Target tracking methods, devices, and equipment based on multi-agent behavior planning

Abstract

This application relates to a target tracking method, apparatus, and device based on multi-agent behavior planning, comprising: dividing the planned behaviors of a probe drone and a target drone into multiple agents; constructing a tracking and localization model, which includes multiple independent behavior networks and evaluation networks, with each agent corresponding to one behavior network and one evaluation network; calculating the localization accuracy of the probe drone based on the probe drone's localization position relative to the target drone and the target drone's actual position; assigning reward and penalty values ​​to the corresponding agents based on the localization accuracy; guiding the optimization of the behavior networks and evaluation networks based on the reward and penalty values; and obtaining a trained tracking and localization model for tracking and localization when the tracking and localization model meets a preset iteration termination condition. This invention solves the problem of traditional machine learning methods failing to converge in complex environments and improves the accuracy of the localization system.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a target tracking method, apparatus, and device based on multi-agent behavior planning. Background Technology

[0002] In the field of UAV detection, electromagnetic signal-based positioning systems offer good positioning performance and low operating costs, making them widely used in electronic reconnaissance in complex electromagnetic environments. The timing and location measurement errors of radar sites, as well as the geometric configuration of the geographical location, are the main factors affecting positioning accuracy. Using UAVs instead of fixed sites for positioning can improve the geometric configuration of the positioning system, thereby improving positioning accuracy. However, using UAVs to detect and locate other UAVs presents new challenges: the effective range of UAV detection and positioning is limited, while the target UAV has mobility. Therefore, the detection UAV needs to adjust its own position according to the target's location to keep the target within its effective range. The movement between the target UAV and the detection UAV can be viewed as an adversarial process.

[0003] To improve the accuracy of positioning systems, traditional machine learning methods are typically used to plan mobile positioning platforms, making static deployments only when the target takes a specific route. However, compared to the specific route conditions of the target, the adversarial relationship between the positioning system and the target increases the uncertainty and complexity of the environment. When the target route changes, traditional machine learning methods often require replanning and deployment due to the complex target distribution in the solution space, and the training efficiency is low. Summary of the Invention

[0004] Therefore, it is necessary to provide a target tracking method, device, and equipment based on multi-agent behavior planning to address the above-mentioned technical problems, so as to maintain good information capture capabilities in the face of complex and ever-changing environments and improve the positioning accuracy of the positioning system.

[0005] A target tracking method based on multi-agent behavior planning, the method comprising:

[0006] The planned behaviors of the detection drone and the target drone are divided into multiple intelligent agents respectively;

[0007] A tracking and localization model is constructed; the tracking and localization model includes multiple independent behavioral networks and evaluation networks; each agent corresponds to one behavioral network and one evaluation network.

[0008] The system obtains the location of the detection drone relative to the target drone and the actual location of the target drone. Based on the location and the actual location, it calculates the positioning accuracy of the detection drone and assigns reward and penalty values ​​to the corresponding intelligent agents based on the positioning accuracy.

[0009] The reward value and the penalty value guide the optimization of the behavior network and the evaluation network corresponding to each agent. When the tracking and localization model meets the preset iteration termination condition, the trained tracking and localization model is obtained.

[0010] The trained tracking and positioning model is used to track and locate the target UAV.

[0011] A target tracking device based on multi-agent behavior planning, the device comprising:

[0012] The behavior segmentation module is used to divide the planned behaviors of the detection drone and the target drone into multiple intelligent agents respectively;

[0013] A model building module is used to build a tracking and localization model; the tracking and localization model includes multiple independent behavioral networks and evaluation networks; each agent corresponds to one behavioral network and one evaluation network.

[0014] The reward and punishment allocation module is used to obtain the positioning position of the detection drone relative to the target drone and the actual position of the target drone, calculate the positioning accuracy of the detection drone based on the positioning position and the actual position, and allocate reward and punishment values ​​to the corresponding intelligent agents based on the positioning accuracy.

[0015] The network optimization module is used to guide the optimization of the behavior network and the evaluation network corresponding to each agent according to the reward value and the penalty value. When the tracking and localization model meets the preset iteration termination condition, the trained tracking and localization model is obtained.

[0016] The tracking and positioning module is used to track and locate the target UAV using the trained tracking and positioning model.

[0017] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement any of the steps in the above-described target tracking method based on multi-agent behavior planning.

[0018] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the steps of the above-described target tracking method based on multi-agent behavior planning.

[0019] The aforementioned target tracking method, apparatus, and device based on multi-agent behavior planning first divides the planned behaviors of the probe drone and the target drone into multiple agents. Then, a tracking and localization model is constructed, comprising multiple independent behavior networks and an evaluation network, with each agent corresponding to one behavior network. Next, the positioning position of the probe drone relative to the target drone and the actual position of the target drone are obtained. The positioning accuracy of the probe drone is calculated based on the positioning position and the actual position, and reward and penalty values ​​are assigned to the corresponding agents based on the positioning accuracy. Then, the reward and penalty values ​​guide the optimization of the behavior networks and evaluation networks corresponding to each agent. When the tracking and localization model meets a preset iteration termination condition, a trained tracking and localization model is obtained. Finally, the trained tracking and localization model can be used to track and locate the target drone. This invention considers the high degree of freedom in the behavior of the probe drone and the target drone, and their complex spatial relationships. Therefore, the actions to be planned are decomposed into multiple agents to distribute and control the units. The behavior networks corresponding to each agent are executed in a distributed manner, while the evaluation network performs centralized training, solving the problem of convergence difficulties in complex environments using traditional machine learning methods. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a target tracking method based on multi-agent behavior planning in one embodiment;

[0021] Figure 2 This is a schematic diagram of the network parameters of the behavior network and the evaluation network in one embodiment, where (a) is a schematic diagram of the network parameters of the behavior network Actor, and (b) is a schematic diagram of the network parameters of the evaluation network Critic.

[0022] Figure 3 This is a structural block diagram of a target tracking method device based on multi-agent behavior planning in one embodiment;

[0023] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0025] In one embodiment, such as Figure 1 As shown, a target tracking method based on multi-agent behavior planning is provided, including the following steps:

[0026] Step 102: Divide the planned behaviors of the detection drone and the target drone into multiple intelligent agents.

[0027] The planned behaviors for the detection drone include adjusting its angle and distance relative to the origin of the coordinate system and / or whether to replace it with a fixed radar station, and which fixed radar station to replace it with. The planned behaviors for the target drone include adjusting its angle and / or distance relative to the origin of the coordinate system. The origin of the coordinate system can be the initial position of the detection drone. Replacing it with a fixed radar station means, for example, if there were radar stations A, B, C, and D, the drone changes from replacing station A to replacing station B, C, or D. Using drones for dynamic aerial detection reduces the impact of terrain and other factors on detection accuracy.

[0028] The agents of the detection drone cooperate with each other to keep the target drone within the detection range as much as possible, while the agents of the target drone cooperate with each other to keep the target drone as undetectable as possible. It can be seen that the agents of the detection drone and the agents of the target drone are in conflict with each other. The cooperative or antagonistic behavior between the agents constitutes the macroscopic event of the drone tracking and locating the target.

[0029] Step 104: Construct a tracking and positioning model.

[0030] The tracking and localization model comprises multiple independent behavioral networks and evaluation networks, with each agent corresponding to one behavioral network and one evaluation network. Each behavioral network takes the environmental state within the agent's observation range as input, while the evaluation network takes the overall environmental state.

[0031] Step 106: Obtain the location of the target drone from the detection drone and the actual location of the target drone. Calculate the positioning accuracy of the detection drone based on the location and actual location. Assign reward and penalty values ​​to the corresponding intelligent agents based on the positioning accuracy.

[0032] If the positioning accuracy is good, a reward value is assigned to the agent corresponding to the detecting drone, and a penalty value is assigned to the agent corresponding to the target drone; if the positioning accuracy is poor, a penalty value is assigned to the agent corresponding to the detecting drone, and a reward value is assigned to the agent corresponding to the target drone.

[0033] Step 108: Based on the reward value and penalty value, guide the optimization of the behavior network and evaluation network corresponding to each agent. When the tracking and localization model meets the preset iteration termination condition, the trained tracking and localization model is obtained.

[0034] The behavioral networks corresponding to each agent will be optimized in the direction of higher reward value and avoid penalty value. When optimizing, the evaluation network will take into account all behavioral networks to maximize the global reward value.

[0035] Step 110: Use the trained tracking and positioning model to track and locate the target UAV.

[0036] In the aforementioned target tracking method based on multi-agent behavior planning, the planned behaviors of both the probe drone and the target drone are first divided into multiple agents. Then, a tracking and localization model is constructed, comprising multiple independent behavior networks and an evaluation network, with each agent corresponding to one behavior network. Next, the probe drone's location relative to the target drone and the target drone's actual location are obtained. The localization accuracy of the probe drone is calculated based on the location and actual location, and reward and penalty values ​​are assigned to the corresponding agents based on the localization accuracy. The reward and penalty values ​​then guide the optimization of the behavior networks and evaluation networks corresponding to each agent. When the tracking and localization model meets the iteration termination condition, a trained tracking and localization model is obtained. Finally, the trained tracking and localization model can be used to track and locate the target drone. This method considers the high degree of freedom in the behavior of both the probe drone and the target drone, and their complex spatial relationships. Therefore, the actions to be planned are decomposed into multiple agents to distribute the control units, observe the environment globally, and allow the policy to use additional information to simplify training. The behavior networks corresponding to each agent are executed distributedly, while the evaluation network performs centralized training, solving the problem of convergence difficulties in complex environments encountered by traditional machine learning methods. This method also simulates the condition that the target intentionally evades the positioning system, enhances the tracking capability of the mobile platform, directly solves the problem of NPA distribution configuration adjustment, adjusts the dominant area closer to the target, and improves the accuracy of the positioning system.

[0037] In summary, to enhance the practicality of positioning system planning and address the issue of low training efficiency in traditional machine learning methods, this approach employs swarm intelligence decision-making to plan UAV adversarial behavior. By training to find the optimal guard position strategy, the positioning system achieves higher accuracy.

[0038] In one embodiment, a unified coordinate system is established with the initial position of the probe drone as the origin, and the positions of the probe drone and the target drone in the subsequent process are described according to the world coordinate system.

[0039] The location of the target drone relative to the detection drone and the actual location of the target drone are obtained. Based on the location and actual location, the positioning accuracy of the detection drone is calculated as follows:

[0040]

[0041] Wherein, NPA represents positioning accuracy, P(x,y,z) represents the actual position of the target UAV, and P(x′,y′,z′) represents the positioning position of the probe UAV relative to the target UAV.

[0042] In one embodiment, reward and penalty values ​​are assigned to the corresponding agents based on positioning accuracy, including:

[0043] Set the positioning accuracy threshold;

[0044] When the positioning accuracy is less than the positioning accuracy threshold, the number of successful positioning is reset to zero and the number of positioning failures is continuously counted. The penalty value obtained by the agent corresponding to the detection drone and the reward value obtained by the agent corresponding to the target drone increase with the increase of the number of positioning failures.

[0045] When the positioning accuracy is greater than the positioning accuracy threshold, the number of positioning loss counts is reset to zero and the number of successful positioning counts is started continuously. The reward value obtained by the agent corresponding to the detection drone and the penalty value obtained by the agent corresponding to the target drone increase with the increase of the number of successful positioning counts.

[0046] In one embodiment, the iteration termination condition is designed based on the number of successful positioning attempts or the number of positioning failures. The steps for designing the iteration termination condition include: setting the number of successful positioning attempts or the number of positioning failures exceeding a preset number as the iteration termination condition for the tracking and positioning model.

[0047] If the positioning accuracy remains good for a certain period of time, that is, the number of successful positioning exceeds a certain number, it indicates that the positioning system is training well and the iterative training of the tracking and positioning model can be terminated to obtain a well-trained tracking and positioning model. Conversely, if the positioning accuracy is poor for a certain period of time, that is, the number of positioning failures exceeds a certain number, it indicates that the training direction of the positioning system may be wrong, and iterative training needs to be interrupted, the environment terminated and reset, and iterative training restarted.

[0048] In one embodiment, the optimization of the behavior network and evaluation network corresponding to each agent is guided by reward and penalty values, including:

[0049] The system acquires the local state observations of each agent at the current stage, inputs the local state observations into the corresponding behavior network to obtain the actions performed by the agent, obtains the local state observations of the next stage and the corresponding reward and penalty values ​​based on the actions, and updates the policy parameters of the corresponding behavior network according to the local state observations and actions of the current stage.

[0050] The local state observations of the current stage and the next stage are summarized to obtain the global state observations of the current stage and the next stage, respectively. These are then input into the evaluation network. The action value network parameters of the evaluation network are updated with the goal of maximizing the global reward value. In other words, the global observations of the evaluation network guide the training of each agent, while the local observations of the behavior network guide the execution of training.

[0051] Through numerous experiments, parameter settings and the degree of prior information provision suitable for the context of this invention were established, such as the positioning accuracy threshold and corresponding reward / penalty values. In a suitable incentive mechanism for the agent, sufficient prior information ensures the training convergence speed, while the absence of excessive prior information prevents the agent from being constrained by the subjectivity of prior information, allowing for the training of new strategies.

[0052] In one embodiment, such as Figure 2 The diagram shows the network parameters of the behavior network and the evaluation network. The Actor represents the behavior network, the Critic represents the evaluation network, and the Linear layer is a linear classification layer. The left side of the arrow indicates the activation function, and the right side indicates the input or output size of each layer. The training steps for the tracking and localization model include:

[0053] S1. Initialize network state;

[0054] S2, Enter iteration;

[0055] (1) Each behavioral network independently acquires the feature vector of the local state observation of the agent at the current stage.

[0056] (2) Each behavioral network collects its data Store in the experience replay pool; where s represents the local state observation of the current stage, a represents the action executed according to s, μ represents the gradient policy, r represents the reward / penalty value allocated according to a, s′ represents the local state observation of the next stage obtained according to a, and a′ represents the action executed according to s′.

[0057] (3) Each behavioral network uses its own Update strategy parameters;

[0058] (3) Each evaluation network updates the action value network parameters. During the update, the data of all behavior networks are considered. The optimization goal is to maximize the contribution of the evaluation network to the whole, that is, after the update, each agent trains in the direction of obtaining the highest reward value.

[0059] It should be understood that although Figure 1The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0060] In one embodiment, such as Figure 3 As shown, a target tracking device based on multi-agent behavior planning is provided, including: a behavior segmentation module, a model building module, a reward and punishment allocation module, a network optimization module, and a tracking and localization module, wherein:

[0061] The behavior segmentation module is used to divide the planned behaviors of the detection drone and the target drone into multiple intelligent agents respectively;

[0062] The model building module is used to build the tracking and localization model; the tracking and localization model includes multiple independent behavior networks and evaluation networks; each agent corresponds to one behavior network and one evaluation network;

[0063] The reward and punishment allocation module is used to obtain the location of the detection drone relative to the target drone and the actual location of the target drone, calculate the positioning accuracy of the detection drone based on the location and actual location, and allocate reward and punishment values ​​to the corresponding intelligent agents according to the positioning accuracy.

[0064] The network optimization module is used to guide the optimization of the behavior network and evaluation network of each agent based on the reward value and the penalty value. When the tracking and localization model meets the preset iteration termination condition, the trained tracking and localization model is obtained.

[0065] The tracking and positioning module is used to track and locate the target UAV using a trained tracking and positioning model.

[0066] In one embodiment, the reward and punishment allocation module is also used to calculate the positioning accuracy of the detection drone based on the location and the actual location:

[0067]

[0068] Wherein, NPA represents positioning accuracy, P(x,y,z) represents the actual position of the target UAV, and P(x′,y′,z′) represents the positioning position of the probe UAV relative to the target UAV.

[0069] Set a positioning accuracy threshold. When the positioning accuracy is less than the threshold, the number of successful positioning attempts is reset to zero and the number of positioning failures is continuously counted. The penalty value obtained by the agent corresponding to the probe drone and the reward value obtained by the agent corresponding to the target drone increase with the number of positioning failures. When the positioning accuracy is greater than the threshold, the number of positioning failures is reset to zero and the number of successful positioning attempts is continuously counted. The reward value obtained by the agent corresponding to the probe drone and the penalty value obtained by the agent corresponding to the target drone increase with the number of successful positioning attempts.

[0070] In one embodiment, the network optimization module is further configured to acquire the local state observations of each agent at the current stage, input the local state observations into the corresponding behavior network to obtain the actions performed by the agent, obtain the local state observations of the next stage and the corresponding reward and penalty values ​​based on the actions, and update the policy parameters of the corresponding behavior network according to the local state observations and actions of the current stage.

[0071] Summarize the local state observations of the current stage and the local state observations of the next stage to obtain the global state observations of the current stage and the global state observations of the next stage, and input them into the evaluation network. Update the action value network parameters of the evaluation network with the highest global reward value as the optimization objective.

[0072] Specific limitations regarding the target tracking device based on multi-agent behavior planning can be found in the limitations of the target tracking method based on multi-agent behavior planning mentioned above, and will not be repeated here. Each module in the aforementioned target tracking device based on multi-agent behavior planning can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0073] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and the database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data such as the location of the UAV. The network interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a target tracking method based on multi-agent behavior planning.

[0074] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0075] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described above.

[0076] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0077] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0078] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0079] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A target tracking method based on multi-agent behavior planning, characterized in that, The method includes: The planned behaviors of the detection drone and the target drone are divided into multiple intelligent agents respectively; A tracking and localization model is constructed; the tracking and localization model includes multiple independent behavioral networks and evaluation networks; each agent corresponds to one behavioral network and one evaluation network. The system obtains the location of the detection drone relative to the target drone and the actual location of the target drone. Based on the location and the actual location, it calculates the positioning accuracy of the detection drone and assigns reward and penalty values ​​to the corresponding intelligent agents based on the positioning accuracy. The reward value and the penalty value guide the optimization of the behavior network and the evaluation network corresponding to each agent. When the tracking and localization model meets the preset iteration termination condition, the trained tracking and localization model is obtained. The trained tracking and localization model is used to track and locate the target UAV. Before dividing the planned behaviors of the reconnaissance drone and the target drone into multiple agents, the following is also included: A unified coordinate system is established with the initial position of the probe drone as the origin; The planned behaviors of the detection drone include adjusting the angle and distance of the detection drone relative to the origin of the coordinate system and / or whether to replace the fixed radar station and which fixed radar station to replace; the planned behaviors of the target drone include adjusting the angle and / or distance of the target drone relative to the origin of the coordinate system.

2. The method according to claim 1, characterized in that, Calculating the positioning accuracy of the detection drone based on the stated location and the actual location includes: The positioning accuracy of the detection drone is calculated based on the stated location and the actual location as follows: in, Indicates positioning accuracy. Indicates the actual location of the target drone. This indicates the location of the target drone relative to the detection drone.

3. The method according to claim 1, characterized in that, Based on the positioning accuracy, reward and penalty values ​​are assigned to the corresponding intelligent agents, including: Set the positioning accuracy threshold; When the positioning accuracy is less than the positioning accuracy threshold, the number of successful positioning is reset to zero and the number of positioning failures is continuously counted. The penalty value obtained by the agent corresponding to the detection drone and the reward value obtained by the agent corresponding to the target drone increase with the increase of the number of positioning failures. When the positioning accuracy is greater than the positioning accuracy threshold, the number of positioning loss counts is reset to zero and the number of successful positioning counts is started continuously. The reward value obtained by the agent corresponding to the detection drone and the penalty value obtained by the agent corresponding to the target drone increase with the increase of the number of successful positioning counts.

4. The method according to claim 3, characterized in that, The iteration termination condition is designed based on the number of successful positioning attempts or the number of positioning failures. The steps for designing the iteration termination condition include: The number of successful positioning attempts or the number of positioning failures exceeding a preset number are used as the termination condition for the iteration of the tracking and positioning model.

5. The method according to claim 3, characterized in that, The optimization of the behavior network and evaluation network corresponding to each agent is guided by the reward value and the penalty value, including: The system acquires the local state observations of each agent at the current stage, inputs the local state observations into the corresponding behavior network to obtain the actions performed by the agent, obtains the local state observations of the next stage and the corresponding reward and penalty values ​​based on the actions, and updates the policy parameters of the corresponding behavior network according to the local state observations of the current stage and the actions. The local state observations of the current stage and the local state observations of the next stage are summarized to obtain the global state observations of the current stage and the global state observations of the next stage, and then input into the evaluation network. The action value network parameters of the evaluation network are updated with the goal of maximizing the global reward value.

6. A target tracking device based on multi-agent behavior planning implemented in the method of any one of claims 1-5, characterized in that, The device includes: The behavior segmentation module is used to divide the planned behaviors of the detection drone and the target drone into multiple intelligent agents respectively; A model building module is used to build a tracking and localization model; the tracking and localization model includes multiple independent behavioral networks and evaluation networks; each agent corresponds to one behavioral network and one evaluation network. The reward and punishment allocation module is used to obtain the positioning position of the detection drone relative to the target drone and the actual position of the target drone, calculate the positioning accuracy of the detection drone based on the positioning position and the actual position, and allocate reward and punishment values ​​to the corresponding intelligent agents based on the positioning accuracy. The network optimization module is used to guide the optimization of the behavior network and the evaluation network corresponding to each agent according to the reward value and the penalty value. When the tracking and localization model meets the preset iteration termination condition, the trained tracking and localization model is obtained. The tracking and positioning module is used to track and locate the target UAV using the trained tracking and positioning model.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images