Deep Learning-Based Unmanned Aerial Vehicle Emergency Defense Control System and Method

Abstract

This invention relates to the field of unmanned aerial vehicle (UAV) control, specifically a deep learning-based UAV emergency defense control system and method. It utilizes deep neural networks to learn abnormal state data around a base area to predict potential intrusions and set UAV swarm inspection strategies, rapidly initiating UAV surveillance at the initial stage of an intrusion. It identifies local interference experienced by the UAV swarm during the inspection process, adjusts the flight layout of the UAV swarm and the control strategies for at least some UAVs, ensuring that high-performing UAVs maintain normal overall swarm flight. Based on dynamic data of the relative spatial relationship between the intruder and the UAV swarm, it adjusts the UAV flight state to achieve accurate containment and control of the intruder. Furthermore, based on the intruder's flight trajectory, it controls the UAV swarm to initiate non-contact interference against the intruder until the intruder's flight state changes. This effectively adapts to different intruder flight change scenarios, improving defense reliability and timeliness.

Description

Technical Field

[0001] This invention relates to the technical field of unmanned aerial vehicle (UAV) control, and in particular to a deep learning-based UAV emergency defense control system and method. Background Technology

[0002] Unmanned aerial vehicles (UAVs), characterized by their miniaturization, ease of operation, flexible flight, and wide inspection range, have become a primary tool for aerial surveillance and defense. Current methods of using UAVs for defense primarily focus on monitoring the flight paths of intruding objects and then controlling the UAVs to track them. This approach is inherently passive and uncertain. If the flight path of an intruding object changes significantly, or if the intruding object actively interferes with the UAVs, the UAVs will be unable to promptly and accurately change their flight trajectories, failing to achieve close-range tracking and containment control of the intruding object. This severely impacts the efficiency and accuracy of intrusion control. Therefore, adapting the flight layout of UAV swarms and proactively interfering with intruding objects throughout the entire process of containing and controlling intrusion events is crucial for improving the accuracy and timeliness of intrusion identification and for rapidly controlling intruding objects. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a deep learning-based UAV emergency defense control system and method. It utilizes deep neural networks to learn abnormal state data around the base area to predict potential intrusions and set UAV swarm inspection strategies, rapidly initiating UAV surveillance at the initial stage of an intrusion. It identifies local interference experienced by the UAV swarm during the inspection process, adjusts the flight layout of the UAV swarm and the control strategies for at least some UAVs, ensuring that high-performing UAVs maintain normal overall swarm flight. Based on dynamic data of the relative spatial relationship between the intruder and the UAV swarm, it adjusts the UAV flight status to achieve accurate containment and control of the intruder, effectively restricting the intruder's flight activities. Furthermore, based on the intruder's flight trajectory, it controls the UAV swarm to initiate non-contact interference against the intruder until the intruder's flight status changes. This effectively adapts to different intruder flight change scenarios, accurately and proactively driving away intruders, improving defense reliability and timeliness.

[0004] This invention provides a deep learning-based emergency defense control method for unmanned aerial vehicles (UAVs), comprising:

[0005] Step S1: Perform deep neural network learning on abnormal state data around the base area to predict potential intrusion events, thereby setting a patrol strategy for the drone swarm around the base area.

[0006] Step S2: Monitor and identify the interference encountered by the drone swarm during the inspection process, and adjust the flight layout of the drone swarm and the control strategy for at least some of the drones.

[0007] Step S3: Based on the dynamic data of the relative spatial relationship between the intruders around the base area and the drone cluster, adjust the flight status of the corresponding drones in the drone cluster so that the intruders and the drone cluster meet the preset positional relationship conditions.

[0008] Step S4: Based on the flight trajectory of the intruder, control at least one drone in the drone cluster to initiate non-contact interference against the intruder until the intruder's flight state changes.

[0009] In one embodiment disclosed in this application, before performing deep neural network learning on the abnormal state data surrounding the base area in step S1, the following steps are included:

[0010] Visual and radar scanning detections are performed on the area surrounding the base to obtain environmental images and radar data of the surrounding environment of the base area.

[0011] Analyze the surrounding environmental radar data to determine the flight dynamic paths of suspected foreign objects around the base area;

[0012] Based on the flight dynamic path, the surrounding environment image is locally identified to obtain the flight attitude of the suspected foreign object;

[0013] A time-domain evolution analysis is performed on the flight dynamic path and flight attitude of the suspected foreign object to obtain abnormal state data around the base area; wherein, the abnormal state data includes the flight dynamic path data and flight attitude data of the suspected foreign object approaching the base area.

[0014] In one embodiment disclosed in this application, in step S1, deep neural network learning is performed on abnormal state data around the base area to predict potential intrusion events, including:

[0015] After noise removal and reliable data screening of the flight dynamic path data and flight attitude data contained in the abnormal state data, the flight dynamic path data and flight attitude data are input into a deep neural network model for analysis and learning to obtain the flight trend of the suspected foreign object relative to the base area.

[0016] Based on the flight trends, predict the spatial distribution of potential intrusion events in the base area.

[0017] In one embodiment disclosed in this application, step S1 involves setting a patrol strategy for the drone swarm around the base area, including:

[0018] Based on the spatial distribution of potential intrusion events in the base area and the geographical topography surrounding the base area, identify the intruders and their spatial distribution of intrusion locations around the base area;

[0019] Based on the spatial distribution of the intrusion locations and the readiness status of the drone clusters within the base area, a patrol strategy for drone clusters around the base area is set; wherein, the readiness status of the drone clusters includes the parking positions and takeoff acceleration characteristics of each drone cluster within the base area; the drone cluster patrol strategy includes the patrol flight coverage area of ​​the drone clusters around the base area.

[0020] In one embodiment disclosed in this application, step S2, monitoring and identifying interference locally experienced by the drone swarm during the inspection process, includes:

[0021] Monitor the external signals received locally by each drone in the drone cluster during the inspection process, and extract the interference signal components of the external signals;

[0022] Monitor the flight trajectory and attitude drift change data of all drones under the drone cluster during the inspection process, analyze the flight trajectory and attitude drift change data, and obtain the time of the drone flight anomaly.

[0023] By comparing the time series of the interference signal components with the time of the drone flight anomaly, the drones experiencing flight instability within the drone cluster can be identified.

[0024] In one embodiment disclosed in this application, step S2, adjusting the flight layout of the drone swarm and the control strategy for at least some of the drones, includes:

[0025] Based on the distribution location of the drones experiencing flight instability within the drone swarm and the relative flight speed between the drones experiencing flight instability and other drones, the flight layout of the drone swarm is adjusted, switching the drones experiencing flight instability to the inner area of ​​the drone swarm and switching the other drones to the outer area of ​​the drone swarm.

[0026] Based on the communication control strategy between the drone experiencing flight instability and the ground base station, the drone experiencing flight instability is switched to a working state controlled by other drones, thereby enabling other drones to locate and control the drone experiencing flight instability.

[0027] In one embodiment disclosed in this application, step S3 further includes:

[0028] Obtain the predicted flight trajectories of intruders surrounding the base area and the planned flight trajectories of the entire drone swarm;

[0029] By spatially comparing the predicted flight trajectory with the planned flight trajectory, dynamic data on the relative spatial relationship between intruders around the base area and the drone cluster is obtained; wherein, the dynamic data on the relative spatial relationship includes dynamic data on the relative spatial distance and relative spatial orientation between intruders around the base area and the drone cluster.

[0030] In one embodiment disclosed in this application, in step S3, the flight status of the corresponding drones in the drone cluster is adjusted according to the dynamic data of the relative spatial relationship between the intruders around the base area and the drone cluster, so that the intruders and the drone cluster meet the preset positional relationship conditions, including:

[0031] The relative spatial distance and relative spatial orientation dynamic data between the intruders around the base area and the drone cluster are analyzed in time to predict the range of the encounter location between the drone cluster and the intruders.

[0032] Based on the range of the encounter location, the flight trajectories and / or speeds of the drones in the outer area of ​​the drone swarm are adjusted, thereby enabling the drone swarm to surround the intruder.

[0033] In one embodiment of this application, in step S4, based on the flight trajectory of the intruder, at least one drone in the drone cluster is controlled to initiate non-contact interference against the intruder until the intruder's flight state changes, including:

[0034] Based on the flight trajectory of the intruder, determine the drone that is closest to the intruder within the drone cluster;

[0035] Based on the relative orientation change characteristics between the nearest drone and the intruder, the intensity and duration of the operation of the nearest drone to launch electromagnetic interference and / or laser interference against the intruder are controlled until the intruder flies away from the base area or stops flying.

[0036] This invention also provides a deep learning-based drone emergency defense control system, comprising:

[0037] The deep learning and inspection strategy setting module is used to perform deep neural network learning on abnormal state data around the base area to predict potential intrusion events, thereby setting an inspection strategy for the drone swarm around the base area.

[0038] Interference monitoring and identification module, used to monitor and identify local interference experienced by the drone swarm during the inspection process;

[0039] The flight layout and control strategy adjustment module is used to adjust the flight layout of the UAV cluster and the control strategy for at least some of the UAVs.

[0040] The cluster flight status adjustment module is used to adjust the flight status of the corresponding drones in the drone cluster based on the dynamic data of the relative spatial relationship between the intruders around the base area and the drone cluster, so that the intruders and the drone cluster meet the preset positional relationship conditions.

[0041] The interference initiation control module is used to control at least one UAV in the UAV cluster to initiate non-contact interference against the intruder based on the flight trajectory of the intruder, until the intruder changes its flight state.

[0042] Compared to existing technologies, this deep learning-based UAV emergency defense control system and method utilizes deep neural networks to learn abnormal state data around the base area to predict potential intrusions and set UAV swarm inspection strategies. It rapidly initiates UAV surveillance at the initial stage of an intrusion; identifies local interference experienced by the UAV swarm during inspection, adjusts the flight layout of the UAV swarm and the control strategies for at least some UAVs, ensuring that high-performing UAVs maintain normal overall swarm flight; adjusts the UAV flight status based on dynamic data of the relative spatial relationship between the intruder and the UAV swarm, achieving accurate containment and control of the intruder by the UAV swarm, effectively limiting the intruder's flight activities; and controls the UAV swarm to initiate non-contact interference against the intruder based on its flight trajectory until the intruder's flight status changes. This effectively adapts to different intruder flight change scenarios, accurately and proactively driving away intruders, improving defense reliability and timeliness.

[0043] Other features and advantages of the invention will be set forth in the description which follows, 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, claims, and drawings.

[0044] 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

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating the deep learning-based UAV emergency defense control method provided by this invention.

[0047] Figure 2 A schematic diagram illustrating the adjustment of the flight layout for a drone swarm.

[0048] Figure 3 This is a schematic diagram of a drone swarm launching non-contact interference against an intruder.

[0049] Figure 4 A schematic diagram of the structure of the deep learning-based UAV emergency defense control system provided by the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] See Figure 1 This is a flowchart illustrating a deep learning-based drone emergency defense control method provided in an embodiment of the present invention. The deep learning-based drone emergency defense control method includes:

[0052] Step S1: Perform deep neural network learning on abnormal state data around the base area to predict potential intrusion events, thereby setting a patrol strategy for the drone swarm around the base area.

[0053] Step S2: Monitor and identify the local interference encountered by the drone swarm during the inspection process, and adjust the flight layout of the drone swarm and the control strategy for at least some of the drones.

[0054] Step S3: Based on the dynamic data of the relative spatial relationship between the intruders and the drone cluster around the base area, adjust the flight status of the corresponding drones in the drone cluster so that the intruders and the drone cluster meet the preset positional relationship conditions.

[0055] Step S4: Based on the flight trajectory of the intruder, control at least one drone in the drone cluster to initiate non-contact interference against the intruder until the intruder's flight status changes.

[0056] The beneficial effects of the above technical solution are as follows: This deep learning-based UAV emergency defense control method uses deep neural networks to learn abnormal state data around the base area to predict potential intrusions and set UAV swarm inspection strategies, quickly initiating UAV surveillance in the initial stage of intrusion; it identifies local interference experienced by the UAV swarm during the inspection process, adjusts the flight layout of the UAV swarm and the control strategies for at least some UAVs, ensuring that high-performance UAVs maintain normal flight of the swarm as a whole; it adjusts the flight state of UAVs based on dynamic data of the relative spatial relationship between the intruder and the UAV swarm, achieving accurate containment and control of the intruder by the UAV swarm, effectively restricting the flight activities of the intruder; and it controls the UAV swarm to initiate non-contact interference against the intruder based on the intruder's flight trajectory until the intruder's flight state changes, effectively adapting to different intruder flight change scenarios, accurately and proactively driving away the intruder, and improving the reliability and timeliness of defense.

[0057] Preferably, in step S1, before performing deep neural network learning on the abnormal state data surrounding the base area, the following steps are included:

[0058] Visual and radar scanning detections were performed on the area surrounding the base to obtain images and radar data of the surrounding environment.

[0059] Analyze the surrounding environmental radar data to determine the flight dynamic paths of suspected foreign objects around the base area;

[0060] Based on the flight dynamic path, local identification is performed on the surrounding environmental images to obtain the flight attitude of suspected foreign objects.

[0061] Temporal evolution analysis was performed on the flight dynamic path and flight attitude of suspected foreign objects to obtain abnormal state data around the base area; among which, the abnormal state data included the flight dynamic path data and flight attitude data of suspected foreign objects near the base area.

[0062] Base areas such as sports fields and parks, being open spaces, are susceptible to external intrusion and surveillance. To promptly monitor potential intrusions, cameras and radar sensors can be deployed at different locations within the base area. These cameras and sensors can perform panoramic visual and radar scans of the surrounding environment, respectively, to obtain environmental images and radar data. Specifically, the environmental sub-images and radar sub-data generated by each camera and radar sensor can be collected first. Then, based on the coordinates of all cameras' locations within the base area, the environmental sub-images generated by all cameras are integrated into a single environmental image, and the radar sub-data generated by all radar sensors are integrated into a single environmental radar data. This provides a dynamic representation of the distribution of external objects in the surrounding environment of the base area at both the visible visual and radar visual levels.

[0063] Considering that radar sensors have a longer object detection range than cameras, when there are foreign objects in the surrounding environment of the base area, radar sensors will detect their presence before cameras. Furthermore, radar sensors can perform real-time dynamic monitoring of these objects, obtaining their movement information. Therefore, we first extract several radar sub-data points corresponding to several time points from the surrounding environmental radar data, and then perform temporal evolution analysis on these sub-data points to determine the flight dynamic paths of suspected foreign objects around the base area. These flight dynamic paths represent the spatial range where suspected foreign objects exist around the base area. Then, based on the spatial range of the suspected foreign object's flight dynamic path, we extract local image segments of the suspected foreign object from the surrounding environmental images, and perform object contour recognition on these local image segments to obtain the flight attitude of the suspected foreign object. By combining radar sensors and cameras, the system can quickly locate and dynamically track foreign objects when they are far from the base area, obtaining their flight paths. When the foreign objects are close to the base area, it can perform local image recognition, comprehensively and accurately determining the flight attitude of the foreign objects. This provides a reliable basis for subsequent judgments on whether the foreign objects constitute an abnormal impact on the surrounding area of ​​the base area.

[0064] In actual monitoring, not all foreign objects surrounding the base area will ultimately intrude into the base area. Some foreign objects may interfere with the surrounding area while flying along pre-planned paths; these should not be considered potential intruders. Others may intrude by gradually approaching the base area. To avoid assuming all foreign objects surrounding the base area are intruders and placing excessive pressure on the drone swarm's defense mission, it is necessary to identify and filter abnormal state data around the base area. Specifically, temporal evolution analysis is performed on the flight dynamic paths and flight attitudes of suspected foreign objects to identify those gradually approaching the base area. This flight dynamic path and flight attitude data of the suspected foreign objects are then used as abnormal state data around the base area to facilitate further identification of potential intrusion events.

[0065] Preferably, in step S1, deep neural network learning is performed on the abnormal state data around the base area to predict potential intrusion events, including:

[0066] After noise removal and reliable data screening of the flight dynamic path data and flight attitude data contained in the abnormal state data, the flight dynamic path data and flight attitude data are input into a deep neural network model for analysis and learning to obtain the flight trend of suspected foreign objects relative to the base area.

[0067] Based on flight trends, predict the spatial distribution of potential intrusion events in the base area.

[0068] The abnormal status data includes flight dynamic path data and flight attitude data of suspected foreign objects approaching the base area. Considering the significant uncertainty in the flight of suspected foreign objects, to accurately predict their flight trend relative to the base area, the flight dynamic path data and flight attitude data included in the abnormal status data are first processed by noise removal and reliable data screening. Then, the flight dynamic path data and flight attitude data are input into a deep neural network model for analysis and learning to comprehensively and accurately predict the flight trend of the suspected foreign object relative to the base area. This flight trend may include, but is not limited to, the changing trends in the flight distance and flight orientation of the suspected foreign object relative to the base area. Based on the flight trend of the suspected foreign object relative to the base area, it is determined whether the current flight behavior of the suspected foreign object constitutes a potential intrusion event to the base area, thereby predicting the spatial distribution of potential intrusion events occurring in the base area. This spatial distribution may include, but is not limited to, the spatial location range of potential intrusion events relative to the base area.

[0069] Preferably, in step S1, a patrol strategy for the drone swarm around the base area is set, including:

[0070] Based on the spatial distribution of potential intrusion events in the base area and the geographical topography surrounding the base area, identify the intruders and their spatial distribution intrusion locations around the base area.

[0071] Based on the spatial distribution of intrusion sites and the readiness status of drone swarms within the base area, a drone swarm inspection strategy is set for the area surrounding the base area. The readiness status of the drone swarms includes the parking positions and takeoff acceleration characteristics of all drone swarms within the base area. The drone swarm inspection strategy includes the range of drone swarms' flight coverage area around the base area.

[0072] When a potential intrusion event is identified in the base area, a drone swarm needs to be initiated to intercept it. To ensure the immediacy of the drone swarm interception, a suitable drone swarm needs to be selected from within the base area to perform the interception mission. Furthermore, potential intrusion events in the base area are also affected by geographical and topographical factors such as hillsides around the base area, requiring intruders to bypass these obstacles to enter the base area. To accurately determine the actual intruders and their locations around the base area, a spatial comparison needs to be made between the spatial distribution of potential intrusion events in the base area and the distribution of geographical and topographical obstacles such as hillsides around the base area to identify the intruders that actually cause disturbances to the base area and their spatial distribution.

[0073] To ensure that the standby drone clusters within the base area can respond promptly to intrusions, based on the spatial distribution of intrusion sites around the base area and the parking locations and takeoff acceleration characteristics of all drone clusters within the base area, drone clusters with parking locations close to the intrusion sites and high takeoff acceleration are selected. These selected drone clusters are then assigned to patrol and fly within the perimeter of the base area to ensure that the drone clusters can take off in time and reach the patrol and fly within the aforementioned patrol and fly coverage area to effectively track and intercept the intruders.

[0074] Preferably, in step S2, monitoring and identifying local interference experienced by the drone swarm during the inspection process includes:

[0075] Monitor the external signals received locally by each drone in the drone cluster during the inspection process, and extract the interference signal components of the external signals;

[0076] Monitor the flight trajectory and attitude drift changes of all drones in the drone cluster during the inspection process, analyze the flight trajectory and attitude drift change data, and obtain the time when the drone flight anomaly occurred.

[0077] By comparing the time series of interference signal components with the time of drone flight anomalies, drones experiencing flight instability events within a drone swarm can be identified.

[0078] Intruders approaching the base area may cause electromagnetic interference to the drone swarm, affecting the normal flight of at least some drones within the swarm. The impact of this interference varies among different drones in the swarm; some drones experience greater interference than others. Drones experiencing significant interference may experience flight malfunctions or be unable to communicate normally with the ground base station. To accurately identify drones experiencing significant interference and unstable flight, the system first monitors the external signals received locally by each drone during its inspection process, identifying useful command signals and interference signals. It also monitors and analyzes the flight trajectory and attitude drift changes of each drone during the inspection process. If the flight trajectory drift or attitude drift of a drone exceeds a preset drift threshold within a certain time interval during the inspection process, this time interval is determined as the time of the drone flight anomaly. Next, the time series of the interference signal components in the external signal is compared with the time when the drone flight anomaly occurs. If the time series overlaps with the time when the drone flight anomaly occurs, it is determined that the drone has experienced a flight instability event; otherwise, it is determined that the drone has not experienced a flight instability event. This facilitates timely and accurate adjustment of the drone's position within the cluster, minimizing the interference impact on the drone.

[0079] Preferably, in step S2, adjusting the flight layout of the drone swarm and the control strategy for at least some of the drones includes:

[0080] Based on the distribution location of the drone experiencing flight instability within the drone swarm and the relative flight speed between the drone experiencing flight instability and other drones, the flight layout of the drone swarm is adjusted, switching the drone experiencing flight instability to the inner area of ​​the drone swarm and switching the other drones to the outer area of ​​the drone swarm.

[0081] Based on the communication control strategy between the drone experiencing flight instability and the ground base station, the drone experiencing flight instability is switched to a working state controlled by other drones, thereby enabling other drones to locate and control the drone experiencing flight instability.

[0082] Drones experiencing flight instability events may lose their location and communication with ground base stations due to interference from intruders. This can cause them to deviate from their planned flight paths, increasing the risk of crashes or collisions with other drones in the swarm. To mitigate the impact of intruder interference, the flight layout of the drone swarm should be adjusted. For example... Figure 2 The diagram shows the distribution of drones before and after a flight layout adjustment of a drone swarm, where UAV1 represents the drone that experienced the flight instability event. Before the flight layout adjustment, all drones in the swarm were arranged in a circular array. After the adjustment, the drone that experienced the instability event was moved to the inner area of ​​the swarm (the area inside the dashed arc), while the other drones were moved to the outer area (the area outside the dashed arc). This physically isolates the drone that experienced the instability event. Specifically, based on the distribution of the drone that experienced the instability event within the swarm and its relative flight speed to other drones, the drone that experienced the instability event was moved to the inner area of ​​the swarm, while the other drones were moved to the outer area, ensuring that no collisions occur during the flight layout adjustment process. In addition, in order to ensure that the drone experiencing flight instability can still obtain positioning signals and flight control command signals after switching positions, the drone experiencing flight instability is switched to a working state controlled by other drones according to the communication control strategy between the drone experiencing flight instability and the ground base station. This avoids external interference to the communication between the drone experiencing flight instability and the ground base station, and instead relies on other drones to locate and control the drone experiencing flight instability.

[0083] Preferably, step S3 further includes:

[0084] Obtain the predicted flight trajectories of intruders around the base area and the planned flight trajectories of the entire drone swarm;

[0085] By spatially comparing the predicted flight trajectory with the planned flight trajectory, dynamic data on the relative spatial relationship between intruders and drone swarms around the base area is obtained; among which, the dynamic data on the relative spatial relationship includes dynamic data on the relative spatial distance and relative spatial orientation between intruders and drone swarms around the base area.

[0086] To enable real-time tracking and precise containment of intruders surrounding the base area, it is necessary to control the drone swarm to effectively encircle the intruders in a physical space, ensuring that the relative spatial relationship between the drone swarm and the intruders matches during flight. Specifically, by comparing the predicted flight trajectories of the intruders around the base area with the planned flight trajectories of the drone swarm as a whole, dynamic data on the relative spatial distance and relative spatial orientation between the intruders and the drone swarm are obtained, providing a reliable spatial location reference for subsequent drone control to contain the intruders.

[0087] Preferably, in step S3, based on the dynamic data of the relative spatial relationship between the intruders surrounding the base area and the drone swarm, the flight status of the corresponding drones in the drone swarm is adjusted so that the intruders and the drone swarm meet the preset positional relationship conditions, including:

[0088] Time-evolution analysis of the dynamic data of relative spatial distance and relative spatial orientation between intruders and drone swarms around the base area is performed to predict the range of encounter locations between drone swarms and intruders.

[0089] Based on the range of the encounter location, adjust the flight trajectory and / or flight speed of drones in the outer area of ​​the drone swarm, thereby enabling the drone swarm to surround the intruder.

[0090] Dynamic data on the relative spatial distance and orientation between intruders and drone swarms around the base area are analyzed in real time to predict the encounter range between the drone swarm and the intruders, precisely narrowing down the space where the drone swarm can effectively contain the intruders. Based on this encounter range, the flight trajectories and / or speeds of drones in the outer areas of the drone swarm are adjusted to allow the drone swarm to encircle the intruders. This ensures precise physical containment and isolation of the intruders during the encounter, facilitating subsequent interference and effectively improving the efficiency of intruder control.

[0091] Preferably, in step S4, based on the flight trajectory of the intruder, at least one drone in the drone cluster is controlled to initiate non-contact interference against the intruder until the intruder's flight state changes, including:

[0092] Based on the flight trajectory of the intruder, identify the drone that is closest to the intruder within the drone swarm;

[0093] Based on the relative orientation change characteristics between the nearest drone and the intruder, control the intensity and duration of electromagnetic and / or laser interference operations launched by the nearest drone against the intruder until the intruder flies away from the base area or stops flying.

[0094] Once the drone swarm surrounds the intruder, it adjusts its own flight path based on the intruder's trajectory, effectively encircling it for a certain period. At this point, the interference previously experienced by drones within the swarm due to flight instability is negligible after the control strategy is switched. The drone closest to the intruder within the swarm is then identified, and based on the relative positional changes between the closest drone and the intruder, the intensity and duration of electromagnetic and / or laser interference are controlled to continue until the intruder flies away from the base area or ceases flight. This process of initiating non-contact interference against the intruder is described below. Figure 3 As shown, the above methods can quickly snipe and control intruders, reducing their impact on the base area.

[0095] See Figure 4 This is a schematic diagram of the structure of a deep learning-based UAV emergency defense control system provided in an embodiment of the present invention. The deep learning-based UAV emergency defense control system includes:

[0096] The deep learning and inspection strategy setting module is used to perform deep neural network learning on abnormal state data around the base area to predict potential intrusion events, thereby setting inspection strategies for drone swarms around the base area.

[0097] Interference monitoring and identification module, used to monitor and identify local interference experienced by the drone swarm during the inspection process;

[0098] The flight layout and control strategy adjustment module is used to adjust the flight layout of the drone swarm and the control strategy for at least some of the drones.

[0099] The cluster flight status adjustment module is used to adjust the flight status of the corresponding drones in the drone cluster based on the dynamic data of the relative spatial relationship between the intruders around the base area and the drone cluster, so that the intruders and the drone cluster meet the preset positional relationship conditions.

[0100] The interference initiation control module is used to control at least one drone in the drone swarm to initiate non-contact interference against the intruder based on the intruder's flight trajectory until the intruder's flight state changes.

[0101] The operation and effect of the deep learning-based UAV emergency defense control system of the present invention are consistent with the deep learning-based UAV emergency defense control method described above, and will not be described again here.

[0102] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A deep learning-based emergency defense control method for unmanned aerial vehicles, characterized in that, include: Step S1: Perform deep neural network learning on abnormal state data around the base area to predict potential intrusion events, thereby setting a patrol strategy for the drone swarm around the base area. Step S2 involves monitoring and identifying local interference experienced by the drone swarm during the inspection process, and adjusting the flight layout of the drone swarm and the control strategy for at least some of the drones, including: Monitor the external signals received locally by each drone in the drone cluster during the inspection process, and extract the interference signal components of the external signals; Monitor the flight trajectory and attitude drift change data of all drones under the drone cluster during the inspection process, analyze the flight trajectory and attitude drift change data, and obtain the time of the drone flight anomaly. By comparing the time series of the external signals with the time of the drone flight anomaly, the drones that experienced flight instability events within the drone cluster can be identified. Based on the distribution location of the drones experiencing flight instability within the drone swarm and the relative flight speed between the drones experiencing flight instability and other drones, the flight layout of the drone swarm is adjusted, switching the drones experiencing flight instability to the inner area of ​​the drone swarm and switching the other drones to the outer area of ​​the drone swarm. Based on the communication control strategy between the drone experiencing flight instability and the ground base station, the drone experiencing flight instability is switched to a working state controlled by other drones, thereby enabling other drones to locate and control the drone experiencing flight instability. Step S3: Based on the dynamic data of the relative spatial relationship between the intruders around the base area and the drone cluster, adjust the flight status of the corresponding drones in the drone cluster so that the intruders and the drone cluster meet the preset positional relationship conditions. Step S4: Based on the flight trajectory of the intruder, control at least one drone in the drone cluster to initiate non-contact interference against the intruder until the intruder's flight state changes.

2. The deep learning-based UAV emergency defense control method as described in claim 1, characterized in that: In step S1, before performing deep neural network learning on the abnormal state data around the base area, the following steps are included: Visual and radar scanning detections are performed on the area surrounding the base to obtain environmental images and radar data of the surrounding environment of the base area. Analyze the surrounding environmental radar data to determine the flight dynamic paths of suspected foreign objects around the base area; Based on the flight dynamic path, the surrounding environment image is locally identified to obtain the flight attitude of the suspected foreign object; A time-domain evolution analysis is performed on the flight dynamic path and flight attitude of the suspected foreign object to obtain abnormal state data around the base area; wherein, the abnormal state data includes the flight dynamic path data and flight attitude data of the suspected foreign object approaching the base area.

3. The deep learning-based UAV emergency defense control method as described in claim 2, characterized in that: In step S1, deep neural network learning is performed on the abnormal state data around the base area to predict potential intrusion events, including: After noise removal and reliable data screening of the flight dynamic path data and flight attitude data contained in the abnormal state data, the flight dynamic path data and flight attitude data are input into a deep neural network model for analysis and learning to obtain the flight trend of the suspected foreign object relative to the base area. Based on the flight trends, predict the spatial distribution of potential intrusion events in the base area.

4. The deep learning-based UAV emergency defense control method as described in claim 3, characterized in that: In step S1, a patrol strategy for the drone swarm around the base area is set, including: Based on the spatial distribution of potential intrusion events in the base area and the geographical topography surrounding the base area, identify the intruders and their spatial distribution of intrusion locations around the base area; Based on the spatial distribution of the intrusion locations and the readiness status of the drone clusters within the base area, a patrol strategy for drone clusters around the base area is set; wherein, the readiness status of the drone clusters includes the parking positions and takeoff acceleration characteristics of each drone cluster within the base area; the drone cluster patrol strategy includes the patrol flight coverage area of ​​the drone clusters around the base area.

5. The deep learning-based UAV emergency defense control method as described in claim 1, characterized in that: Step S3 further includes: Obtain the predicted flight trajectories of intruders surrounding the base area and the planned flight trajectories of the entire drone swarm; By spatially comparing the predicted flight trajectory with the planned flight trajectory, dynamic data on the relative spatial relationship between intruders around the base area and the drone cluster is obtained; wherein, the dynamic data on the relative spatial relationship includes dynamic data on the relative spatial distance and relative spatial orientation between intruders around the base area and the drone cluster.

6. The deep learning-based UAV emergency defense control method as described in claim 5, characterized in that: In step S3, based on the dynamic data of the relative spatial relationship between the intruders surrounding the base area and the drone swarm, the flight status of the corresponding drones in the drone swarm is adjusted so that the intruders and the drone swarm meet the preset positional relationship conditions, including: The relative spatial distance and relative spatial orientation dynamic data between the intruders around the base area and the drone cluster are analyzed in time to predict the range of the encounter location between the drone cluster and the intruders. Based on the range of the encounter location, the flight trajectories and / or speeds of the drones in the outer area of ​​the drone swarm are adjusted, thereby enabling the drone swarm to surround the intruder.

7. The deep learning-based UAV emergency defense control method as described in claim 1, characterized in that: In step S4, based on the flight trajectory of the intruder, at least one drone in the drone cluster is controlled to initiate non-contact interference against the intruder until the intruder's flight state changes, including: Based on the flight trajectory of the intruder, determine the drone that is closest to the intruder within the drone cluster; Based on the relative orientation change characteristics between the nearest drone and the intruder, the intensity and duration of the operation of the nearest drone to launch electromagnetic interference and / or laser interference against the intruder are controlled until the intruder flies away from the base area or stops flying.

8. A deep learning-based unmanned aerial vehicle (UAV) emergency defense control system, characterized in that, include: The deep learning and inspection strategy setting module is used to perform deep neural network learning on abnormal state data around the base area to predict potential intrusion events, thereby setting an inspection strategy for the drone swarm around the base area. The interference monitoring and identification module is used to monitor and identify local interference experienced by the drone swarm during the inspection process, including: Monitor the external signals received locally by each drone in the drone cluster during the inspection process, and extract the interference signal components of the external signals; Monitor the flight trajectory and attitude drift change data of all drones under the drone cluster during the inspection process, analyze the flight trajectory and attitude drift change data, and obtain the time of the drone flight anomaly. By comparing the time series of the external signals with the time of the drone flight anomaly, the drones that experienced flight instability events within the drone cluster can be identified. A flight layout and control strategy adjustment module is used to adjust the flight layout of the UAV swarm and the control strategy for at least some of the UAVs, including: Based on the distribution location of the drones experiencing flight instability within the drone swarm and the relative flight speed between the drones experiencing flight instability and other drones, the flight layout of the drone swarm is adjusted, switching the drones experiencing flight instability to the inner area of ​​the drone swarm and switching the other drones to the outer area of ​​the drone swarm. Based on the communication control strategy between the drone experiencing flight instability and the ground base station, the drone experiencing flight instability is switched to a working state controlled by other drones, thereby enabling other drones to locate and control the drone experiencing flight instability. The cluster flight status adjustment module is used to adjust the flight status of the corresponding drones in the drone cluster based on the dynamic data of the relative spatial relationship between the intruders around the base area and the drone cluster, so that the intruders and the drone cluster meet the preset positional relationship conditions. The interference initiation control module is used to control at least one UAV in the UAV cluster to initiate non-contact interference against the intruder based on the flight trajectory of the intruder, until the intruder changes its flight state.

Images