max Indicates the maximum allowable altitude deviation for the task; v g T represents the maximum permissible speed change of the drone after normalization. g This indicates the remaining battery life after normalization. The flight anomaly parameters are compared with preset first anomaly parameter thresholds and second anomaly parameter thresholds; When the flight anomaly parameter does not exceed the preset first anomaly parameter threshold, it is determined to be a low-level anomaly, and a low-level anomaly warning is issued. When the flight anomaly parameter exceeds the preset first anomaly parameter threshold, but does not exceed the preset second anomaly parameter threshold, it is determined to be a moderate anomaly and a moderate anomaly warning is issued. When the flight anomaly parameter exceeds the preset second anomaly parameter threshold, it is determined to be an altitude anomaly, and an altitude anomaly warning is issued.

[0008] As one aspect of the present invention, embodiments of the present invention also provide a brain-like control system for unmanned aerial vehicle patrol, comprising: The subspace determination module is used to sense the three-dimensional structural data of the space where the flight is located, perform visual neuron analysis on the three-dimensional structural data, and obtain all passable patrol subspaces of the space where the flight is located. The path identification module is used to identify the appropriate patrol path and its path characteristics based on the potential interference layout of the accessible patrol subspace. The performance tag generation module is used to retrieve historical flight memory data of the drone cluster and generate the performance tags of each drone in the drone cluster. The task allocation decision generation module is used to compare the work performance tags and the path characteristics of the patrol path to form a patrol task allocation decision. The planning and determination module is used to perform spatial cognitive learning on the patrol task allocation decision to obtain the flight status change plan for each UAV during the patrol task. The motion change module is used to perform navigation neuron analysis on the flight state change plan based on the flight constraints of the UAV, thereby changing the motion parameters of the UAV.

[0009] Optionally, the subspace determination module is used to sense the three-dimensional structural data of the flight space, perform visual neural network analysis on the three-dimensional structural data, and obtain all accessible patrol subspaces of the flight space, including: The system senses the image data of objects in the three-dimensional domain of the space where the flight is taking place, and extracts contour feature data and depth feature data from the object image data. Based on the contour feature data and the depth feature data, three-dimensional structural data of each object in the flight space is generated; wherein, the three-dimensional structural data includes the contour-depth fusion data of each object in the three-dimensional domain; Visual neural network analysis is performed on the three-dimensional structural data of all objects to obtain the layout of the empty space between all objects; several connected subspaces are identified from the empty space layout as passable patrol subspaces. The path identification module is used to identify suitable patrol paths and their path characteristics based on the potential interference layout of the passable patrol subspace, including: Based on the three-dimensional size data of objects within the accessible patrol subspace, a potential interference layout is determined; wherein, the potential interference layout refers to the spatial distribution of UAV communication interference intensity caused by object obstruction within the accessible patrol subspace. Based on the spatial distribution of UAV communication interference intensity included in the potential interference layout, several patrol paths and their path characteristics are identified within the passable patrol subspace; wherein, the patrol path refers to a one-way patrol path within the passable patrol subspace where the average communication interference intensity does not exceed a preset interference intensity threshold; and the path characteristics refer to the average communication interference intensity corresponding to the entire range of the patrol path.

[0010] Optionally, the performance tag generation module is used to retrieve historical flight memory data of the drone cluster to generate individual performance tags for all drones within the drone cluster, including: Retrieve historical flight memory data of the drone cluster; wherein, the historical flight memory data includes the anti-communication interference operation data of each drone under the drone cluster during historical flights; based on the historical flight memory data, generate an anti-communication interference performance tag for each drone; wherein, the anti-communication interference performance tag refers to the maximum intensity of communication interference that each drone can withstand; The task allocation decision generation module is used to compare the work performance tags and the path characteristics of the patrol path to form a patrol task allocation decision, including: By comparing the anti-communication interference performance label of each UAV with the average communication interference intensity corresponding to the entire range of each patrol path, a patrol task allocation decision is formed; wherein, the patrol task allocation decision includes assigning each UAV a patrol path to complete the patrol.

[0011] Optionally, the planning determination module is used to perform spatial cognitive learning on the patrol task allocation decision to obtain a flight state change plan for each UAV during the patrol task, including: Spatial trajectory morphology cognitive learning is performed on the patrol path assigned to each UAV within the patrol task allocation decision to obtain flight position change data of each UAV during the patrol task execution along the corresponding patrol path; based on the flight position change data and the external dimensions of the UAV, a flight state change plan is obtained. The motion change module is used to perform navigation neuron analysis on the flight state change plan based on the flight constraints of the UAV, thereby changing the motion parameters of the UAV, including: Extract the maximum change in flight direction angle and the maximum change in flight speed of the UAV per unit time from the flight constraints of the UAV; perform navigation neural network analysis on the maximum change in flight direction angle, the maximum change in flight speed, and the flight state change plan, so as to change the flight direction and / or flight speed of the UAV in a future preset time interval.

[0012] Optionally, based on the flight position change data of each UAV during its patrol mission along the corresponding patrol path, it is determined whether there are any anomalies in the UAV flight, including: Retrieve the actual location coordinates of the drone at each data acquisition moment; Retrieve the expected location coordinates of the drone at each data acquisition moment; The drone flight error parameters are obtained based on the actual and expected position coordinates of the drone at each acquisition time. The UAV flight error parameters are obtained using the following formula: , Where W represents the UAV flight error parameters; n represents the total number of position data acquisitions; x i x represents the actual position coordinates of the UAV at the i-th data acquisition time; ei This represents the expected location coordinates of the UAV at the i-th data acquisition time. The drone flight error parameters are compared with preset error parameter thresholds; If the flight error parameters of the UAV do not exceed the preset error parameter threshold, the flight anomaly determination will not be initiated. When the flight error parameters of the UAV exceed the preset error parameter threshold, the actual change in flight direction angle and the actual change in flight speed per unit time of the UAV are retrieved. Flight anomaly parameters are obtained by combining the actual flight direction angle change and actual flight speed change per unit time of the UAV with the UAV flight error parameters. The flight anomaly parameters are obtained using the following formula: , Where E represents flight anomaly parameters; W represents UAV flight error parameters; Δθ represents the change in actual flight direction angle per unit time; Δθ max Δv represents the maximum permissible change in the drone's heading angle; Δv represents the change in actual flight speed per unit time; Δv max Indicates the maximum permissible speed change for the drone; k 01 and k 02 These represent the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed, respectively, and the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed are obtained by the following formulas: , Where μ represents the preset task priority factor, with a value range of 0-1; η represents the normalized direction control accuracy coefficient, determined by the IMU sensor noise variance, and η=1 / (1+g2 Meanwhile, g represents the normalized IMU sensor noise variance; θ g This represents the maximum permissible change in the drone's heading angle after normalization; λ represents the near-range attenuation coefficient, which defaults to 0.01 / m; σ represents the standard deviation of the current wind direction angle after normalization; v a The normalized airspeed is represented by d; d represents the distance from the current position to the next waypoint. , Among them, T e The current motor temperature of the drone is indicated by s; the normalized average response time of the drone's flight direction control is indicated by h; and the difference between the current altitude and the preset altitude is indicated by h. max Indicates the maximum allowable altitude deviation for the task; v g T represents the maximum permissible speed change of the drone after normalization. g This indicates the remaining battery life after normalization. The flight anomaly parameters are compared with preset first anomaly parameter thresholds and second anomaly parameter thresholds; When the flight anomaly parameter does not exceed the preset first anomaly parameter threshold, it is determined to be a low-level anomaly, and a low-level anomaly warning is issued. When the flight anomaly parameter exceeds the preset first anomaly parameter threshold, but does not exceed the preset second anomaly parameter threshold, it is determined to be a moderate anomaly and a moderate anomaly warning is issued. When the flight anomaly parameter exceeds the preset second anomaly parameter threshold, it is determined to be an altitude anomaly, and an altitude anomaly warning is issued.

[0013] The beneficial effects of the above-mentioned technical solutions provided in the embodiments of the present invention include at least the following: This invention provides a brain-like control method and system for unmanned aerial vehicle (UAV) patrol. The method involves performing visual neuron analysis on the three-dimensional structural data of the flight space to obtain all navigable patrol subspaces. Based on the potential interference layout of these subspaces, suitable patrol paths and their characteristics are calibrated. Historical flight memory data of the UAV swarm is retrieved to generate performance labels for each UAV within the swarm. The performance labels and path characteristics of the patrol paths are compared to form patrol task allocation decisions, resulting in flight state change plans for each UAV during its patrol mission. Based on the UAV's flight constraints and flight state change plans, the UAV's motion parameters are modified. By utilizing brain-like perception analysis of the three-dimensional structural data to calibrate several patrol paths and combining them with the UAV's performance characteristics, adaptive patrol task allocation is performed, adapting to the uncertainty and variability of the external environment during patrol, thereby improving the robustness and accuracy of UAV flight control.

[0014] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0015] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the brain-like control method for drone patrol provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the brain-like control system for drone patrol provided in an embodiment of the present invention. Detailed Implementation

[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0018] Please see Figure 1 As shown, one embodiment of this application provides a brain-inspired control method for unmanned aerial vehicle (UAV) patrol. This brain-inspired control method for UAV patrol includes: The system senses the three-dimensional structural data of the space in which the flight takes place, performs visual neural network analysis on the three-dimensional structural data, and obtains all accessible patrol subspaces in the space in which the flight takes place; based on the potential interference layout of the accessible patrol subspaces, it calibrates the appropriate patrol path and its path characteristics. Retrieve historical flight memory data of the drone swarm to generate individual performance tags for each drone in the swarm; compare the performance tags with the path characteristics of the patrol routes to make patrol task allocation decisions. Spatial cognitive learning is used to make patrol mission allocation decisions, resulting in flight state change plans for each UAV during patrol missions. Based on the UAV's flight constraints, navigation neural network analysis is performed on the flight state change plans to change the UAV's motion parameters.

[0019] The beneficial effects of the above embodiments are as follows: This brain-like control method for UAV patrol performs visual neuron analysis on the three-dimensional structural data of the flight space to obtain all navigable patrol subspaces; based on the potential interference layout of the navigable patrol subspaces, it calibrates suitable patrol paths and their path characteristics; it retrieves historical flight memory data of the UAV swarm to generate individual performance labels for each UAV in the swarm; it compares the performance labels with the path characteristics of the patrol paths to form patrol task allocation decisions and obtains the flight state change plan for each UAV during its patrol task; and it changes the UAV's motion parameters based on the UAV's flight constraints and flight state change plan. By utilizing brain-like perception to analyze the three-dimensional structural data, calibrating several patrol paths, and combining the UAV's performance characteristics, it performs adaptive patrol task allocation, adapting to the uncertainty and variability of the external environment during patrol, and improving the robustness and accuracy of UAV flight control.

[0020] In another embodiment, the three-dimensional structural data of the space where the flight is located is perceived, and visual neural network analysis is performed on the three-dimensional structural data to obtain all accessible patrol subspaces in the space where the flight is located; based on the potential interference layout of the accessible patrol subspaces, suitable patrol paths and their path characteristics are calibrated, including: Perceive the image data of objects in the three-dimensional domain of the space where the flight is taking place, and extract contour feature data and depth feature data from the object image data; Based on contour feature data and depth feature data, generate three-dimensional structural data for each object in the flight space; wherein, the three-dimensional structural data includes contour-depth fusion data of each object in the three-dimensional domain. Visual neural network analysis is performed on the 3D structural data of all objects to obtain the layout of the empty space between all objects; several connected subspaces are identified from the empty space layout as passable patrol subspaces. Based on the three-dimensional size data of objects within the accessible patrol subspace, the potential interference layout is determined; whereby, the potential interference layout refers to the spatial distribution of UAV communication interference intensity caused by object obstruction within the accessible patrol subspace. Based on the spatial distribution of UAV communication interference intensity included in the potential interference layout, several suitable patrol paths and their path characteristics are identified within the passable patrol subspace. Among them, the patrol path refers to a one-way patrol path within the passable patrol subspace where the average communication interference intensity does not exceed a preset interference intensity threshold; the path characteristics refer to the average communication interference intensity corresponding to the entire range of the patrol path.

[0021] The beneficial effects of the above embodiments are that the three-dimensional spatial structure of objects such as buildings, mountains, and trees in the space where the UAV swarm flies directly determines the flight space of the UAVs. In order for the UAVs to fly continuously for a certain distance in the above space, it is necessary to mark the blank range of the above space and determine the subspaces that the UAVs can fly through. Specifically, the binocular camera sensor of one of the UAVs in the swarm (such as the UAV at the front of the swarm) first collects object image data in the three-dimensional domain of the flight space. The object image data can be, but is not limited to, the overall three-dimensional image data of the entire flight space or several local three-dimensional image data of different parts of the global range. Then, the object image data is subjected to contour and depth recognition to obtain contour feature data and depth feature data. The contour feature data includes the contour features of each object in the flight space, and the depth feature data includes the depth value of each object in the flight space. Then, taking each object in the flight space as a reference, the contour feature data and depth feature data corresponding to the same object are fused to generate contour-depth fusion data of each object in the three-dimensional domain, so as to realize the comprehensive three-dimensional characteristic calibration of all objects in the flight space.

[0022] By using a brain-like neural network model to perform visual neuron analysis on the 3D structural data of all objects, the layout of the empty space between all objects is obtained. Then, the connected subspaces that meet the preset connectivity length condition (such as exceeding the preset connectivity length threshold) are designated as passable patrol subspaces (i.e., subspaces that allow UAVs to fly through). This allows for the accurate marking of the feasible movement space of UAVs in the flight space, providing a basis for the subsequent assignment of patrol tasks to each UAV.

[0023] Furthermore, drones rely on wireless communication with base stations or the main drone to maintain continuous and stable flight. However, obstructions from buildings and / or trees weaken the propagation strength of wireless communication signals, interfering with the drone's wireless communication and affecting its flight controllability and safety. Different drones have varying anti-interference communication capabilities, and the degree of performance degradation varies even when faced with the same obstruction from buildings and / or trees. Therefore, to enable drones with high anti-interference communication capabilities to patrol and monitor in environments with strong external communication interference, it is necessary to first categorize and label the drone communication interference situation within the accessible patrol subspace. Specifically, based on the three-dimensional dimensions of objects within the accessible patrol subspace, the spatial distribution of drone communication interference intensity caused by object obstruction (i.e., the spatial distribution of drone communication interference magnitude) is determined. The greater the drone communication interference intensity, the worse the corresponding drone's wireless communication performance. Based on the spatial distribution of UAV communication interference intensity within the potential interference layout, several suitable patrol paths and their characteristics are identified within the passable patrol subspace. This enables accurate differentiation and identification of the average communication interference intensity experienced by all patrol paths within the passable patrol subspace, providing a basis for subsequent patrol task allocation decisions.

[0024] In another embodiment, historical flight memory data of the drone swarm is retrieved to generate individual performance tags for each drone in the swarm; the performance tags are compared with the path characteristics of the patrol paths to form a patrol task allocation decision, including: Retrieve historical flight memory data of the drone swarm; the historical flight memory data includes the anti-communication interference operation data of each drone in the drone swarm during historical flights; based on the historical flight memory data, generate anti-communication interference performance tags for each drone; the anti-communication interference performance tag refers to the maximum intensity of communication interference that each drone can withstand. By comparing the anti-communication interference performance label of each UAV with the average communication interference intensity corresponding to the entire range of each patrol path, a patrol task allocation decision is formed; the patrol task allocation decision includes assigning each UAV a patrol path to complete the patrol.

[0025] The beneficial effects of the above embodiments are that the drone swarm records process data of its anti-communication interference operations during historical patrol and monitoring. This anti-communication interference operation process data may include, but is not limited to, operations such as increasing the strength of its own wireless communication signal transmission in response to the current level of communication interference. Based on this process data, the maximum communication interference intensity that each drone can withstand is generated, quantifying the anti-communication interference performance of each drone. Furthermore, by comparing the anti-communication interference performance label of each drone with the average communication interference intensity corresponding to the entire patrol path, a patrol task allocation decision is formed, including the patrol path assigned to each drone. This ensures that each drone effectively resists the communication interference of its assigned patrol path during patrol and monitoring, maintaining normal and stable wireless communication and ensuring the safe flight of the drones.

[0026] In another embodiment, spatial cognitive learning is performed on the patrol task allocation decision to obtain the flight state change plan for each UAV during the patrol task; based on the UAV's flight constraints, navigation neural network analysis is performed on the flight state change plan to change the UAV's motion parameters, including: Spatial trajectory morphology cognition learning is performed on the patrol path assigned to each UAV within the patrol task allocation decision to obtain flight position change data of each UAV during the patrol task execution along the corresponding patrol path; based on the flight position change data and the external dimensions of the UAV, a flight state change plan is obtained. Extract the maximum change in flight direction angle and maximum change in flight speed per unit time from the flight constraints of the UAV; perform navigation neural network analysis on the maximum change in flight direction angle, maximum change in flight speed, and flight state change planning to change the flight direction and / or flight speed of the UAV in a future preset time interval.

[0027] The beneficial effects of the above embodiments are as follows: After each UAV is assigned a corresponding patrol path, each UAV will fly along the corresponding patrol path. To ensure that the UAV flies accurately and stably along the patrol path, the spatial trajectory morphology recognition learning is first performed on the patrol path assigned to each UAV within the patrol task allocation decision, obtaining the flight position change data of each UAV during the patrol task execution along the corresponding patrol path. Combined with the UAV's own external dimensions (such as the length, width, wing or propeller length, etc.), a flight state change plan is obtained. The flight state change plan may include, but is not limited to, the change plan of the UAV's flight direction and / or flight speed at the corresponding position or segment within the patrol path. Furthermore, the maximum change in flight direction angle and the maximum change in flight speed of the UAV per unit time are extracted from the UAV's flight constraints. Then, navigation neural network analysis is performed on the above maximum change in flight direction angle, maximum change in flight speed, and flight state change plan to change the UAV's flight direction and / or flight speed in the future preset time interval, ensuring that the UAV flies accurately along the patrol path without deviation, adapting to the uncertainty and variability of the external environment during the patrol process, and improving the robustness and accuracy of UAV flight control.

[0028] In another embodiment, determining whether there are any abnormalities in the flight of the UAVs based on the flight position change data of each UAV during its patrol mission along the corresponding patrol path includes: Retrieve the actual location coordinates of the drone at each data acquisition moment; Retrieve the expected location coordinates of the drone at each data acquisition moment; The drone flight error parameters are obtained based on the actual and expected position coordinates of the drone at each acquisition time. The UAV flight error parameters are obtained using the following formula: , Where W represents the UAV flight error parameters; n represents the total number of position data acquisitions; x i x represents the actual position coordinates of the UAV at the i-th data acquisition time; ei This represents the expected location coordinates of the UAV at the i-th data acquisition time. The drone flight error parameters are compared with preset error parameter thresholds; If the flight error parameters of the UAV do not exceed the preset error parameter threshold, the flight anomaly determination will not be initiated. When the flight error parameters of the UAV exceed the preset error parameter threshold, the actual change in flight direction angle and the actual change in flight speed per unit time of the UAV are retrieved. Flight anomaly parameters are obtained by combining the actual flight direction angle change and actual flight speed change per unit time of the UAV with the UAV flight error parameters. The flight anomaly parameters are obtained using the following formula: , Where E represents flight anomaly parameters; W represents UAV flight error parameters; Δθ represents the change in actual flight direction angle per unit time; Δθ max Δv represents the maximum permissible change in the drone's heading angle; Δv represents the change in actual flight speed per unit time; Δv max Indicates the maximum permissible speed change for the drone; k 01 and k 02 These represent the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed, respectively, and the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed are obtained by the following formulas: , Where μ represents the preset task priority factor, with a value range of 0-1; η represents the normalized direction control accuracy coefficient, determined by the IMU sensor noise variance, and η=1 / (1+g 2 Meanwhile, g represents the normalized IMU sensor noise variance; θ g This represents the maximum permissible change in the drone's heading angle after normalization; λ represents the near-range attenuation coefficient, which defaults to 0.01 / m; σ represents the standard deviation of the current wind direction angle after normalization; v a The normalized airspeed is represented by d; d represents the distance from the current position to the next waypoint. , Among them, T e The current motor temperature of the drone is indicated by s; the normalized average response time of the drone's flight direction control is indicated by h; and the difference between the current altitude and the preset altitude is indicated by h. max Indicates the maximum allowable altitude deviation for the task; v g T represents the maximum permissible speed change of the drone after normalization. g This indicates the remaining battery life after normalization. The flight anomaly parameters are compared with preset first anomaly parameter thresholds and second anomaly parameter thresholds; When the flight anomaly parameter does not exceed the preset first anomaly parameter threshold, it is determined to be a low-level anomaly, and a low-level anomaly warning is issued. When the flight anomaly parameter exceeds the preset first anomaly parameter threshold, but does not exceed the preset second anomaly parameter threshold, it is determined to be a moderate anomaly and a moderate anomaly warning is issued. When the flight anomaly parameter exceeds the preset second anomaly parameter threshold, it is determined to be an altitude anomaly, and an altitude anomaly warning is issued.

[0029] The beneficial effects of the above embodiments are as follows: The actual and expected position coordinates of the UAV at each collection moment along the patrol path are retrieved. A three-dimensional data model including position deviation, rate of change of azimuth angle, and rate of change of velocity is constructed to form a full-time-domain flight state mapping. Based on spatial trajectory morphology cognitive learning, flight error parameters and anomaly parameters are calculated in real time, constructing a full-link technology system from data acquisition to hierarchical judgment. A calculation formula for flight anomaly parameters integrating position error, azimuth angle constraint, and velocity constraint is proposed. By introducing a task priority factor, sensor noise compensation coefficient, and environmental disturbance adaptive term, the limitations of traditional single-parameter threshold detection are overcome. The dynamic sensitivity coefficient model achieves cross-domain coupling of task characteristics, hardware status, and environmental parameters for the first time, constructing a nonlinear adaptive adjustment mechanism to solve the problem of dynamic balance of control parameters in complex scenarios.

[0030] Flight error parameters are used to quantitatively assess trajectory deviation through Euclidean distance accumulation. Combined with normalized processing of changes in azimuth angle and velocity, a three-dimensional anomaly feature vector is formed, encompassing spatial position, motion attitude, and dynamic characteristics. This method effectively captures complex anomalies in UAVs, such as over-turning at waypoints, sudden velocity changes, and long-term trajectory deviations, improving anomaly identification accuracy by 40% compared to traditional single-parameter detection. The sensitivity coefficient calculation formula incorporates multi-source parameters such as task priority (μ), sensor accuracy (η), environmental interference (σ, va), and energy state (Tg), achieving dynamic calibration through nonlinear forms such as exponential functions and S-curves. For example, in low-altitude complex terrain scenarios, the azimuth angle sensitivity coefficient is automatically increased to enhance waypoint tracking accuracy; in low-battery scenarios, the velocity sensitivity coefficient is dynamically reduced to optimize energy allocation, achieving self-optimization of the control strategy throughout the entire mission cycle.

[0031] A dual-threshold-based anomaly classification system is established. By setting a first anomaly parameter threshold (low-level anomaly) and a second anomaly parameter threshold (moderate anomaly), the severity of anomalies can be accurately defined. Low-level anomalies trigger sensor verification and control algorithm fault tolerance, moderate anomalies initiate redundant navigation system switching, and altitude anomalies directly activate emergency return-to-base commands, forming a gradient response mechanism from early warning to response, ensuring optimal decision-making under different risk levels. For complex environments such as strong winds, electromagnetic interference, and terrain undulations, an environmental disturbance acceleration compensation term (a_env) and an altitude deviation correction factor (Δh / h_max) are designed to adjust the sensitivity coefficient weight allocation in real time. Experimental data show that in strong wind environments with wind speeds ≥15m / s, the direction angle anomaly detection delay is reduced from 2.5s in the traditional method to 0.8s, and the speed fluctuation amplitude is reduced by 35%. In low-altitude canyon scenarios, the position deviation detection accuracy is improved by 28%, effectively solving the problems of misjudgment and missed judgment in complex environments.

[0032] By introducing look-ahead adjustment of the near-distance attenuation coefficient (λ) and waypoint distance (d), pre-compensation control for nearby turning points is achieved, improving waypoint turning accuracy by 25% (heading angle error ≤ 1.2°). Combined with deep coupling between navigation neural network analysis and hardware constraint parameters (Δθ_max, Δv_max), the calculation delay for abnormal parameters is controlled within 50ms, meeting the requirements of real-time tasks such as inspection and mapping.

[0033] The task priority factor (μ) supports dynamic switching between multiple task modes such as surveying, inspection, and logistics. For example, in surveying tasks, the orientation angle sensitivity coefficient is increased by 30% to ensure image acquisition accuracy, while in high-speed inspection tasks, the speed sensitivity coefficient is increased by 40% to optimize path efficiency. Simultaneously, through status feedback of remaining battery life (T_bat) and motor temperature (T_e), deep integration of energy management and hardware protection is achieved, improving energy utilization by 18% throughout the mission cycle and increasing hardware fault warning accuracy to 92%. A standardized threshold library based on UAV model, task type, and environmental level is established, supporting rapid parameter configuration and cross-platform portability. This is achieved by using IMU sensor noise variance (g... 2 Hardware parameters such as maximum permissible height deviation (h_max) are deeply embedded in the algorithm model, forming a highly interpretable basis for anomaly judgment, significantly reducing the parameter calibration cost in engineering applications, and promoting the large-scale deployment of technical solutions in fields such as UAV swarm monitoring and power line inspection.

[0034] Please see Figure 2 As shown, one embodiment of this application provides a brain-inspired control system for drone patrol. This brain-inspired control system for drone patrol includes: The subspace determination module is used to perceive the three-dimensional structural data of the space where the flight is located, perform visual neuron analysis on the three-dimensional structural data, and obtain all accessible patrol subspaces in the space where the flight is located. The path identification module is used to identify the appropriate patrol path and its path characteristics based on the potential interference layout of the passable patrol subspace. The performance tag generation module is used to retrieve the historical flight memory data of the drone cluster and generate the performance tags for each drone in the cluster. The task allocation decision generation module is used to compare the work performance tags and the path characteristics of the patrol path to form a patrol task allocation decision. The planning and determination module is used to perform spatial cognitive learning on patrol mission allocation decisions to obtain flight status change plans for each UAV during patrol missions. The motion change module is used to perform navigation neuron analysis on the flight state change plan based on the flight constraints of the UAV, thereby changing the motion parameters of the UAV.

[0035] The beneficial effects of the above embodiments are as follows: The brain-like control system for UAV patrol performs visual neuron analysis on the three-dimensional structural data of the flight space to obtain all navigable patrol subspaces; based on the potential interference layout of the navigable patrol subspaces, it calibrates suitable patrol paths and their path characteristics; it retrieves historical flight memory data of the UAV swarm to generate individual performance labels for each UAV in the swarm; it compares the performance labels with the path characteristics of the patrol paths to form patrol task allocation decisions and obtains flight state change plans for each UAV during patrol tasks; and it changes the UAV's motion parameters according to the UAV's flight constraints and flight state change plans. By utilizing brain-like perception to analyze three-dimensional structural data, calibrating several patrol paths, and combining UAV performance to perform adaptive patrol task allocation, it adapts to the uncertainty and variability of the external environment during patrols, improving the robustness and accuracy of UAV flight control.

[0036] In another embodiment, the subspace determination module is used to perceive the three-dimensional structural data of the space where the flight is located, perform visual neural network analysis on the three-dimensional structural data, and obtain all accessible patrol subspaces of the space where the flight is located, including: Perceive the image data of objects in the three-dimensional domain of the space where the flight is taking place, and extract contour feature data and depth feature data from the object image data; Based on contour feature data and depth feature data, generate three-dimensional structural data for each object in the flight space; wherein, the three-dimensional structural data includes contour-depth fusion data of each object in the three-dimensional domain. Visual neural network analysis is performed on the 3D structural data of all objects to obtain the layout of the empty space between all objects; several connected subspaces are identified from the empty space layout as passable patrol subspaces. The path identification module is used to identify suitable patrol paths and their characteristics based on the potential interference layout of the passable patrol subspace, including: Based on the three-dimensional size data of objects within the accessible patrol subspace, the potential interference layout is determined; whereby, the potential interference layout refers to the spatial distribution of UAV communication interference intensity caused by object obstruction within the accessible patrol subspace. Based on the spatial distribution of UAV communication interference intensity included in the potential interference layout, several suitable patrol paths and their path characteristics are identified within the passable patrol subspace. Among them, the patrol path refers to a one-way patrol path within the passable patrol subspace where the average communication interference intensity does not exceed a preset interference intensity threshold; the path characteristics refer to the average communication interference intensity corresponding to the entire range of the patrol path.

[0037] The beneficial effects of the above embodiments are that the three-dimensional spatial structure of objects such as buildings, mountains, and trees in the space where the UAV swarm flies directly determines the flight space of the UAVs. In order for the UAVs to fly continuously for a certain distance in the above space, it is necessary to mark the blank range of the above space and determine the subspaces that the UAVs can fly through. Specifically, the binocular camera sensor of one of the UAVs in the swarm (such as the UAV at the front of the swarm) first collects object image data in the three-dimensional domain of the flight space. The object image data can be, but is not limited to, the overall three-dimensional image data of the entire flight space or several local three-dimensional image data of different parts of the global range. Then, the object image data is subjected to contour and depth recognition to obtain contour feature data and depth feature data. The contour feature data includes the contour features of each object in the flight space, and the depth feature data includes the depth value of each object in the flight space. Then, taking each object in the flight space as a reference, the contour feature data and depth feature data corresponding to the same object are fused to generate contour-depth fusion data of each object in the three-dimensional domain, so as to realize the comprehensive three-dimensional characteristic calibration of all objects in the flight space.

[0038] By using a brain-like neural network model to perform visual neuron analysis on the 3D structural data of all objects, the layout of the empty space between all objects is obtained. Then, the connected subspaces that meet the preset connectivity length condition (such as exceeding the preset connectivity length threshold) are designated as passable patrol subspaces (i.e., subspaces that allow UAVs to fly through). This allows for the accurate marking of the feasible movement space of UAVs in the flight space, providing a basis for the subsequent assignment of patrol tasks to each UAV.

[0039] Furthermore, drones rely on wireless communication with base stations or the main drone to maintain continuous and stable flight. However, obstructions from buildings and / or trees weaken the propagation strength of wireless communication signals, interfering with the drone's wireless communication and affecting its flight controllability and safety. Different drones have varying anti-interference communication capabilities, and the degree of performance degradation varies even when faced with the same obstruction from buildings and / or trees. Therefore, to enable drones with high anti-interference communication capabilities to patrol and monitor in environments with strong external communication interference, it is necessary to first categorize and label the drone communication interference situation within the accessible patrol subspace. Specifically, based on the three-dimensional dimensions of objects within the accessible patrol subspace, the spatial distribution of drone communication interference intensity caused by object obstruction (i.e., the spatial distribution of drone communication interference magnitude) is determined. The greater the drone communication interference intensity, the worse the corresponding drone's wireless communication performance. Based on the spatial distribution of UAV communication interference intensity within the potential interference layout, several suitable patrol paths and their characteristics are identified within the passable patrol subspace. This enables accurate differentiation and identification of the average communication interference intensity experienced by all patrol paths within the passable patrol subspace, providing a basis for subsequent patrol task allocation decisions.

[0040] In another embodiment, the performance tag generation module is used to retrieve historical flight memory data of the drone swarm to generate individual performance tags for each drone in the swarm, including: Retrieve historical flight memory data of the drone swarm; the historical flight memory data includes the anti-communication interference operation data of each drone in the drone swarm during historical flights; based on the historical flight memory data, generate anti-communication interference performance tags for each drone; the anti-communication interference performance tag refers to the maximum intensity of communication interference that each drone can withstand. The task allocation decision generation module compares the work performance labels and the path characteristics of the patrol routes to form patrol task allocation decisions, including: By comparing the anti-communication interference performance label of each UAV with the average communication interference intensity corresponding to the entire range of each patrol path, a patrol task allocation decision is formed; the patrol task allocation decision includes assigning each UAV a patrol path to complete the patrol.

[0041] The beneficial effects of the above embodiments are that the drone swarm records process data of its anti-communication interference operations during historical patrol and monitoring. This anti-communication interference operation process data may include, but is not limited to, operations such as increasing the strength of its own wireless communication signal transmission in response to the current level of communication interference. Based on this process data, the maximum communication interference intensity that each drone can withstand is generated, quantifying the anti-communication interference performance of each drone. Furthermore, by comparing the anti-communication interference performance label of each drone with the average communication interference intensity corresponding to the entire patrol path, a patrol task allocation decision is formed, including the patrol path assigned to each drone. This ensures that each drone effectively resists the communication interference of its assigned patrol path during patrol and monitoring, maintaining normal and stable wireless communication and ensuring the safe flight of the drones.

[0042] In another embodiment, the planning and determination module is used to perform spatial cognitive learning on patrol task allocation decisions to obtain a flight state change plan for each UAV during its patrol task, including: Spatial trajectory morphology cognition learning is performed on the patrol path assigned to each UAV within the patrol task allocation decision to obtain flight position change data of each UAV during the patrol task execution along the corresponding patrol path; based on the flight position change data and the external dimensions of the UAV, a flight state change plan is obtained. The motion change module is used to perform navigation neural network analysis on the flight state change plan based on the UAV's flight constraints, thereby changing the UAV's motion parameters, including: Extract the maximum change in flight direction angle and maximum change in flight speed per unit time from the flight constraints of the UAV; perform navigation neural network analysis on the maximum change in flight direction angle, maximum change in flight speed, and flight state change planning to change the flight direction and / or flight speed of the UAV in a future preset time interval.

[0043] The beneficial effects of the above embodiments are as follows: After each UAV is assigned a corresponding patrol path, each UAV will fly along the corresponding patrol path. To ensure that the UAV flies accurately and stably along the patrol path, the spatial trajectory morphology recognition learning is first performed on the patrol path assigned to each UAV within the patrol task allocation decision, obtaining the flight position change data of each UAV during the patrol task execution along the corresponding patrol path. Combined with the UAV's own external dimensions (such as the length, width, wing or propeller length, etc.), a flight state change plan is obtained. The flight state change plan may include, but is not limited to, the change plan of the UAV's flight direction and / or flight speed at the corresponding position or segment within the patrol path. Furthermore, the maximum change in flight direction angle and the maximum change in flight speed of the UAV per unit time are extracted from the UAV's flight constraints. Then, navigation neural network analysis is performed on the above maximum change in flight direction angle, maximum change in flight speed, and flight state change plan to change the UAV's flight direction and / or flight speed in the future preset time interval, ensuring that the UAV flies accurately along the patrol path without deviation, adapting to the uncertainty and variability of the external environment during the patrol process, and improving the robustness and accuracy of UAV flight control.

[0044] In another embodiment, determining whether there are any abnormalities in the flight of the UAVs based on the flight position change data of each UAV during its patrol mission along the corresponding patrol path includes: Retrieve the actual location coordinates of the drone at each data acquisition moment; Retrieve the expected location coordinates of the drone at each data acquisition moment; The drone flight error parameters are obtained based on the actual and expected position coordinates of the drone at each acquisition time. The UAV flight error parameters are obtained using the following formula:

[0045] Where W represents the UAV flight error parameters; n represents the total number of position data acquisitions; x i x represents the actual position coordinates of the UAV at the i-th data acquisition time; ei This represents the expected location coordinates of the UAV at the i-th data acquisition time. The drone flight error parameters are compared with preset error parameter thresholds; If the flight error parameters of the UAV do not exceed the preset error parameter threshold, the flight anomaly determination will not be initiated. When the flight error parameters of the UAV exceed the preset error parameter threshold, the actual change in flight direction angle and the actual change in flight speed per unit time of the UAV are retrieved. Flight anomaly parameters are obtained by combining the actual flight direction angle change and actual flight speed change per unit time of the UAV with the UAV flight error parameters. The flight anomaly parameters are obtained using the following formula: , Where E represents flight anomaly parameters; W represents UAV flight error parameters; Δθ represents the change in actual flight direction angle per unit time; Δθ max Δv represents the maximum permissible change in the drone's heading angle; Δv represents the change in actual flight speed per unit time; Δv max Indicates the maximum permissible speed change for the drone; k 01 and k 02 These represent the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed, respectively, and the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed are obtained by the following formulas: , Where μ represents the preset task priority factor, with a value range of 0-1; η represents the normalized direction control accuracy coefficient, determined by the IMU sensor noise variance, and η=1 / (1+g 2 Meanwhile, g represents the normalized IMU sensor noise variance; θ g This represents the maximum permissible change in the drone's heading angle after normalization; λ represents the near-range attenuation coefficient, which defaults to 0.01 / m; σ represents the standard deviation of the current wind direction angle after normalization; v a The normalized airspeed is represented by d; d represents the distance from the current position to the next waypoint. , Among them, T e The current motor temperature of the drone is indicated by s; the normalized average response time of the drone's flight direction control is indicated by h; and the difference between the current altitude and the preset altitude is indicated by h. max Indicates the maximum allowable altitude deviation for the task; v g T represents the maximum permissible speed change of the drone after normalization. g This indicates the remaining battery life after normalization. The flight anomaly parameters are compared with preset first anomaly parameter thresholds and second anomaly parameter thresholds; When the flight anomaly parameter does not exceed the preset first anomaly parameter threshold, it is determined to be a low-level anomaly, and a low-level anomaly warning is issued. When the flight anomaly parameter exceeds the preset first anomaly parameter threshold, but does not exceed the preset second anomaly parameter threshold, it is determined to be a moderate anomaly and a moderate anomaly warning is issued. When the flight anomaly parameter exceeds the preset second anomaly parameter threshold, it is determined to be an altitude anomaly, and an altitude anomaly warning is issued.

[0046] The beneficial effects of the above embodiments are as follows: The actual and expected position coordinates of the UAV at each collection moment along the patrol path are retrieved. A three-dimensional data model including position deviation, rate of change of azimuth angle, and rate of change of velocity is constructed to form a full-time-domain flight state mapping. Based on spatial trajectory morphology cognitive learning, flight error parameters and anomaly parameters are calculated in real time, constructing a full-link technology system from data acquisition to hierarchical judgment. A calculation formula for flight anomaly parameters integrating position error, azimuth angle constraint, and velocity constraint is proposed. By introducing a task priority factor, sensor noise compensation coefficient, and environmental disturbance adaptive term, the limitations of traditional single-parameter threshold detection are overcome. The dynamic sensitivity coefficient model achieves cross-domain coupling of task characteristics, hardware status, and environmental parameters for the first time, constructing a nonlinear adaptive adjustment mechanism to solve the problem of dynamic balance of control parameters in complex scenarios.

[0047] Flight error parameters are used to quantitatively assess trajectory deviation through Euclidean distance accumulation. Combined with normalized processing of changes in azimuth angle and velocity, a three-dimensional anomaly feature vector is formed, encompassing spatial position, motion attitude, and dynamic characteristics. This method effectively captures complex anomalies in UAVs, such as over-turning at waypoints, sudden velocity changes, and long-term trajectory deviations, improving anomaly identification accuracy by 40% compared to traditional single-parameter detection. The sensitivity coefficient calculation formula incorporates multi-source parameters such as task priority (μ), sensor accuracy (η), environmental interference (σ, va), and energy state (Tg), achieving dynamic calibration through nonlinear forms such as exponential functions and S-curves. For example, in low-altitude complex terrain scenarios, the azimuth angle sensitivity coefficient is automatically increased to enhance waypoint tracking accuracy; in low-battery scenarios, the velocity sensitivity coefficient is dynamically reduced to optimize energy allocation, achieving self-optimization of the control strategy throughout the entire mission cycle.

[0048] A dual-threshold-based anomaly classification system is established. By setting a first anomaly parameter threshold (low-level anomaly) and a second anomaly parameter threshold (moderate anomaly), the severity of anomalies can be accurately defined. Low-level anomalies trigger sensor verification and control algorithm fault tolerance, moderate anomalies initiate redundant navigation system switching, and altitude anomalies directly activate emergency return-to-base commands, forming a gradient response mechanism from early warning to response, ensuring optimal decision-making under different risk levels. For complex environments such as strong winds, electromagnetic interference, and terrain undulations, an environmental disturbance acceleration compensation term (a_env) and an altitude deviation correction factor (Δh / h_max) are designed to adjust the sensitivity coefficient weight allocation in real time. Experimental data show that in strong wind environments with wind speeds ≥15m / s, the direction angle anomaly detection delay is reduced from 2.5s in the traditional method to 0.8s, and the speed fluctuation amplitude is reduced by 35%. In low-altitude canyon scenarios, the position deviation detection accuracy is improved by 28%, effectively solving the problems of misjudgment and missed judgment in complex environments.

[0049] By introducing look-ahead adjustment of the near-distance attenuation coefficient (λ) and waypoint distance (d), pre-compensation control for nearby turning points is achieved, improving waypoint turning accuracy by 25% (heading angle error ≤ 1.2°). Combined with deep coupling between navigation neural network analysis and hardware constraint parameters (Δθ_max, Δv_max), the calculation delay for abnormal parameters is controlled within 50ms, meeting the requirements of real-time tasks such as inspection and mapping.

[0050] The task priority factor (μ) supports dynamic switching between multiple task modes such as surveying, inspection, and logistics. For example, in surveying tasks, the orientation angle sensitivity coefficient is increased by 30% to ensure image acquisition accuracy, while in high-speed inspection tasks, the speed sensitivity coefficient is increased by 40% to optimize path efficiency. Simultaneously, through status feedback of remaining battery life (T_bat) and motor temperature (T_e), deep integration of energy management and hardware protection is achieved, improving energy utilization by 18% throughout the mission cycle and increasing hardware fault warning accuracy to 92%. A standardized threshold library based on UAV model, task type, and environmental level is established, supporting rapid parameter configuration and cross-platform portability. This is achieved by using IMU sensor noise variance (g... 2 Hardware parameters such as maximum permissible height deviation (h_max) are deeply embedded in the algorithm model, forming a highly interpretable basis for anomaly judgment, significantly reducing the parameter calibration cost in engineering applications, and promoting the large-scale deployment of technical solutions in fields such as UAV swarm monitoring and power line inspection.

[0051] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. This disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims. Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. A brain-like control method for unmanned aerial vehicle (UAV) patrol, characterized in that, include: The system senses the three-dimensional structural data of the space in which the flight is located, performs visual neural network analysis on the three-dimensional structural data, and obtains all accessible patrol subspaces in the space in which the flight is located; based on the potential interference layout of the accessible patrol subspaces, it calibrates the appropriate patrol path and its path characteristics. Historical flight memory data of the drone swarm is retrieved to generate performance tags for each drone in the swarm; the performance tags are compared with the path characteristics of the patrol path to form a patrol task allocation decision. Spatial cognitive learning is performed on the patrol task allocation decision to obtain the flight state change plan for each UAV during the patrol task; based on the flight constraints of the UAV, navigation neural network analysis is performed on the flight state change plan to change the motion parameters of the UAV.

2. The brain-like control method for unmanned aerial vehicle patrol as described in claim 1, characterized in that: The system senses the three-dimensional structural data of the flight space, performs visual neural network analysis on the three-dimensional structural data to obtain all accessible patrol subspaces within the flight space, and, based on the potential interference layout of the accessible patrol subspaces, calibrates suitable patrol paths and their characteristics, including: The system senses the image data of objects in the three-dimensional domain of the space where the flight is taking place, and extracts contour feature data and depth feature data from the object image data. Based on the contour feature data and the depth feature data, three-dimensional structural data of each object in the flight space is generated; wherein, the three-dimensional structural data includes the contour-depth fusion data of each object in the three-dimensional domain; Visual neural network analysis is performed on the three-dimensional structural data of all objects to obtain the layout of the empty space between all objects; several connected subspaces are identified from the empty space layout as passable patrol subspaces. Based on the three-dimensional size data of objects within the accessible patrol subspace, a potential interference layout is determined; wherein, the potential interference layout refers to the spatial distribution of UAV communication interference intensity caused by object obstruction within the accessible patrol subspace. Based on the spatial distribution of UAV communication interference intensity included in the potential interference layout, several patrol paths and their path characteristics are identified within the passable patrol subspace; wherein, the patrol path refers to a one-way patrol path within the passable patrol subspace where the average communication interference intensity does not exceed a preset interference intensity threshold; and the path characteristics refer to the average communication interference intensity corresponding to the entire range of the patrol path.

3. The brain-like control method for unmanned aerial vehicle patrol as described in claim 2, characterized in that: Retrieve historical flight memory data of the drone swarm to generate individual performance tags for each drone in the swarm; compare the performance tags with the path characteristics of the patrol path to form a patrol task allocation decision, including: Retrieve historical flight memory data of the drone cluster; wherein, the historical flight memory data includes the anti-communication interference operation data of each drone under the drone cluster during historical flights; based on the historical flight memory data, generate an anti-communication interference performance tag for each drone; wherein, the anti-communication interference performance tag refers to the maximum intensity of communication interference that each drone can withstand; By comparing the anti-communication interference performance label of each UAV with the average communication interference intensity corresponding to the entire range of each patrol path, a patrol task allocation decision is formed; wherein, the patrol task allocation decision includes assigning each UAV a patrol path to complete the patrol.

4. The brain-like control method for unmanned aerial vehicle patrol as described in claim 3, characterized in that: Spatial cognitive learning is performed on the patrol task allocation decision to obtain the flight state change plan for each UAV during the patrol task; based on the flight constraints of the UAV, navigation neural network analysis is performed on the flight state change plan to change the motion parameters of the UAV, including: Spatial trajectory morphology cognitive learning is performed on the patrol path assigned to each UAV within the patrol task allocation decision to obtain flight position change data of each UAV during the patrol task execution along the corresponding patrol path; based on the flight position change data and the external dimensions of the UAV, a flight state change plan is obtained. Extract the maximum change in flight direction angle and the maximum change in flight speed of the UAV per unit time from the flight constraints of the UAV; perform navigation neural network analysis on the maximum change in flight direction angle, the maximum change in flight speed, and the flight state change plan, so as to change the flight direction and / or flight speed of the UAV in a future preset time interval.

5. The brain-like control method for unmanned aerial vehicle patrol as described in claim 4, characterized in that: Based on the flight position change data of each drone during its patrol mission along the corresponding patrol path, determine whether there are any anomalies in the drone's flight, including: Retrieve the actual location coordinates of the drone at each data acquisition moment; Retrieve the expected location coordinates of the drone at each data acquisition moment; The drone flight error parameters are obtained based on the actual and expected position coordinates of the drone at each acquisition time. The UAV flight error parameters are obtained using the following formula: ， Where W represents the UAV flight error parameters; n represents the total number of position data acquisitions; x i x represents the actual position coordinates of the UAV at the i-th data acquisition time; ei This represents the expected location coordinates of the UAV at the i-th data acquisition time. The drone flight error parameters are compared with preset error parameter thresholds; If the flight error parameters of the UAV do not exceed the preset error parameter threshold, the flight anomaly determination will not be initiated. When the flight error parameters of the UAV exceed the preset error parameter threshold, the actual change in flight direction angle and the actual change in flight speed per unit time of the UAV are retrieved. Flight anomaly parameters are obtained by combining the actual flight direction angle change and actual flight speed change per unit time of the UAV with the UAV flight error parameters. The flight anomaly parameters are obtained using the following formula: ， Where E represents flight anomaly parameters; W represents UAV flight error parameters; Δθ represents the change in actual flight direction angle per unit time; Δθ max Δv represents the maximum permissible change in the drone's heading angle; Δv represents the change in actual flight speed per unit time; Δv max Indicates the maximum permissible speed change for the drone; k 01 and k 02 These represent the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed, respectively, and the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed are obtained by the following formulas: ， Where μ represents the preset task priority factor, with a value range of 0-1; η represents the normalized direction control accuracy coefficient, determined by the IMU sensor noise variance, and η=1 / (1+g 2 Meanwhile, g represents the normalized IMU sensor noise variance; θ g This represents the maximum permissible change in the drone's heading angle after normalization; λ represents the near-range attenuation coefficient, which defaults to 0.01 / m; σ represents the standard deviation of the current wind direction angle after normalization; v a The normalized airspeed is represented by d; d represents the distance from the current position to the next waypoint. ， Among them, T e The current motor temperature of the drone is indicated by s; the normalized average response time of the drone's flight direction control is indicated by h; and the difference between the current altitude and the preset altitude is indicated by h. max Indicates the maximum allowable altitude deviation for the task; v g T represents the maximum permissible speed change of the drone after normalization. g This indicates the remaining battery life after normalization. The flight anomaly parameters are compared with preset first anomaly parameter thresholds and second anomaly parameter thresholds; When the flight anomaly parameter does not exceed the preset first anomaly parameter threshold, it is determined to be a low-level anomaly, and a low-level anomaly warning is issued. When the flight anomaly parameter exceeds the preset first anomaly parameter threshold, but does not exceed the preset second anomaly parameter threshold, it is determined to be a moderate anomaly and a moderate anomaly warning is issued. When the flight anomaly parameter exceeds the preset second anomaly parameter threshold, it is determined to be an altitude anomaly, and an altitude anomaly warning is issued.

6. A brain-like control system for unmanned aerial vehicle (UAV) patrols, characterized in that, include: The subspace determination module is used to sense the three-dimensional structural data of the space where the flight is located, perform visual neuron analysis on the three-dimensional structural data, and obtain all passable patrol subspaces of the space where the flight is located. The path identification module is used to identify the appropriate patrol path and its path characteristics based on the potential interference layout of the accessible patrol subspace. The performance tag generation module is used to retrieve historical flight memory data of the drone cluster and generate the performance tags of each drone in the drone cluster. The task allocation decision generation module is used to compare the work performance tags and the path characteristics of the patrol path to form a patrol task allocation decision. The planning and determination module is used to perform spatial cognitive learning on the patrol task allocation decision to obtain the flight status change plan for each UAV during the patrol task. The motion change module is used to perform navigation neuron analysis on the flight state change plan based on the flight constraints of the UAV, thereby changing the motion parameters of the UAV.

7. The brain-like control system for unmanned aerial vehicle patrol as described in claim 6, characterized in that: The subspace determination module is used to perceive the three-dimensional structural data of the flight space, perform visual neural network analysis on the three-dimensional structural data, and obtain all accessible patrol subspaces of the flight space, including: The system senses the image data of objects in the three-dimensional domain of the space where the flight is taking place, and extracts contour feature data and depth feature data from the object image data. Based on the contour feature data and the depth feature data, three-dimensional structural data of each object in the flight space is generated; wherein, the three-dimensional structural data includes the contour-depth fusion data of each object in the three-dimensional domain; Visual neural network analysis is performed on the three-dimensional structural data of all objects to obtain the layout of the empty space between all objects; several connected subspaces are identified from the empty space layout as passable patrol subspaces. The path identification module is used to identify suitable patrol paths and their path characteristics based on the potential interference layout of the passable patrol subspace, including: Based on the three-dimensional size data of objects within the accessible patrol subspace, a potential interference layout is determined; wherein, the potential interference layout refers to the spatial distribution of UAV communication interference intensity caused by object obstruction within the accessible patrol subspace. Based on the spatial distribution of UAV communication interference intensity included in the potential interference layout, several patrol paths and their path characteristics are identified within the passable patrol subspace; wherein, the patrol path refers to a one-way patrol path within the passable patrol subspace where the average communication interference intensity does not exceed a preset interference intensity threshold; and the path characteristics refer to the average communication interference intensity corresponding to the entire range of the patrol path.

8. The brain-like control system for unmanned aerial vehicle patrol as described in claim 7, characterized in that: The performance tag generation module is used to retrieve historical flight memory data of the drone cluster to generate performance tags for each drone in the cluster, including: Retrieve historical flight memory data of the drone cluster; wherein, the historical flight memory data includes the anti-communication interference operation data of each drone under the drone cluster during historical flights; based on the historical flight memory data, generate an anti-communication interference performance tag for each drone; wherein, the anti-communication interference performance tag refers to the maximum intensity of communication interference that each drone can withstand; The task allocation decision generation module is used to compare the work performance tags and the path characteristics of the patrol path to form a patrol task allocation decision, including: By comparing the anti-communication interference performance label of each UAV with the average communication interference intensity corresponding to the entire range of each patrol path, a patrol task allocation decision is formed; wherein, the patrol task allocation decision includes assigning each UAV a patrol path to complete the patrol.

9. The brain-like control system for unmanned aerial vehicle patrol as described in claim 8, characterized in that: The planning and determination module is used to perform spatial cognitive learning on the patrol task allocation decision to obtain the flight state change plan for each UAV during the patrol task, including: Spatial trajectory morphology cognitive learning is performed on the patrol path assigned to each UAV within the patrol task allocation decision to obtain flight position change data of each UAV during the patrol task execution along the corresponding patrol path; based on the flight position change data and the external dimensions of the UAV, a flight state change plan is obtained. The motion change module is used to perform navigation neuron analysis on the flight state change plan based on the flight constraints of the UAV, thereby changing the motion parameters of the UAV, including: Extract the maximum change in flight direction angle and the maximum change in flight speed of the UAV per unit time from the flight constraints of the UAV; perform navigation neural network analysis on the maximum change in flight direction angle, the maximum change in flight speed, and the flight state change plan, so as to change the flight direction and / or flight speed of the UAV in a future preset time interval.

10. The brain-like control system for unmanned aerial vehicle patrol as described in claim 9, characterized in that: Based on the flight position change data of each drone during its patrol mission along the corresponding patrol path, determine whether there are any anomalies in the drone's flight, including: Retrieve the actual location coordinates of the drone at each data acquisition moment; Retrieve the expected location coordinates of the drone at each data acquisition moment; The drone flight error parameters are obtained based on the actual and expected position coordinates of the drone at each acquisition time. The UAV flight error parameters are obtained using the following formula: ， Where W represents the UAV flight error parameters; n represents the total number of position data acquisitions; x i x represents the actual position coordinates of the UAV at the i-th data acquisition time; ei This represents the expected location coordinates of the UAV at the i-th data acquisition time. The drone flight error parameters are compared with preset error parameter thresholds; If the flight error parameters of the UAV do not exceed the preset error parameter threshold, the flight anomaly determination will not be initiated. When the flight error parameters of the UAV exceed the preset error parameter threshold, the actual change in flight direction angle and the actual change in flight speed per unit time of the UAV are retrieved. Flight anomaly parameters are obtained by combining the actual flight direction angle change and actual flight speed change per unit time of the UAV with the UAV flight error parameters. The flight anomaly parameters are obtained using the following formula: ， Where E represents flight anomaly parameters; W represents UAV flight error parameters; Δθ represents the change in actual flight direction angle per unit time; Δθ max Δv represents the maximum permissible change in the drone's heading angle; Δv represents the change in actual flight speed per unit time; Δv max Indicates the maximum permissible speed change for the drone; k 01 and k 02 These represent the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed, respectively, and the sensitivity coefficients corresponding to the changes in flight direction angle and flight speed are obtained by the following formulas: ， Where μ represents the preset task priority factor, with a value range of 0-1; η represents the normalized direction control accuracy coefficient, determined by the IMU sensor noise variance, and η=1 / (1+g 2 Meanwhile, g represents the normalized IMU sensor noise variance; θ g This represents the maximum permissible change in the drone's heading angle after normalization; λ represents the near-range attenuation coefficient, which defaults to 0.01 / m; σ represents the standard deviation of the current wind direction angle after normalization; v a The normalized airspeed is represented by d; d represents the distance from the current position to the next waypoint. ， Among them, T e The current motor temperature of the drone is indicated by s; the normalized average response time of the drone's flight direction control is indicated by h; and the difference between the current altitude and the preset altitude is indicated by h. max Indicates the maximum allowable altitude deviation for the task; v g T represents the maximum permissible speed change of the drone after normalization. g This indicates the remaining battery life after normalization. The flight anomaly parameters are compared with preset first anomaly parameter thresholds and second anomaly parameter thresholds; When the flight anomaly parameter does not exceed the preset first anomaly parameter threshold, it is determined to be a low-level anomaly, and a low-level anomaly warning is issued. When the flight anomaly parameter exceeds the preset first anomaly parameter threshold, but does not exceed the preset second anomaly parameter threshold, it is determined to be a moderate anomaly and a moderate anomaly warning is issued. When the flight anomaly parameter exceeds the preset second anomaly parameter threshold, it is determined to be an altitude anomaly, and an altitude anomaly warning is issued.

Images