Unmanned aerial vehicle cluster distributed area search method and system based on global connectivity

Abstract

The application provides a kind of global connectivity-based unmanned aerial vehicle cluster distributed area search method and system, initialize unmanned aerial vehicle cooperative search task environment;Unmanned aerial vehicle updates cluster search cognitive map;Multi-unmanned aerial vehicle cooperative search task allocation modeling;CBAA-based cooperative search pre-allocation;Global connectivity-based cooperative search redistribution;Collision time cooperative guidance-based unmanned aerial vehicle waypoint tracking control;Output unmanned aerial vehicle cluster cooperative search result.The application uses CBAA to realize distributed decision-making among multiple unmanned aerial vehicles, relies on its best suboptimal solution characteristics in multi-agent task allocation problem to obtain the suboptimal action strategy of unmanned aerial vehicle;And introduce disconnected individual removal, action selection taboo list and mobile base station switching strategy, so that CBAA can effectively exclude the action strategy of disconnected individual, and by releasing the action space of decision lag individual, the unmanned aerial vehicle cluster can effectively improve search efficiency and coverage performance under the premise of ensuring global connectivity.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) target search technology, and in particular to a distributed regional search method and system for UAV swarms based on global connectivity. Background Technology

[0002] Area search, as a core application scenario and research focus of unmanned aerial vehicle (UAV) systems, is crucial for supporting UAVs in completing complex tasks such as area reconnaissance, communication relay, and emergency rescue. A swarm system formed by multiple UAVs working collaboratively can significantly improve the search efficiency of a single UAV. Currently, UAV swarms mainly adopt three organizational architectures: centralized, decentralized, and distributed. Given the insufficient robustness and scalability of centralized architectures, and the susceptibility of decentralized architectures to decision-making conflicts, distributed architecture has become the mainstream design solution.

[0003] While distributed architectures are favored in drone swarm applications, limitations in the drones' perception, communication, and mobility make it difficult to achieve optimal search path planning in mission environments lacking global information. Relying solely on local perception and interaction is insufficient. In practical applications, drones often get bogged down in repetitive searches of local areas due to detection noise interference. Furthermore, limited communication range forces drones to maintain global connectivity of the communication topology when exploring unknown areas, highlighting the conflict between search coverage and connectivity maintenance.

[0004] To address these challenges, current research often incorporates connectivity-preserving control strategies from multi-agent systems theory. However, these traditional methods often sacrifice search efficiency and UAV maneuverability to ensure connectivity. When the mission area has complex terrain and numerous obstacles, overly conservative connectivity-preserving strategies not only severely compress the UAV's action space and exacerbate local optima problems, but also easily lead to decision-making conflicts among swarm members, resulting in impaired connectivity, search stagnation, or even collisions. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a distributed region search method and system for UAV swarms based on global connectivity. By equating the region search problem to a discrete grid allocation problem for UAV swarms, the CBAA algorithm is used to obtain suboptimal action strategies for UAVs. Simultaneously, disconnected individual removal, action selection taboo tables, and mobile base station handover strategies are introduced to enable CBAA to effectively eliminate action strategies of disconnected individuals. Furthermore, by releasing the action space of individuals with decision lag, the UAV swarm can consistently find suboptimal action strategies while maintaining global connectivity, thereby maximizing search efficiency and coverage performance.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a distributed region search method for UAV swarms based on global connectivity, comprising:

[0008] Initialize the UAV cooperative search mission environment, including the UAV kinematic model, detection probability model, and cooperative search cognitive map;

[0009] Based on the detection probability model, Bayes' theorem is used to obtain posterior observation information to calculate the probability estimate of the fused target and update the collaborative search cognitive map;

[0010] Based on the UAV kinematics model and the updated cooperative search cognitive map, we update the action space, model the equivalent task allocation problem and design the objective function to construct a task allocation model for UAV swarm cooperative search.

[0011] Based on the task allocation model, collaborative search pre-allocation is performed according to the consensus auction algorithm, the bidding value is calculated, and the task allocation list is updated through consensus interaction.

[0012] Based on the task allocation list, global connectivity is used for collaborative search and reallocation. Mobile base stations are identified and connected subgraphs are generated through depth-first search. Cluster search action strategies that satisfy connectivity constraints are output.

[0013] The system controls UAV waypoint tracking based on collision time guidance, calculates segmented velocity commands and normal acceleration commands, and outputs collaborative search results for UAV swarms, including probabilistic maps, deterministic maps, and target lists.

[0014] Secondly, the present invention provides a globally connected UAV swarm distributed area search system, comprising:

[0015] The initialization module is configured to initialize the UAV cooperative search mission environment, including the UAV kinematic model, the detection probability model, and the cooperative search cognitive map.

[0016] The map update module is configured to use Bayes' theorem to obtain posterior observation information based on the detection probability model, calculate the probability estimate of the fused target, and update the collaborative search cognitive map.

[0017] The task construction module is configured to update the action space, model the equivalent task allocation problem, and design the objective function based on the UAV kinematic model and the updated cooperative search cognitive map, and construct the task allocation model for UAV swarm cooperative search.

[0018] The information allocation module is configured to perform collaborative search pre-allocation based on the task allocation model and the consensus auction algorithm, calculate the bidding value, and update the task allocation list through consensus interaction.

[0019] The connectivity maintenance module is configured to perform collaborative search and reallocation based on global connectivity according to the task allocation list, determine mobile base stations and generate a connected subgraph through depth-first search, and output a cluster search action strategy that satisfies connectivity constraints.

[0020] The dynamic adjustment and output module is configured to control UAV waypoint tracking based on the collision time guidance law, calculate segmented velocity commands and normal acceleration commands, and output UAV swarm collaborative search results, including probabilistic maps, deterministic maps and target lists.

[0021] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the distributed area search method for UAV swarms based on global connectivity described in the first aspect.

[0022] Fourthly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the distributed area search method for UAV swarms based on global connectivity described in the first aspect.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] This invention transforms the collaborative area search problem of UAV swarms into a discrete grid allocation problem, utilizes the CBAA algorithm to achieve distributed decision-making among multiple UAVs, and leverages the suboptimal solution characteristics of CBAA in multi-agent task allocation problems to obtain suboptimal action strategies for UAVs, thereby improving the overall search efficiency of the swarm. Simultaneously, it introduces strategies such as disconnected individual removal, action selection taboo tables, and mobile base station handover, enabling CBAA to effectively eliminate action strategies of disconnected individuals. By releasing the action space of lagging individuals, the UAV swarm can always find feasible solutions while maintaining global connectivity, thus overcoming the conservative drawbacks of traditional distributed decision-making in terms of search efficiency and coverage performance.

[0025] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0026] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute a limitation thereof.

[0027] Figure 1 A main flowchart of a distributed region search method for UAV swarms based on global connectivity provided in an embodiment of the present invention;

[0028] Figure 2 A detailed flowchart of a distributed region search method for UAV swarms based on global connectivity, provided for an embodiment of the present invention;

[0029] Figure 3 This is a schematic diagram of task area discretization and UAV action space provided in an embodiment of the present invention;

[0030] Figure 4 This is a schematic diagram of the true target probability distribution in the task area in the simulation example provided by the embodiments of the present invention;

[0031] Figure 5 A schematic diagram of the UAV position distribution and deterministic map at t = 80s, provided as a simulation example of an embodiment of the present invention;

[0032] Figure 6 A schematic diagram of the probability map obtained by the UAV search at t = 80s, provided as a simulation example of an embodiment of the present invention;

[0033] Figure 7 A schematic diagram of the UAV position distribution and deterministic map at t = 200s, provided as a simulation example of an embodiment of the present invention;

[0034] Figure 8 A schematic diagram of the probability map obtained by the UAV search at t = 200s, provided as a simulation example of an embodiment of the present invention;

[0035] Figure 9 This is a schematic diagram showing the comparison of task area coverage changes of related algorithms in the simulation examples provided in this embodiment of the invention.

[0036] Figure 10 This is a schematic diagram of the topological connectivity index change curve of a drone cluster in a simulation example provided in this embodiment of the invention. Detailed Implementation

[0037] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0038] While traditional distributed search architectures can improve robustness and scalability, they are limited by the local perception and communication range of drones, making them prone to repetitive searches (local optima) due to greedy strategies, and causing conflicts between the goals of search coverage and connectivity maintenance.

[0039] This invention optimizes distributed decision-making through an auction mechanism, and solves the problems of low search efficiency and local optima caused by conservative connectivity constraints in traditional methods, while ensuring global connectivity of the communication topology.

[0040] By equating the region search to a discrete grid allocation problem, an improved consensus auction algorithm (CBAA) is introduced. Combined with disconnected individual removal, action selection taboo table and mobile base station handover strategy, the decision-making process is ensured to exclude disconnected action schemes. By releasing the action space of lagging individuals, the cluster always generates suboptimal search strategies under connectivity constraints.

[0041] By releasing action space and switching base stations, search efficiency and coverage performance are improved while ensuring connectivity, and the probability of local optima is effectively reduced, making it particularly suitable for continuous exploration in areas with dense obstacles.

[0042] Example 1

[0043] like Figure 1 As shown, this embodiment discloses a distributed region search method for UAV swarms based on global connectivity, including the following steps:

[0044] S1: Initialize the UAV cooperative search mission environment, including the UAV kinematics model, detection probability model, and cooperative search cognitive map;

[0045] S2: Based on the detection probability model, Bayes' theorem is used to obtain posterior observation information to calculate the probability estimate of the fused target and update the collaborative search cognitive map;

[0046] S3: Based on the UAV kinematics model and the updated cooperative search cognitive map, update the action space, model the equivalent task allocation problem and design the objective function to build a task allocation model for UAV swarm cooperative search;

[0047] S4: Based on the task allocation model, perform collaborative search pre-allocation according to the consensus auction algorithm, calculate the bidding value, and update the task allocation list through consensus interaction.

[0048] S5: Based on the task allocation list, global connectivity is used for collaborative search and reallocation. Mobile base stations are determined and connected subgraphs are generated through depth-first search. Cluster search action strategies that satisfy connectivity constraints are output.

[0049] S6: Controls UAV waypoint tracking based on collision time guidance law, calculates segmented velocity commands and normal acceleration commands; outputs UAV swarm collaborative search results, including probabilistic maps, deterministic maps and target lists.

[0050] Next, combined Figure 2 This embodiment provides a detailed description of a distributed region search method for unmanned aerial vehicle (UAV) swarms based on global connectivity.

[0051] In S1, the drone cooperative search mission environment is initialized, specifically including:

[0052] S11. Initialize the two-dimensional kinematic model of the UAV, considering the following two-dimensional planar kinematic equations:

[0053] (1)

[0054] in, For the first i The radial distance between the drone and the desired waypoint, and and They represent drones i and the two-dimensional position vector of the desired waypoint, Indicates drone i Speed ​​command, Its heading angle, Indicates drone i Line-of-Sight (LOS) relative to the desired waypoint. For drones i Normal acceleration command, n This indicates the number of drones. Each drone is counted according to the control cycle. Output control commands According to the decision-making cycle The system conducts area search and action decision-making. This embodiment requires that the UAVs simultaneously reach their respective desired waypoints and conduct area searches synchronously according to a prescribed time sequence. Furthermore, each UAV has a maximum detection radius. and maximum communication distance .

[0055] S12. Initialize the UAV detection model

[0056] Discretize the task space as A grid of equal size, let k Represents the raster index, defining a Boolean variable. for t Time Drone i In the grid k The detection events at the location, among which Indicates in k Target detected. Indicates no target detected; define a Boolean variable. Represents grid k The target exists in the event, where For actual existing goals, Assume there is no target; let For grid k The center position coordinates, For drones i With grid k The relative distance between them is based on the relative distance. The following UAV detection model is defined:

[0057] (2)

[0058] in, Indicates the first i The drone successfully detected the grid. k The probability of detecting a target. This indicates the sensor's maximum detection efficiency. This represents the attenuation coefficient of the drone detection probability with respect to relative distance. When multiple drones conduct cooperative detection on the same grid, the joint detection probability is defined as follows:

[0059] (3)

[0060] in, express t Always able to cover the grid k The collection of all drones.

[0061] S13. Initialize the collaborative search cognitive map

[0062] Drone swarms consider two types of cognitive maps: probabilistic maps and deterministic maps. Probabilistic maps store the distribution information of detected targets, while deterministic maps guide the swarm to conduct coverage exploration of unknown areas. Definition for t Time Grid k The probability of the existence of the target. Let its determinism coefficient be... and These respectively represent the drones in t Probability and deterministic maps at each time step, and initialization. , , .

[0063] In S2, the drones update the cluster search cognitive map, specifically including:

[0064] S21. Obtaining posterior observation information based on Bayes' theorem

[0065] make and Let represent the start time of the current decision cycle and the start time of the previous decision cycle, respectively, and satisfy . Then the drone at the decision moment Target detection is performed within the grid area to obtain real-time observations. Let... t Time Drone i For grid k The target detection results are ,but Indicates drone i For grid k Therefore, according to Bayes' theorem, the following equation holds:

[0066] (4)

[0067] in, express t Time Drone i About grids k The detection probability, express t Time Drone i About grids k Posterior probability estimation, Indicates drone i The probability of a false alarm.

[0068] S22. Obtain the probability estimate of the fused target in the region grid.

[0069] At the moment of decision The drones share the posterior probability estimates of their independent observations. And calculate its probability estimate of the fusion target in the task area according to the following formula:

[0070] (5)

[0071] in, This represents a collection of drones. Grid obtained from synchronous observation k Posterior fusion probability estimate of the central location.

[0072] S23, Update Probability Map and Deterministic Map

[0073] At the moment of decision The drone updates the deterministic map using the following formula:

[0074] (6)

[0075] in, For the learning factors of deterministic maps, As defined in equation (3), This is a discount factor for the deterministic coefficient, applied when a certain grid region has not been detected for a long time. This will significantly increase its uncertainty, thereby prompting drones to increase their exploration of the area.

[0076] At the same time, the probability map is updated using the following formula:

[0077] (7)

[0078] in, , is the learning factor for the probability map. Same as the definition in equation (5).

[0079] In S3, the task allocation modeling for multi-UAV cooperative search specifically includes:

[0080] S31, Drone Update Action Space

[0081] like Figure 3 As shown, at the decision-making moment drones i Based on current location Choose the one adjacent to it. Each grid index serves as its action space. ,in Indicates drone i The action capacity. Let Indicates the start time of the next decision-making cycle, then the drone is in the action space. Select the target grid as the current decision period. The desired waypoint, in order to make the next decision. Arrive at the waypoint. When the drone is at the area boundary, remove the boundary-crossing grid. Define the cluster joint action space as follows: , This allows each drone to store a list of action selection taboos. and initialized to .

[0082] S32. Modeling the Equivalent Task Allocation Problem for Cluster Search

[0083] Joint action space of drone swarms Consider it as a set of tasks to be assigned, define For drone index set, Let be the set of task indices, and assume that each drone can independently undertake one task (access one specific grid) within a single decision cycle. Then, the cooperative search problem of the drone swarm can be equivalently transformed into the following nonlinear integer programming problem:

[0084] (8)

[0085] in, Let the Boolean variable represent the drone. i Have you been assigned a task? j , Indicates drone i Assigned to a task j ,otherwise ; Indicates dronei Regarding the task j The revenue value; This represents the upper limit of the number of tasks a cluster can undertake within a single decision-making cycle, and is generally considered to be... .

[0086] S33. Design of the Objective Function for Cluster Search

[0087] The collaborative search of UAV swarms adopts the following task allocation objective function:

[0088] (9)

[0089] in, Indicates drone i Select task j The raster index corresponding to the time. Represents the target grid position vector, Represents the joint action space. j One element, Indicates drone i In the grid The set of grid cells that can be covered by a given location. This represents the total grid set of the task space. Then, for a task... Its task allocation benefit value have

[0090] (10)

[0091] In this embodiment, the UAV updates its action space, accurately determining available grids based on its own position. This allows the UAV's actions to better align with the actual situation, avoiding ineffective searches and improving search efficiency. The modeling of the equivalent task allocation problem for swarm search transforms the joint action space into tasks to be assigned, constructing a nonlinear integer programming problem. This ensures that task allocation fully utilizes the capabilities of each UAV and achieves rational resource allocation, while avoiding excessive dispersion or aggregation of UAVs through reasonable task arrangement, thereby maintaining the overall communication topology of the swarm. The swarm search objective function design comprehensively considers multiple factors, accurately measures task benefits, and guides UAVs to prioritize high-reward tasks, making the entire swarm search more targeted and effective. This comprehensively improves the performance of the UAV swarm in distributed area search, enabling it to better complete search tasks.

[0092] In S4, the cooperative search pre-assignment based on CBAA specifically includes:

[0093] S41, CBAA Bid Value Calculation

[0094] like Figure 1 As shown, let The number of iterations in a single CBAA is defined as follows: For drones i Task assignment list Indicates drone i The task bidding value list, both in CBBA No. Each iterative calculation connects to the results of the previous calculation, i.e. , And both initial values ​​are zero. If in the... Step by step This means that there are currently no tasks assigned to drones. i Then calculate its feasible task list according to the following formula. :

[0095] (11)

[0096] like If not all values ​​are zero, update the drone data according to the following formula. i Task allocation list and bid value list:

[0097] (12)

[0098] in, Indicates drone i In the The selected target task is iterated step by step.

[0099] S42, CBAA Bidding Consistency Calculation

[0100] set up This represents the adjacency matrix of the cluster communication topology at the current moment, i.e., when the drone... i With drones k The relative distance between them is less than the maximum communication radius. Sometimes, In this case, the two are adjacent to each other; otherwise, In particular, , At the moment of decision-making drones i To neighboring drones k Send the list of bid values and receive from k List of bid values Then in the first... Iterate step by step, updating the task allocation list and bid value list according to the following formula:

[0101] (13)

[0102] in, Indicates dronei After the first The list of winners after iterative consensus interaction. This indicates a drone. i In the The information about the task obtained through step iterations j The winning bidder's number. For any drone. When the first Task allocation list for each iteration If none of the values ​​are zero, terminate the iteration and output the task assignment list; otherwise, proceed to step S43 and update the iteration step. .

[0103] S43. Resetting the action space of individuals with decision-making lag

[0104] definition The threshold for the number of iterations assigned to a single UAV CBAA mission, if And drones still exist i Its task assignment list If all values ​​are zero, then in the current action space No valid solution can be found within the individual. i Becoming a decision-making lag. At this point, the drone needs to be reset. i Action Selection Taboos If the condition is met, proceed to step S41; otherwise, proceed directly to step S41.

[0105] In S5, the cooperative search reallocation based on global connectivity specifically includes:

[0106] S51. Determine the cluster topology of the mobile base station.

[0107] In the decision-making cycle Within the drone swarm, one drone will be designated as a mobile base station to participate in the connectivity construction of the communication topology and will act as the starting node to establish communication links with other drones. The drone swarm makes decisions at the moment... The mobile base stations for this period are determined by the following formula. :

[0108] (14)

[0109] in, This represents the iteration step size when the CBAA pre-allocation algorithm terminates for the first time. For drones i The index of pre-allocated tasks. For example... Figure 1 As shown, the number of iterations in the decision algorithm is defined as follows: The iteration threshold of the algorithm is If the decision model is in If convergence still fails, then the currently selected mobile base station will be... Remove from the comparison queue of Equation (14) and reselect the mobile base station according to Equation (14), then proceed to step S52; otherwise, proceed directly to step S52.

[0110] S52. Determining Cluster Connectivity Subgraphs Based on Depth-First Search

[0111] make The decision algorithm is represented in the first step. The CBAA pre-allocation algorithm terminates at step size 1, then the 1st step... The pre-allocated cluster task list of Step CBAA is as follows The decision algorithm is then obtained from the following formula. Pre-assigned target grid positions for each drone:

[0112] (15)

[0113] in, Indicates drone i The pre-assigned raster index. Determined by the pre-assigned target location. The current iteration step is obtained by calculating the relative distance between each pair of elements. The pre-allocated adjacency matrix and with mobile base stations Starting with the node, the Depth-First Search (DFS) algorithm is used to... Perform a traversal to obtain information about Connected subgraph For any drone Retain its task assignment list and a list of bid values Otherwise, reset. , and make upcoming mission Add drones i Action selection taboos table.

[0114] S53, Output Cluster Search Action Strategy

[0115] Judge the first Iterative Cluster Task List With the Step Cluster Task List Check if the pairs are equal and contain no zero elements. If so, the decision algorithm has converged, and the output should be... As a search action strategy for drone swarms; otherwise, let Proceed to step S41.

[0116] In this embodiment, determining the cluster topology mobile base station enables the selection of suitable UAVs as starting nodes to build communication links within the decision-making cycle, ensuring the connectivity of the cluster communication topology and laying the foundation for effective information transmission. Determining the cluster connectivity subgraph based on depth-first search accurately identifies the set of UAVs connected to the mobile base station, allowing for the reasonable retention or reset of their task allocations, avoiding invalid task allocations, optimizing resource allocation, and making the overall cluster action more coordinated and unified. In the cluster search action strategy output stage, the convergence of iteration results determines whether to continue optimization, ensuring that the final output action strategy enables the UAV cluster to achieve efficient collaborative search globally, avoiding local optima, improving the entire cluster's ability to achieve global goals in distributed area search, and effectively promoting the achievement of global objectives.

[0117] In S6, the UAV waypoint tracking control based on collision time cooperative guidance specifically includes:

[0118] S61, Command to calculate UAV waypoint tracking segment speed

[0119] At the moment of decision ,make The number of iterations at which the decision algorithm converges indicates the current decision cycle. The cluster search action strategy is Therefore, the desired grid position for each drone is calculated using the following formula:

[0120] (16)

[0121] in, Indicates drone i The target raster index. Then the drone. i In the current decision-making cycle The waypoint tracking speed command within the system has

[0122] (17)

[0123] in, This is an adjustment factor used to control the path length of the UAV to the desired waypoint.

[0124] S62. Calculation of normal acceleration control command based on collision time guidance law

[0125] In the decision-making cycle Inside, drones i According to the control cycle Calculate the arrival time at the desired waypoint using the following formula Remaining time estimate :

[0126] (18)

[0127] Define drones i The collision time error is The guide angle is Then its normal acceleration control command can be calculated using the collision time guidance law shown in the following formula.

[0128] (19)

[0129] in, This is the gain coefficient. Each UAV operates according to the control cycle. Calculate equations (18) and (19), output normal acceleration control commands, and in the decision cycle Within, output the speed command according to formula (17) so that it reaches the predetermined time. Arrive at the desired waypoint accurately This enables each machine to simultaneously conduct collaborative detection, information exchange, and search decision-making, while ensuring the connectivity of the cluster communication topology at the moment of decision-making.

[0130] Furthermore, the output of drone swarm collaborative search results specifically includes:

[0131] Determine the current time t Has the scheduled task deadline been reached? If so, output the probability map. and deterministic maps The system retrieves the target list and other status data; otherwise, it returns to step two and updates the search time according to the decision cycle. and control cycle Conduct collaborative detection, search decision-making, and target waypoint tracking and control using unmanned aerial vehicles (UAVs).

[0132] Example 1

[0133] The effectiveness of the method proposed in this embodiment will be verified through a specific example below. In this example, consider 10 UAVs conducting a cooperative search of an unknown area of ​​2400m × 2400m. The true target probability distribution in the task area is as follows: Figure 4 As shown. The mission area is discretized into a 50×50 grid space. Each UAV is equipped with detection and communication equipment with a sensing range of 150m and a communication range of 150m. Its initial position is located in the lower left corner of the mission area, ensuring connectivity at the starting position. The initial airspeed is 40m / s, and the heading angle is 0. A simulation mission time limit is set. Control cycle Decision-making cycle At the start of the exercise, the UAVs have no prior target distribution information for the mission area. Their mission objective is to efficiently explore the mission area through waypoint planning and collaborative exploration to quickly reduce regional uncertainty and obtain a high-confidence target distribution and probability map. Simultaneously, the communication topology connectivity of the UAV swarm is maintained at the decision-making stage. To verify the effectiveness of the swarm decision-making method in this embodiment, a comparative simulation with a potential game theory model was added. A potential function and a precise potential game utility function were constructed using a deterministic map, and connectivity constraints were added for UAV action selection to achieve search path planning for the UAV swarm while maintaining connectivity. Specifically, this embodiment allows the swarm communication topology to change during the search process due to the switchability of the mobile base station, while the potential game decision-making model uses a fixed topology, meaning each UAV must maintain a predetermined communication topology during the search. The simulation environment for this example is configured with an Intel i7-8565U processor, 8GB of memory, and MATLAB R2018a software.

[0134] The specific steps of this embodiment are as follows:

[0135] Step 1: Initialize the UAV cooperative search mission environment, which includes:

[0136] S11. Initialize the two-dimensional kinematic model of the UAV.

[0137] Given A number of unmanned aerial vehicles (UAVs) are used, employing the kinematic model of equation (1). Each UAV starts at the lower left corner of the mission area with an initial velocity of 40 m / s and a heading and normal acceleration of 0. The control period for each UAV is... Decision-making cycle It has the largest detection radius and maximum communication distance .

[0138] S12. Initialize the UAV detection model

[0139] The task area is discretized into 50×50 equally sized grid cells. The UAV then uses the perception model shown in equation (2) to detect targets in the area based on the relative distance to neighboring grid cells, where the sensor's maximum detection efficiency is... attenuation coefficient When multiple drones conduct collaborative detection on the same grid, the joint detection probability shown in equation (3) is adopted.

[0140] S13. Initialize the collaborative search cognitive map

[0141] Both the probabilistic and deterministic maps of the drone swarm use the same 50×50 grid space as the discretization format of the mission area, and are initialized... , , .

[0142] Step 2: The drone updates the cluster search cognitive map, specifically including:

[0143] S21. Obtaining posterior observation information based on Bayes' theorem

[0144] At each decision-making moment, the UAV performs target detection on the grid area it is in to obtain real-time observation values. Based on equation (4), it calculates the posterior probability of the presence of the target in the covered grid, where the false alarm probability is... .

[0145] S22. Obtain the probability estimate of the fused target in the region grid.

[0146] At each decision moment, the UAV shares its independently observed posterior target probability distribution with other UAVs and uses Equation (5) to calculate the fused target probability estimate of the UAV cluster for the current grid space.

[0147] S23, Update Probability Map and Deterministic Map

[0148] For any grid cell in the grid space, the deterministic map is updated using Equation (6) by combining the joint detection probability of Equation (3), where the learning factor... Discount factor Simultaneously, the probability map is updated using equation (7), where the learning factor... .

[0149] Step 3: Task allocation modeling for multi-UAV cooperative search, specifically including:

[0150] S31, Drone Update Action Space

[0151] like Figure 2 As shown, at each decision-making moment, the drone selects its neighboring drones based on its current location. Each grid index serves as its action space. When the drone is located at the area boundary, its action space removes adjacent grids that cross the boundary, thus increasing the action capacity. The UAV selects a target grid within its action space as the desired waypoint for the current decision cycle and aims to reach that waypoint at the next decision time. Simultaneously, each UAV stores a separate action selection taboo table, initialized as the union of the action spaces of other UAVs, to avoid selection conflicts with target grids.

[0152] S32. Modeling the Equivalent Task Allocation Problem for Cluster Search

[0153] Treating the joint action space of the UAV swarm as a set of tasks to be assigned, and assuming that each UAV can independently undertake one task within a single decision cycle, the cooperative search problem of the UAV swarm is equivalently transformed into the nonlinear integer programming problem shown in equation (8), where the upper limit of the number of tasks undertaken by the UAV swarm within a single decision cycle is... , This refers to the space capacity for cluster joint actions.

[0154] S33. Design of the Objective Function for Cluster Search

[0155] The collaborative search of the UAV swarm adopts the task allocation objective function shown in Equation (9), and the task allocation benefit value is calculated according to Equation (10).

[0156] Step 4: CBAA-based collaborative search pre-assignment, specifically including:

[0157] S41, CBAA Bid Value Calculation

[0158] For each drone, initialize its CBAA iteration steps. and make the starting step , , Judgment of the first Does the step satisfy? If so, calculate the list of feasible tasks for the UAV according to formula (11). Otherwise, proceed to step S42. Determine... If not all values ​​are zero, then update the drone task allocation list and bidding value list according to formula (12).

[0159] S42, CBAA Bidding Consistency Calculation

[0160] Calculate the cluster communication topology adjacency matrix at the current moment. At each decision moment, the drone sends a list of bidding values ​​to its neighboring drones. And receive a list of bid values ​​from neighboring drones. Based on this, update the task allocation list and bidding value list for each UAV according to equation (13). For any UAV, when the first... Task allocation list for each iteration If none of the values ​​are zero, terminate the iteration and output the task assignment list; otherwise, proceed to step S43 and update the list. .

[0161] S43. Resetting the action space of individuals with decision-making lag

[0162] Set a threshold for the number of iterations of the UAV CBAA. ,like If a drone's task assignment list is still entirely zero, then reinitialize the drone's action selection taboo list and proceed to step S41; otherwise, proceed directly to step S41.

[0163] Step 5: Collaborative search and reallocation based on global connectivity, specifically including:

[0164] S51. Determine the cluster topology of the mobile base station.

[0165] At each decision-making moment, the drone swarm determines the mobile base station for the current decision-making cycle according to equation (14). Initialize the number of iterations for the decision algorithm. And set the iteration threshold of the algorithm. Determine if there is a problem in the current iteration step. If so, then the current mobile base station Remove from the comparison queue of Equation (14) and reselect the mobile base station according to Equation (14), then proceed to step S52.

[0166] S52. Determining Cluster Connectivity Subgraphs Based on Depth-First Search

[0167] Based on the CBAA pre-assignment results, the pre-assigned target grid position for each UAV is calculated according to Equation (15), and the current iteration step number is obtained from the pre-assigned target position. The pre-allocated adjacency matrix With mobile base stations Starting with the node, use the DFS algorithm to... Perform a traversal to obtain information about Connected subgraph For any connected drone, retain its task assignment list. and a list of bid values Otherwise, reset its assigned tasks and bid values, and add the reset task to its action selection taboo list.

[0168] S53, Output Cluster Search Action Strategy

[0169] Judge the first Step and the first Check if the cluster task lists of each iteration are equal and contain no zero elements. If so, the decision algorithm has converged, and the output of the first iteration is given. The iterative list of cluster tasks is used as the drone swarm search action strategy; otherwise, let... Proceed to step S41.

[0170] Step Six: UAV waypoint tracking and control based on collision-time cooperative guidance, specifically including:

[0171] S61, Command to calculate UAV waypoint tracking segment speed

[0172] At each decision point, each UAV calculates its desired grid position according to equation (16) and calculates the waypoint tracking speed command for the current decision cycle in conjunction with equation (17), where the adjustment coefficient... .

[0173] S62. Calculation of normal acceleration control command based on collision time guidance law

[0174] Within the current decision-making cycle, the drone operates according to the control cycle. Calculate the estimated remaining time to reach the desired waypoint using equation (18), and calculate the normal acceleration control command using equation (19), where the gain coefficient is... The drone operates according to a control cycle. It outputs normal acceleration control commands and speed commands to enable the aircraft to accurately reach the desired waypoint at the next decision moment, thereby enabling each aircraft to carry out collaborative detection, information exchange and search decision-making simultaneously, and ensuring the connectivity of the cluster communication topology at the decision moment.

[0175] Furthermore, the output of drone swarm collaborative search results specifically includes:

[0176] Determine the current time t Has the scheduled task deadline been reached? If yes, output the probability map and deterministic map, as well as the list of targets found, and other state data; otherwise, return to step two and update the search time according to the decision cycle. and control cycle Conduct collaborative detection, search decision-making, and target waypoint tracking and control using unmanned aerial vehicles (UAVs).

[0177] Figures 5 to 10 The verification results for this example are presented.

[0178] in Figure 5 and Figure 7 The node numbers in the diagram represent the drone numbers, the arrows indicate the direction of drone movement, and the thick solid lines indicate that two drones maintain topological connectivity. Figure 5 and Figure 7 As can be seen, the global connectivity proposed in this embodiment can rationally select mobile base stations based on the coverage of the task area by local UAVs in each decision cycle, maintaining the connectivity of the cluster topology. From the spatial distribution of UAVs, it can be seen that during the search process, except for adjacent boundaries, the UAV cluster mostly exhibits a linear formation pattern with overlapping tails. This phenomenon caters to the task objective of maximizing the coverage of unknown areas by the UAV cluster, quickly reducing regional uncertainty, and exhibiting optimal or near-optimal search behavior.

[0179] Figure 6 and Figure 8This reflects the iteration of the probability map during the search process of the UAV swarm. The simulation results show that the probability map can gradually approach the actual target distribution in the task area as the search progresses, providing reliable situational awareness information for the UAV swarm.

[0180] Figure 9 and Figure 10 The results of the comparison between the method of this embodiment and the potential game theory model are presented.

[0181] Depend on Figure 9 As can be seen, the method in this embodiment effectively improves the collaborative coverage efficiency of the task area compared with the game theory decision-making method, and can continuously improve the area coverage rate within the task time limit, thus overcoming the local optimality problem of the game theory decision-making method.

[0182] Figure 10 This indicates that the connectivity coefficient of the method in this embodiment is lower than that of the game theory decision-making method. However, it can still reduce the binding effect of connectivity constraints on the selection of UAV actions by releasing the UAV action space and switching mobile base stations, thereby outputting action strategies that are more in line with the mission objectives and have better exploratory capabilities under the CBAA decision-making framework.

[0183] Example 2

[0184] This embodiment provides a globally connected UAV swarm distributed area search system, including:

[0185] The initialization module is configured to initialize the UAV cooperative search mission environment, including the UAV kinematic model, the detection probability model, and the cooperative search cognitive map.

[0186] The map update module is configured to use Bayes' theorem to obtain posterior observation information based on the detection probability model, calculate the probability estimate of the fused target, and update the collaborative search cognitive map.

[0187] The task construction module is configured to update the action space, model the equivalent task allocation problem, and design the objective function based on the UAV kinematic model and the updated cooperative search cognitive map, and construct the task allocation model for UAV swarm cooperative search.

[0188] The information allocation module is configured to perform collaborative search pre-allocation based on the task allocation model and the consensus auction algorithm, calculate the bidding value, and update the task allocation list through consensus interaction.

[0189] The connectivity maintenance module is configured to perform collaborative search and reallocation based on global connectivity according to the task allocation list, determine mobile base stations and generate a connected subgraph through depth-first search, and output a cluster search action strategy that satisfies connectivity constraints.

[0190] The dynamic adjustment and output module is configured to control UAV waypoint tracking based on the collision time guidance law, calculate segmented velocity commands and normal acceleration commands, and output UAV swarm collaborative search results, including probabilistic maps, deterministic maps and target lists.

[0191] Example 3

[0192] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of a globally connected UAV swarm distributed area search method as described in Embodiment 1 above.

[0193] Example 4

[0194] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the globally connected distributed area search method for UAV swarms described in Embodiment 1 above.

[0195] The steps or modules involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0196] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A distributed region search method for UAV swarms based on global connectivity, characterized in that, include: Initialize the UAV cooperative search mission environment, including the UAV kinematic model, detection probability model, and cooperative search cognitive map; Based on the detection probability model, Bayes' theorem is used to obtain posterior observation information to calculate the probability estimate of the fused target and update the collaborative search cognitive map; Based on the UAV kinematics model and the updated cooperative search cognitive map, we update the action space, model the equivalent task allocation problem and design the objective function to construct a task allocation model for UAV swarm cooperative search. Based on the task allocation model, collaborative search pre-allocation is performed according to the consensus auction algorithm, the bidding value is calculated, and the task allocation list is updated through consensus interaction. Based on the task allocation list, global connectivity is used for collaborative search and reallocation. Mobile base stations are identified and connected subgraphs are generated through depth-first search. Cluster search action strategies that satisfy connectivity constraints are output. Based on the collision time guidance law, control the waypoint tracking of the UAV and calculate the segmented velocity command and normal acceleration command; Outputs search results for drone swarm collaboration, including probabilistic maps, deterministic maps, and target lists.

2. The distributed region search method for UAV swarms based on global connectivity as described in claim 1, characterized in that, The UAV kinematic model is a two-dimensional planar kinematic equation, and the initialization parameters include control cycle, decision cycle, maximum detection radius, and communication distance.

3. The distributed region search method for UAV swarms based on global connectivity as described in claim 1, characterized in that, The objective function is designed based on grid determinism, and the benefit value is calculated based on the determinism coefficient of the UAV-covered grid set.

4. The distributed region search method for UAV swarms based on global connectivity as described in claim 1, characterized in that, The bid value calculation includes filtering the feasible task list and updating the bid value, and the consistency interaction is achieved through the communication topology adjacency matrix.

5. The distributed region search method for UAV swarms based on global connectivity as described in claim 1, characterized in that, The selection of the mobile base station is determined based on the pre-allocation result of the consensus auction algorithm and the iteration step size. The connected subgraph is generated by traversing the pre-allocated adjacency matrix using DFS.

6. The distributed region search method for UAV swarms based on global connectivity as described in claim 1, characterized in that, The collision time guidance law is used to calculate the normal acceleration command, and the remaining time estimate is obtained based on the radial distance and velocity command between the UAV and the desired waypoint.

7. The distributed region search method for UAV swarms based on global connectivity as described in claim 1, characterized in that, The updates to the probabilistic and deterministic maps are based on the fusion of target probability estimates and joint detection probabilities, with the results output triggered by the mission time limit.

8. A distributed regional search system for unmanned aerial vehicle (UAV) swarms based on global connectivity, characterized in that, include: The initialization module is configured to initialize the UAV cooperative search mission environment, including the UAV kinematic model, the detection probability model, and the cooperative search cognitive map. The map update module is configured to use Bayes' theorem to obtain posterior observation information based on the detection probability model, calculate the probability estimate of the fused target, and update the collaborative search cognitive map. The task construction module is configured to update the action space, model the equivalent task allocation problem, and design the objective function based on the UAV kinematic model and the updated cooperative search cognitive map, and construct the task allocation model for UAV swarm cooperative search. The information allocation module is configured to perform collaborative search pre-allocation based on the task allocation model and the consensus auction algorithm, calculate the bidding value, and update the task allocation list through consensus interaction. The connectivity maintenance module is configured to perform collaborative search and reallocation based on global connectivity according to the task allocation list, determine mobile base stations and generate a connected subgraph through depth-first search, and output a cluster search action strategy that satisfies connectivity constraints. The dynamic adjustment and output module is configured to control UAV waypoint tracking based on the collision time guidance law, calculate segmented velocity commands and normal acceleration commands, and output UAV swarm collaborative search results, including probabilistic maps, deterministic maps and target lists.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the distributed region search method for UAV swarms based on global connectivity as described in any one of claims 1-7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the distributed region search method for UAV swarms based on global connectivity as described in any one of claims 1-7.

Images