Path planning method for anti-swarm interceptor cluster

Abstract

This invention provides a path planning method for an anti-swarm interceptor cluster, belonging to the field of artificial intelligence technology. The method includes: calculating the direction vector between continuous trajectory points of the target swarm based on its location information within the current time period, and predicting the interception zone of the target swarm; adaptively calling target planning strategies from a pre-built strategy library according to the application scenario at the current moment to generate a cooperative path for the interceptor cluster, the endpoint of which is within the interception zone; the strategy library includes a lightweight fast response strategy based on height difference collision avoidance, a progressive optimal search strategy based on random sampling, and a high-safety optimization strategy based on dynamic grid partitioning; if an abnormal situation is determined to occur in the interceptor cluster and / or the target swarm during the interceptor cluster's execution of the interception task based on the cooperative path, the cooperative path is updated. This invention improves the real-time performance, security, and adaptability of path planning by adopting an adaptive path planning strategy.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a path planning method for an anti-swarm interceptor cluster. Background Technology

[0002] With the increasing threat of drone swarms, traditional point defense methods are insufficient to cope with highly mobile and high-density swarm attacks, necessitating a shift in the defense paradigm towards coordinated swarm-to-swarm interception. In this process, target trajectory prediction and interceptor collision avoidance path planning are crucial for achieving efficient interception.

[0003] However, existing technologies have significant shortcomings. In trajectory prediction, methods such as Kalman filtering rely excessively on known target motion models; deep learning models require high computational power and are difficult to deploy in practice; local polynomial fitting suffers from rapid error accumulation and poor adaptability when facing highly maneuverable and abruptly changing targets. In path planning, traditional optimization-based methods incur enormous computational costs in high-dimensional spaces with dense obstacles, severely limiting real-time response efficiency; while reinforcement learning-based methods suffer from high training costs, low sample efficiency, and difficulty in ensuring safety in complex scenarios. Summary of the Invention

[0004] This invention provides a path planning method for an anti-swarm interceptor cluster, which addresses the shortcomings of existing anti-swarm interceptor cluster path planning methods in terms of real-time performance, security, and adaptability.

[0005] This invention provides a path planning method for an anti-swarm interceptor cluster, comprising: Based on the location information of the target bee colony within the current time period, the direction vector between the continuous trajectory points of the target bee colony is calculated, and the interception zone of the target bee colony is predicted. Based on the current application scenario, the target planning strategy is adaptively called from the pre-built strategy library to generate a cooperative path for the interceptor cluster. The destination of the cooperative path is within the interception zone. The strategy library includes a lightweight fast response strategy based on height difference collision avoidance, an asymptotic optimal search strategy based on random sampling, and a high-safety optimization strategy based on dynamic grid partitioning. During the process of the interceptor cluster executing the interception task against the target bee colony based on the cooperative path, if it is determined that the interceptor cluster and / or the target bee colony has an abnormal situation, the cooperative path is updated to obtain the updated cooperative path.

[0006] In some embodiments, the step of adaptively invoking a target planning strategy from a pre-built strategy library to generate a collaborative path for the interceptor cluster based on the current application scenario includes: The current position of the interceptors in the interceptor cluster is determined as the starting point, and the corresponding expected interception position within the interception zone is determined as the ending point; In situations where the application scenario indicates a simple environment with sudden maneuvers or strong real-time constraints, the lightweight fast response strategy is invoked to generate the collaborative path based on the starting point and the destination. When the application scenario indicates a complex environment with static or dynamic obstacles, the progressive optimal search strategy is invoked to generate the cooperative path based on the starting point and the ending point. When the application scenario indicates a complex environment with dense obstacles and requirements for collaborative safety and trajectory smoothness, the high-safety optimization strategy is invoked to generate the collaborative path based on the starting point and the ending point.

[0007] In some embodiments, invoking the lightweight fast response strategy to generate the collaborative path based on the starting point and the destination includes: Construct an open list and a closed list, wherein the open list is used to store path nodes to be searched and includes the starting point, and the closed list is used to store path nodes that have been searched; Repeat the following steps until the termination condition of the first iteration is met: If the open list is not empty, based on the pre-set flight altitude layer of the interceptor, the path node closest to the destination is taken from the open list as the current node, and the current node is added to the closed list; If the distance between the current node and the destination is greater than or equal to a preset first distance threshold, then within the spatial neighborhood of the current node, based on the flight altitude layer, select several candidate nodes that are closest to the destination and have not collided with obstacles in the application scenario. These candidate nodes are not in the closed list. Add all the candidate nodes to the open list and record the parent nodes of all the candidate nodes; The first iteration termination condition is that the final distance between the current node and the endpoint is less than the first distance threshold.

[0008] In some embodiments, invoking the asymptotic optimal search strategy to generate the cooperative path based on the starting point and the ending point includes: Initialize the search tree with the starting point as the root node; Repeat the following steps until the second iteration termination condition is met: Within the environmental boundary of the application scenario, sampling points are generated based on the endpoint and a preset probability threshold, and the tree node closest to the sampling point is determined in the search tree; Expand from the nearest tree node toward the sampling point according to a preset step size to obtain a new node; Determine the optimal parent node of the new node within its neighborhood, and optimize the reconnection of existing nodes within the neighborhood. Collision detection is performed on the new path from the nearest tree node to the new node. If the new path has no collision, the new node is added to the search tree. The second iteration termination condition includes: the distance between the final new node and the endpoint is less than a preset second distance threshold.

[0009] In some embodiments, generating sampling points based on the endpoint and a preset probability threshold includes: A random number within a preset range is generated. If the random number is greater than the probability threshold, the endpoint is used as the sampling point. If the random number is less than or equal to the probability threshold, the sampling point is randomly generated within the environmental boundary range.

[0010] In some embodiments, determining the optimal parent node of the new node within its neighborhood includes: A spherical neighborhood with a preset radius is defined with the new node as the center. Each node in the search tree located within the spherical neighborhood is determined as a candidate parent node. Based on each candidate parent node, multiple paths from the starting point to the new node are determined, and each path contains one candidate parent node. Based on a pre-constructed cost function, the path cost of each path is calculated; the cost function is constructed by combining path length and smoothness factors. The candidate parent node contained in the path with the minimum path cost is determined as the optimal parent node, and the new node is connected to the optimal parent node.

[0011] In some embodiments, performing collision detection on the new path from the nearest tree node to the new node includes: The new path is discretized into multiple path points with a fixed length; Based on the location information of all the path points, a preset list of static obstacles is queried, the list of static obstacles including the location information of multiple static obstacles; If it is determined that any of the path points is occupied by the static obstacle, it is determined that a collision has occurred on the new path; If it is determined that none of the path points are occupied by the static obstacles, the new path is considered to be collision-free.

[0012] In some embodiments, invoking the high-security optimization strategy to generate the cooperative path based on the starting point and the ending point includes: Based on the starting point and the ending point, a local effective area is determined from the map covering the task area; Adaptive grid division is performed on the aforementioned effective local area to generate a dynamic local map; With the goal of minimizing path cost, a path search is performed in the three-dimensional grid of the dynamic local map to generate the cooperative path; During the path search process, obstacle avoidance detection is performed, and all grids traversed by the planned path are marked as impassable.

[0013] In some embodiments, the abnormal condition includes at least one of the following: The interceptor cluster's task initialization failed; The actual trajectory of the target bee colony deviates from the expected interception position within the interception zone, and the expected interception position is determined based on the predicted trajectory of the target bee colony. The estimated time for the interceptor cluster to reach the expected interception location is later than the estimated time for the target bee colony to reach the expected interception location; The target bee colony appeared to be hovering.

[0014] In some embodiments, calculating the direction vector between continuous trajectory points of the target bee colony based on its location information within the current time period, and predicting the interception zone of the target bee colony, includes: Obtain the location information of each trajectory point of the target bee colony within the current time period; Select any two consecutive trajectory points from each of the trajectory points to form a trajectory point pair, and calculate the direction vector of each trajectory point pair; The weight of the direction vector for each pair of trajectory points is determined based on preset rules, including that the weight of the direction vector closest to the current time is greater than the weight of the direction vector furthest from the current time. Based on the weight of the direction vector of each of the trajectory point pairs, the direction vectors of all the trajectory point pairs are fused to obtain the predicted direction of the target bee colony. The interception zone is predicted based on the predicted direction and current speed of the target bee colony, and the current speed of the interceptor cluster.

[0015] This invention provides a path planning method for an anti-swarm interceptor cluster. Based on the target swarm's location information within the current time period, it calculates the direction vector between continuous trajectory points of the target swarm to predict its interception zone. According to the application scenario at the current moment, it adaptively calls target planning strategies from a pre-built strategy library to generate a cooperative path for the interceptor cluster. The endpoint of the cooperative path is within the interception zone. The strategy library includes a lightweight, fast-response strategy based on height difference collision avoidance, a progressive optimal search strategy based on random sampling, and a high-safety optimization strategy based on dynamic grid partitioning. During the interceptor cluster's execution of interception tasks against the target swarm based on the cooperative path, if an anomaly is detected in the interceptor cluster and / or the target swarm, the cooperative path is updated, improving the real-time performance, security, and adaptability of path planning. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the path planning method for the anti-swarm interceptor cluster provided in this embodiment of the invention.

[0018] Figure 2 This is a schematic diagram showing the deviation of the actual trajectory of the target bee colony from the expected interception position within the interception zone, as provided in an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram of the process for determining the optimal parent node of a new node within the neighborhood of a new node, provided by an embodiment of the present invention.

[0020] Figure 4 This is a schematic diagram of the process for optimizing the reconnection of existing nodes in the neighborhood, provided by an embodiment of the present invention.

[0021] Figure 5 This is a schematic diagram of the collision detection process based on the time series distance detection method provided in the embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] Figure 1 This is a flowchart illustrating the path planning method for an anti-swarm interceptor cluster provided in an embodiment of the present invention. Figure 1 As shown, a path planning method for an anti-swarm interceptor cluster is provided, including the following steps: step 110, step 120, and step 130. This method's steps are merely one possible implementation of the present invention.

[0024] Step 110: Based on the location information of the target bee colony within the current time period, calculate the direction vector between the continuous trajectory points of the target bee colony and predict the interception zone of the target bee colony.

[0025] In this context, a target swarm can refer to one or more enemy or unidentified drones moving in a coordinated, clustered manner. The location information of the target swarm can come from radar, electro-optical sensors, or other external reconnaissance systems, and is typically represented as a series of discrete three-dimensional spatial coordinate points, i.e., trajectory points, collected within the current time period.

[0026] Optionally, the acquired raw trajectory data can be processed to extract its motion trend. One specific calculation method is to select two temporally adjacent trajectory points and obtain a vector representing the direction and magnitude of motion within that short time interval through vector subtraction. By performing this calculation on multiple pairs of consecutive trajectory points, a set of direction vectors reflecting the recent motion characteristics of the target bee colony can be obtained. The purpose of this step is to predict the future movement of the bee colony without a model by analyzing historical trajectories, avoiding the dependence of traditional methods on specific motion models.

[0027] Optionally, based on the direction vectors, an airspace range in which the interceptor cluster and the target swarm may encounter each other within a future time window is calculated, i.e., the interception zone. The interception zone is a three-dimensional spatial region, rather than a precise point, which provides fault tolerance and flexibility for subsequent path planning. The prediction process can involve fusing all calculated direction vectors, for example, through weighted averaging, to obtain a comprehensive predicted direction. This prediction direction is then combined with information such as the current position and velocity of the target swarm to extrapolate one or more possible future positions, thereby defining the interception zone.

[0028] In some embodiments, step 110 calculates the direction vector between continuous trajectory points of the target bee colony based on the location information of the target bee colony within the current time period, and predicts the interception zone of the target bee colony, including: Step 111: Obtain the location information of each trajectory point of the target bee colony within the current time period.

[0029] Optionally, the location information includes a filtered set of trajectory points: ; ; in, Represents a set of trajectory points. Indicates the total number of trajectory points. Indicates the first A trajectory point, Indicates the first The three-dimensional spatial coordinates of the trajectory points express X-axis coordinates express Y-axis coordinate, express The Z-axis coordinate.

[0030] Step 112: Select any two consecutive trajectory points from each trajectory point to form a trajectory point pair, and calculate the direction vector of each trajectory point pair.

[0031] Optionally, trajectory point pairs can be formed by selecting consecutive trajectory points. Calculate the direction vector between each pair of trajectory points sequentially: ; ; in, express The trajectory points at any given moment express The trajectory points at any given moment express The direction vector at time, i.e., the trajectory point and The direction vector between them express Three-dimensional spatial coordinates, express X-axis coordinates express Y-axis coordinate, express Z-axis coordinate, express Three-dimensional spatial coordinates, express X-axis coordinates express Y-axis coordinate, express The Z-axis coordinate.

[0032] Step 113: Determine the weight of the direction vector for each trajectory point pair based on preset rules. The preset rules include: the weight of the direction vector closest to the current time is greater than the weight of the direction vector farther away from the current time.

[0033] Step 114: Based on the weight of the direction vector of each trajectory point pair, fuse the direction vectors of all trajectory point pairs to obtain the predicted direction.

[0034] Optionally, based on the weights of the direction vectors of each trajectory point pair, the direction vectors of all trajectory point pairs are fused to obtain a fused direction vector. Based on the fused direction vector, the predicted direction is obtained. ; ; ; in, Represents the fusion direction vector. Represents trajectory points and The weights of the direction vectors between them. Indicates the predicted direction.

[0035] Step 115: Based on the predicted direction and current speed of the target bee colony, and the current speed of the interceptor cluster, predict the interception zone.

[0036] Optionally, the interception zone is predicted based on the predicted direction, current speed, and current position of the target bee colony, as well as the current speed and current position of the interceptor cluster.

[0037] Optionally, the current trajectory point The formula for calculating the predicted location is as follows: ; in, Indicates the current trajectory point exist Predicted location at time Indicates the current speed of the target bee colony. This indicates the time interval, i.e., the prediction step size.

[0038] Step 120: Based on the application scenario at the current moment, adaptively call the target planning strategy from the pre-built strategy library to generate the cooperative path of the interceptor cluster. The destination of the cooperative path is within the interception zone. The strategy library includes a lightweight fast response strategy based on height difference collision avoidance, an asymptotic optimal search strategy based on random sampling, and a high-safety optimization strategy based on dynamic grid partitioning.

[0039] The application scenario refers to the specific environment in which the interception task is executed. Application scenarios can include, but are not limited to, the density of obstacles, the complexity of the terrain, the sufficiency of system computing resources, the mobility of the target, and real-time and security requirements. The system needs to perceive or assess the complexity of the current application scenario in real time.

[0040] The interceptor swarm refers to a group of friendly drones performing interception missions. The coordinated path is a set of coordinated and conflict-free flight trajectories planned for each drone in the interceptor swarm. The endpoints of these paths are all located within the predicted interception zone, ensuring that all interceptors converge towards the same target area.

[0041] The strategy library is a pre-built collection of algorithms that includes various path planning algorithms for different application scenarios.

[0042] In some embodiments, step 120 adaptively invokes a target planning strategy from a pre-built strategy library to generate a collaborative path for the interceptor cluster based on the application scenario at the current moment, including: Step 121: Determine the current position of the interceptors in the interceptor cluster as the starting point, and determine the corresponding expected interception position within the interception zone as the ending point.

[0043] Step 122: In situations where the application scenario indicates a simple environment with sudden maneuvers or strong real-time constraints, a lightweight fast response strategy is invoked to generate a collaborative path based on the starting point and the destination.

[0044] Here, "simple environment" refers to low terrain complexity, such as terrain complexity below a preset threshold. The lightweight, fast-response strategy based on height difference collision avoidance is primarily suitable for scenarios with simple environments, sparse obstacles, or extremely stringent response time requirements. Its core idea is to use a lightweight greedy algorithm based on height difference collision avoidance to quickly generate a feasible path, prioritizing timeliness. The collision avoidance method is relatively simple, for example, by presetting different height layers to avoid collisions.

[0045] Step 123: In cases where the application scenario indicates a complex environment with static or dynamic obstacles, invoke the progressive optimal search strategy to generate a collaborative path based on the starting point and the destination.

[0046] Among these, "environmental complexity" refers to high terrain complexity, such as terrain complexity exceeding a preset threshold. The asymptotically optimal search strategy based on random sampling is suitable for scenarios with many obstacles and complex environments. It employs the Rapidly-exploring Random Tree Star (RRT) algorithm. By randomly sampling in space to explore feasible paths, a good balance can be achieved between planning efficiency and path quality, and the solution gradually converges to the optimal solution.

[0047] Step 124: In application scenarios where the environment is complex, there are dense obstacles, and there are requirements for collaborative safety and trajectory smoothness, a high-safety optimization strategy is invoked to generate a collaborative path based on the starting point and the ending point.

[0048] High-safety optimization strategies based on dynamic mesh partitioning are suitable for complex environments with extremely high requirements for path safety, smoothness, and cooperative collision avoidance among interceptors. They typically employ a Multi-Verse Optimizer (MVO) algorithm to perform detailed environmental modeling, finding the optimal path while ensuring absolute safety.

[0049] Optionally, based on the analysis results of the current application scenario, the most suitable strategy can be adaptively selected from the strategy library. For example, if it is determined to be an emergency interception in an open environment, a lightweight fast response strategy is invoked; if it is determined to be a routine interception in a complex environment, an asymptotic optimal search strategy is invoked. After selecting a strategy, the system executes the strategy with the current position of each interceptor as the starting point and one or more positions within the interception zone as the ending point, thereby generating a set of safe and feasible collaborative paths for the entire cluster.

[0050] Step 130: During the process of the interceptor cluster executing the interception task against the target bee colony based on the cooperative path, if it is determined that the interceptor cluster and / or the target bee colony has an abnormal situation, the cooperative path is updated to obtain the updated cooperative path.

[0051] An interception mission refers to the entire process by which an interceptor cluster starts from its origin, flies along a coordinated path, and reaches the interception zone to intercept the target bee swarm.

[0052] Anomalies refer to any unexpected events that may cause the current cooperative path to fail or become suboptimal. Anomalies can include problems with the interceptor cluster itself, such as a malfunction of an interceptor or failure to reach the predetermined location on time; they can also include changes in the behavior of the target bee colony, such as a significant deviation from the prediction in its flight trajectory, or a sudden change in its movement pattern.

[0053] In some embodiments, the abnormal condition includes at least one of the following: Interceptor cluster task initialization failed; The actual trajectory of the target bee colony deviates from the expected interception position within the interception zone, which is determined based on the predicted trajectory of the target bee colony. The estimated time for the interceptor cluster to reach the expected interception location is later than the estimated time for the target bee colony to reach the expected interception location; The target bee colony is hovering.

[0054] Optionally, the status of the interceptor cluster and the target bee colony is continuously monitored during the task execution. When the deviation between the actual status and the planned expectation is detected to exceed a preset threshold, it is determined that an anomaly has occurred.

[0055] Optionally, once an anomaly is identified, a replanning process is immediately triggered. This involves re-predicting the interception zone, reassessing the application scenario, and invoking an appropriate planning strategy to generate a new collaborative path for the interceptor cluster that adapts to the current situation. The interceptor cluster then abandons the old path and executes the updated collaborative path, thereby ensuring effective interception of the target.

[0056] Specifically, if any interceptor in the interceptor cluster fails to meet the basic startup conditions within the specified time at the start of the mission, the mission initialization is deemed a failure, triggering a replanning process. This includes two typical scenarios: first, the interceptor fails to take off and reach the designated altitude within the set time; second, the interceptor fails to complete the unlocking operation within the specified time after startup. This type of replanning aims to reconfigure the mission start point or activate a backup plan to ensure the mission can continue.

[0057] Figure 2 This is a schematic diagram illustrating the deviation of the actual trajectory of the target bee colony from the expected interception position within the interception zone, as provided in an embodiment of the present invention. Figure 2 As shown, by collecting real-time flight path data of the target swarm of drones, the intersection point of its predicted trajectory and the interception zone, i.e., the escape point, can be calculated, and the escape point can be used as the expected interception location. To avoid misjudgment due to errors in escape point estimation, a dual verification mechanism is used for deviation determination: First, the vertical distance h from the target drone's current position to the line connecting the starting point and the escape point is calculated. If this distance exceeds a set threshold, mechanism one is triggered. Second, the angle θ between the target drone's current flight direction and the predetermined interception direction is detected. If the angle deviation exceeds a threshold, mechanism two is triggered. When both mechanisms are triggered simultaneously, it is determined as a trajectory deviation, and the system will replan the interception path to improve adaptability to the target's maneuvering flight.

[0058] Specifically, based on the real-time position, speed, and flight path information of both the interceptor swarm and the target drone swarm, the estimated time to reach the expected interception position is calculated. If the estimated arrival time of the interceptor is later than that of the target drone, it is determined that the resource scheduling is lagging. At this time, the system will initiate replanning, adjust the flight path or speed of the interceptor, or select a better interception point to ensure the effectiveness of the interception mission.

[0059] Specifically, if the target drone suddenly enters a hovering state when the interceptor approaches the interception zone, the original interception point will become invalid. At this time, the system will immediately trigger replanning, regenerating an encirclement or close-range interception path based on the hovering position of the target drone, thus achieving a rapid response to stationary or low-speed targets.

[0060] Understandably, through the above replanning trigger mechanism, the system can respond promptly to various situations such as task initialization anomalies, target trajectory deviations, resource scheduling delays, and sudden changes in target behavior, dynamically adjust the interception scheme, and significantly improve the adaptability and success rate of task execution.

[0061] In this embodiment of the invention, the ability to adapt to unknown maneuvering targets is improved by predicting the interception zone without model dependence; by constructing a strategy library containing multiple strategies and adaptively calling them according to the application scenario, the real-time performance, optimality and security of path planning are taken into account, and it can flexibly cope with combat environments of different complexities; by introducing a dynamic replanning mechanism under abnormal conditions, the robustness and mission success rate of the entire interception system in complex and ever-changing environments are significantly improved.

[0062] In some embodiments, step 122 invokes a lightweight fast response strategy to generate a collaborative path based on the origin and destination, including: Construct an open list and a closed list. The open list stores path nodes to be searched and includes the starting point, while the closed list stores path nodes that have already been searched. Repeat the following steps until the termination condition of the first iteration is met: If the open list is not empty, based on the pre-set flight altitude layer of the interceptor, the path node closest to the destination is taken from the open list as the current node, and the current node is added to the closed list; If the distance between the current node and the destination is greater than or equal to a preset first distance threshold, then within the spatial neighborhood of the current node, select several candidate nodes that are closest to the destination and have not collided with obstacles in the application scenario, based on the flight altitude layer. These candidate nodes are not in the closed list. Add all candidate nodes to the open list and record the parent nodes of all candidate nodes; The first iteration terminates when the distance between the current node and the endpoint is less than a first distance threshold.

[0063] In this embodiment of the invention, a lightweight greedy algorithm based on altitude difference collision avoidance is adopted, which can generate feasible paths in milliseconds, ensuring basic safety and extremely fast response. The specific greedy algorithm process is as follows: S1, Input: Start point s, End point t, Height layer L, Occupancy table occ; S2, Output: Path P; S3. Define an open list Open={s}, a closed list Closed={}, and a parent list Parent={}; S4. Execute when the open list Open is not empty: Extract the node c that is closest to t in Open; If c is in Closed, skip that node; otherwise, add c to Closed. If the planar distance is less than the threshold, jump to S5; Select the three points n that are closest to t and have not collided from the eight neighborhoods, and n is not in Closed; For each n: Set the parent node of n to c, Parent[n] = c; Add n to the Open list; S5. Path P is obtained by tracing back the parent chain from the endpoint.

[0064] In some embodiments, step 123 invokes an asymptotic optimal search strategy to generate a cooperative path based on the start and end points, including: Initialize the search tree with the starting point as the root node; Repeat the following steps until the second iteration termination condition is met: Within the environmental boundary of the application scenario, sampling points are generated based on the endpoint and a preset probability threshold, and the tree node closest to the sampling point is determined in the search tree; Expand from the nearest tree node toward the sampling point according to a preset step size to obtain a new node; Determine the optimal parent node of the new node within its neighborhood and optimize the reconnection of existing nodes within the neighborhood. Perform collision detection on the new path from the nearest tree node to the new node. If the new path has no collision, add the new node to the search tree. The second iteration termination condition includes: the distance between the final new node and the endpoint is less than a preset second distance threshold.

[0065] The specific process is as follows: Step 1: Initialize the environment ,starting point ,end Starting point For the search tree rooted at node `tree`, initialize the algorithm parameters, such as `T`, where `T` is the maximum number of iterations.

[0066] Step 2: Generate sampling points within the environmental boundary using a bias strategy. .

[0067] Step 3: Traverse all nodes in the tree and select the nodes closest to the sampling point. The nearest tree node They are called nearest neighbors.

[0068] Step 4, to avoid adjacent points and sampling points Excessive distance between nodes leads to low exploration efficiency; therefore, a step size (step) is set. If the distance between a neighboring point and the sampling point is less than the step size (step), the sampling point is set as a new node. If the distance between a neighboring point and the sampling point is greater than or equal to the step size (step), then expand the node by a length (step) from the neighboring point toward the sampling point to obtain a new node. .

[0069] Step 5: For the new path - Perform collision detection with the environment; if a collision exists, cancel the operation. Repeat steps 2-4.

[0070] Step 6, Judgment Check if the destination has been reached. If so, a collision-free path has been found, and the search stops. Otherwise, repeat steps 2 to 5 until the maximum number of iterations is reached, and then output the result.

[0071] In some embodiments, generating sampling points based on an endpoint and a preset probability threshold includes: Generate a random number within a preset range. If the random number is greater than the probability threshold, use the endpoint as the sampling point; if the random number is less than or equal to the probability threshold, randomly generate sampling points within the environmental boundary range.

[0072] Optionally, a bias strategy can be incorporated into the sampling process to optimize the sampling selection method, allowing the sampling results to be selected with a certain probability. Extend towards the endpoint to accelerate RRT. The speed at which the algorithm finds its track, after introducing a bias strategy Select as follows: ; In the formula, For sampling points, Let [a] be a random number uniformly distributed between [0, 1]. Greater than the set probability threshold At that time, set the target point For sampling points, when Less than or equal to the probability threshold At that time, random sampling was performed within the boundary area of ​​the environment. This represents a random sampling point.

[0073] In some embodiments, determining the optimal parent node of the new node within its neighborhood includes: Define a spherical neighborhood with a preset radius, centered on the new node. Each node in the spherical neighborhood of the search tree is identified as a candidate parent node. Based on each candidate parent node, multiple paths from the starting point to the new node are determined, and each path contains a candidate parent node. Based on a pre-built cost function, the path cost of each path is calculated separately; the cost function is constructed by combining path length and smoothness factors. The candidate parent node contained in the path with the minimum path cost is determined as the optimal parent node, and the new node is connected to the optimal parent node.

[0074] Figure 3 This is a schematic diagram illustrating the process of determining the optimal parent node of a new node within its neighborhood, as provided in an embodiment of the present invention. Figure 3 As shown, node 9 is a new node generated by node 8, corresponding to the steps described above. and To reselect a parent node for node 9, the original path is 0-6-8-9 with a path cost of 16. At this point, a spherical region is defined with node 9 as the center. The three nodes 5, 6, and 7, excluding the original parent node 8, are selected as candidate parent nodes for node 9. The path cost between the initial node 0 and the new node 9 after connecting the three candidate parent nodes is calculated and compared with the initial path cost. The path with the lowest cost is selected and the parent node is changed. Figure 3 The path cost of path 0-2-5-9 calculated using candidate parent node 5 is 11, the path cost of path 0-6-9 calculated using candidate parent node 6 is 14, and the path cost of path 0-2-5-7-9 calculated using candidate parent node 7 is 12. Therefore, node 5 is reselected as the path cost. The parent node is broken, so the connection between nodes 8 and 9 is severed, and nodes 5-9 are connected.

[0075] Figure 4 This is a schematic diagram illustrating the process of optimizing reconnection of existing nodes within a neighborhood, provided by an embodiment of the present invention. After reselecting the parent node, node 9 still corresponds to the steps described above. The original path is 0-2-5-9. The three nodes 6, 7, and 8, excluding the original parent node 5, are rerouted. Node 9 is used as the parent node of nodes 6, 7, and 8, respectively, forming new paths: 0-2-5-9-6, 0-2-5-9-8, and 0-2-5-9-7. These are compared with the original paths 0-6, 0-6-8, and 0-2-5-7 without rerouting. If the path cost of the new path is lower, node 9 is selected as the parent node of the corresponding node. Figure 4 Only the path 0-2-5-9-8 has a path cost of 12, which is less than the path cost of 15 for 0-6-8. Therefore, node 9 is chosen as the parent node of node 8, that is, the connection between node 6 and node 8 is broken, and the connection between node 9 and node 8 is connected.

[0076] In some embodiments, collision detection is performed on the new path from the nearest tree node to the new node, including: Discretize the new path into multiple path points with a fixed length; Based on the location information of all waypoints, query the preset list of static obstacles. The list of static obstacles includes the location information of multiple static obstacles. If any path point is found to be occupied by a static obstacle, a collision is determined to occur on the new path; If it is determined that no path points are occupied by static obstacles, the new path is considered to have no collision.

[0077] At the obstacle avoidance strategy level, for static obstacles, the method of discretizing the path into a set of points and querying the obstacle status of each point in the map index is used for obstacle avoidance detection.

[0078] Figure 5 This is a schematic diagram illustrating the collision detection process based on time-series distance detection, as provided in an embodiment of the present invention. Figure 5 As shown, for dynamic obstacle avoidance between drones, a time-series distance detection method is adopted, which predicts collisions by introducing a three-time window. This method not only detects whether the drones meet at the planned time t, but also additionally detects the positional relationship between time t and the adjacent times t-1 and t+1, thereby constructing a dynamic safety buffer and effectively identifying and avoiding potential collision risks caused by path intersections.

[0079] In some embodiments, step 124 invokes a high-security optimization strategy to generate a cooperative path based on the start and end points, including: Based on the start and end points, determine the local effective area from the map covering the task area; Adaptive grid division is performed on the effective local area to generate a dynamic local map; With the goal of minimizing path cost, a path search is performed in the 3D grid of a dynamic local map to generate cooperative paths; During the path search process, obstacle avoidance detection is performed, and all grids traversed by the planned path are marked as impassable.

[0080] Optionally, a high-security optimization strategy can be invoked, employing the MVO algorithm for path search. This method uses A Search-based collision avoidance is achieved through conflict resolution and path optimization within a safe neighborhood, and trajectory grid occupancy prediction, enabling highly secure collaborative collision avoidance. Based on 3D A... The algorithm employs a grid-by-grid exploration strategy in a grid map. Its search mechanism uses the classic cost function f(n) = g(n) + h(n), where g(n) represents the actual cost from the starting point to node n, and h(n) represents the heuristically estimated cost from node n to the destination. The algorithm performs a grid search along the x, y, and z dimensions, optimizing the search process by maintaining open and closed lists.

[0081] In terms of obstacle avoidance strategy, obstacle detection is achieved by directly detecting the obstacle occupancy status of each grid. All grids traversed by the previously planned path are marked as impassable, ensuring that subsequent paths are completely spatially isolated from historical paths, thereby providing a higher level of safety guarantee and achieving inter-machine obstacle avoidance. This method effectively avoids potential path conflict risks through strict management of spatial resources.

[0082] RRT When performing path planning, MVO requires a map that includes terrain and obstacles. For path planning problems covering large task areas, a dynamic map segmentation strategy is adopted: a) In the offline phase, a global high-precision map covering the entire task area with uniform resolution is pre-constructed based on map data, and the map is stored in the ground computing node in the form of a layered raster. b) During the online phase, the starting point set of the current batch of interceptors is obtained in real time. And the interception point set {Ij} obtained through interception calculation, based on which it can be calculated to encompass the entire The minimum bounding rectangle R is defined, and the sides of the rectangle are extended outward by a preset safety distance d to form the local effective region for this task. ; c) In the local effective region Internal adaptive mesh remapping: First, perform adaptive mesh remapping... The entire area is divided into fixed, uniform grids. Then, within each grid, the average elevation of the largest connected region is used as the height of that grid, ultimately generating a dynamic local map. .

[0083] d) Dynamic local map As RRT Alternatively, the search space of the MVO algorithm can be used to complete online path planning.

[0084] By first creating a high-precision global map, then performing task-driven cropping, and finally adaptive precision allocation, this embodiment of the invention ensures that no obstacle or terrain details are missed while also taking into account planning accuracy and real-time performance.

[0085] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A path planning method for an anti-swarm interceptor cluster, characterized in that, include: Based on the location information of the target bee colony within the current time period, the direction vector between the continuous trajectory points of the target bee colony is calculated, and the interception zone of the target bee colony is predicted. Based on the current application scenario, the target planning strategy is adaptively called from the pre-built strategy library to generate a cooperative path for the interceptor cluster. The destination of the cooperative path is within the interception zone. The strategy library includes a lightweight fast response strategy based on height difference collision avoidance, an asymptotic optimal search strategy based on random sampling, and a high-safety optimization strategy based on dynamic grid partitioning. During the process of the interceptor cluster executing the interception task against the target bee colony based on the cooperative path, if it is determined that the interceptor cluster and / or the target bee colony has an abnormal situation, the cooperative path is updated to obtain the updated cooperative path.

2. The path planning method for the anti-swarm interceptor cluster according to claim 1, characterized in that, The step of adaptively invoking target planning strategies from a pre-built strategy library to generate a collaborative path for the interceptor cluster based on the current application scenario includes: The current position of the interceptors in the interceptor cluster is determined as the starting point, and the corresponding expected interception position within the interception zone is determined as the ending point; In situations where the application scenario indicates a simple environment with sudden maneuvers or strong real-time constraints, the lightweight fast response strategy is invoked to generate the collaborative path based on the starting point and the destination. When the application scenario indicates a complex environment with static or dynamic obstacles, the progressive optimal search strategy is invoked to generate the cooperative path based on the starting point and the ending point. When the application scenario indicates a complex environment with dense obstacles and requirements for collaborative safety and trajectory smoothness, the high-safety optimization strategy is invoked to generate the collaborative path based on the starting point and the ending point.

3. The path planning method for the anti-swarm interceptor cluster according to claim 2, characterized in that, The step of invoking the lightweight fast response strategy to generate the collaborative path based on the starting point and the ending point includes: Construct an open list and a closed list, wherein the open list is used to store path nodes to be searched and includes the starting point, and the closed list is used to store path nodes that have been searched; Repeat the following steps until the termination condition of the first iteration is met: If the open list is not empty, based on the pre-set flight altitude layer of the interceptor, the path node closest to the destination is taken from the open list as the current node, and the current node is added to the closed list; If the distance between the current node and the destination is greater than or equal to a preset first distance threshold, then within the spatial neighborhood of the current node, based on the flight altitude layer, select several candidate nodes that are closest to the destination and have not collided with obstacles in the application scenario. These candidate nodes are not in the closed list. Add all the candidate nodes to the open list and record the parent nodes of all the candidate nodes; The first iteration termination condition is that the final distance between the current node and the endpoint is less than the first distance threshold.

4. The path planning method for the anti-swarm interceptor cluster according to claim 2, characterized in that, The step of invoking the asymptotic optimal search strategy to generate the cooperative path based on the starting point and the ending point includes: Initialize the search tree with the starting point as the root node; Repeat the following steps until the second iteration termination condition is met: Within the environmental boundary of the application scenario, sampling points are generated based on the endpoint and a preset probability threshold, and the tree node closest to the sampling point is determined in the search tree; Expand from the nearest tree node toward the sampling point according to a preset step size to obtain a new node; Determine the optimal parent node of the new node within its neighborhood, and optimize the reconnection of existing nodes within the neighborhood. Collision detection is performed on the new path from the nearest tree node to the new node. If the new path has no collision, the new node is added to the search tree. The second iteration termination condition includes: the distance between the final new node and the endpoint is less than a preset second distance threshold.

5. The path planning method for the anti-swarm interceptor cluster according to claim 4, characterized in that, The generation of sampling points based on the endpoint and a preset probability threshold includes: A random number within a preset range is generated. If the random number is greater than the probability threshold, the endpoint is used as the sampling point. If the random number is less than or equal to the probability threshold, the sampling point is randomly generated within the environmental boundary range.

6. The path planning method for the anti-swarm interceptor cluster according to claim 4, characterized in that, Determining the optimal parent node of the new node within its neighborhood includes: A spherical neighborhood with a preset radius is defined with the new node as the center. Each node in the search tree located within the spherical neighborhood is determined as a candidate parent node. Based on each candidate parent node, multiple paths from the starting point to the new node are determined, and each path contains one candidate parent node. Based on a pre-constructed cost function, the path cost of each path is calculated; the cost function is constructed by combining path length and smoothness factors. The candidate parent node contained in the path with the minimum path cost is determined as the optimal parent node, and the new node is connected to the optimal parent node.

7. The path planning method for the anti-swarm interceptor cluster according to claim 4, characterized in that, The collision detection of the new path from the nearest tree node to the new node includes: The new path is discretized into multiple path points with a fixed length; Based on the location information of all the path points, a preset list of static obstacles is queried, the list of static obstacles including the location information of multiple static obstacles; If it is determined that any of the path points is occupied by the static obstacle, it is determined that a collision has occurred on the new path; If it is determined that none of the path points are occupied by the static obstacles, the new path is considered to be collision-free.

8. The path planning method for the anti-swarm interceptor cluster according to claim 2, characterized in that, The step of invoking the high-security optimization strategy to generate the cooperative path based on the starting point and the ending point includes: Based on the starting point and the ending point, a local effective area is determined from the map covering the task area; Adaptive grid division is performed on the aforementioned effective local area to generate a dynamic local map; With the goal of minimizing path cost, a path search is performed in the three-dimensional grid of the dynamic local map to generate the cooperative path; During the path search process, obstacle avoidance detection is performed, and all grids traversed by the planned path are marked as impassable.

9. The path planning method for an anti-swarm interceptor cluster according to claim 1, characterized in that, The abnormal situation includes at least one of the following: The interceptor cluster's task initialization failed; The actual trajectory of the target bee colony deviates from the expected interception position within the interception zone, and the expected interception position is determined based on the predicted trajectory of the target bee colony. The estimated time for the interceptor cluster to reach the expected interception location is later than the estimated time for the target bee colony to reach the expected interception location; The target bee colony appeared to be hovering.

10. The path planning method for an anti-swarm interceptor cluster according to claim 1, characterized in that, The step of calculating the direction vector between continuous trajectory points of the target bee colony based on its location information within the current time period, and predicting the interception zone of the target bee colony, includes: Obtain the location information of each trajectory point of the target bee colony within the current time period; Select any two consecutive trajectory points from each of the trajectory points to form a trajectory point pair, and calculate the direction vector of each trajectory point pair; The weight of the direction vector for each pair of trajectory points is determined based on preset rules, including that the weight of the direction vector closest to the current time is greater than the weight of the direction vector furthest from the current time. Based on the weight of the direction vector of each of the trajectory point pairs, the direction vectors of all the trajectory point pairs are fused to obtain the predicted direction of the target bee colony. The interception zone is predicted based on the predicted direction and current speed of the target bee colony, and the current speed of the interceptor cluster.

Images