A swarm robot roundup simulation verification platform and method

Through the cluster robot roundup simulation verification platform, the virtual simulation environment and control algorithm layer are used, combined with the message queue telemetry transmission protocol and fast random tree path planning, the verification problem of cluster robot roundup strategy is solved, and rapid theoretical verification and algorithm performance evaluation are achieved. Reduces actual deployment risks and costs.

CN118690584BActive Publication Date: 2024-12-03ZHEJIANG LAB
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411163690.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-12-03
Estimated Expiration
2044-08-23

AI Technical Summary

Technical Problem

In the cluster robot roundup task, communication and collaboration between robots, autonomous intelligent decision-making of a single robot, collision avoidance planning of multiple robots, object detection and tracking of multiple robots are challenges in building feasible multi-robot roundup strategies, and there is a lack of an effective simulation verification platform. To verify the feasibility of a steady cluster strategy.

Method used

Provide a cluster robot round-up simulation verification platform, including the virtual simulation environment layer, the control algorithm layer and the communication layer. The virtual simulation environment layer includes the user login interface, role selection, round-up scene generation and role control modules. The control algorithm layer is responsible for the robot's motion trajectory. Control, the communication layer realizes data communication through message queue telemetry transmission protocol, and uses algorithms such as fast random tree and B-spline curve for path planning.

Benefits of technology

It realizes verification and testing cluster robot roundup behavior without actual physical risks, reduces the risk of actual deployment, saves hardware costs, and can quickly iterate and evaluate algorithm performance, providing a fast theoretical verification tool.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118690584B_ABST
    Figure CN118690584B_ABST
Patent Text Reader

Abstract

The present invention discloses a cluster robot roundup simulation verification platform and method, which includes a virtual simulation environment layer, a control algorithm layer, and a communication layer; the virtual simulation environment layer is used to build a virtual roundup simulation platform environment to realize the visualization of multi-machine collaborative roundup in different scenarios; the control algorithm layer is used to calculate the motion control calculation of the cluster roundup scene robot; the communication layer provides an information transmission channel between the virtual simulation environment and the control algorithm layer; the cluster robot roundup simulation verification platform includes visualization development of different map sizes and random obstacle environments, and also includes two robot role settings of roundup and escaper; and realizes the pursuit of escaped robots by multiple roundup robots. The cluster robot roundup simulation verification platform disclosed by the present invention has the advantages of high scene generation freedom, convenient operation, stable communication, and easy understanding and verification, and can be extended to the rapid development and verification of other roundup algorithms.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent robots, and in particular to a cluster robot capture simulation verification platform and method. Background Art

[0002] A single system or a single robot will face performance limitations in combat, including load capacity, sensing range, computing power, power consumption, etc. Cluster technology can overcome these limitations of individual machines and improve overall performance and effectiveness. For example, in an unmanned autonomous combat system, clustering can enhance the safety, reliability, flexibility and efficiency of tracking, encircling and striking tasks.

[0003] In the swarm capture task, communication and coordination between robots, autonomous intelligent decision-making of a single robot, collision avoidance planning of multiple robots, and target detection and tracking are the challenges in building a feasible multi-robot capture strategy. In order to verify the feasibility of the swarm strategy, a simulation platform for swarm robots to capture targets needs to be built. Summary of the invention

[0004] The purpose of the present invention is to provide a swarm robot capture simulation verification platform and method in view of the deficiencies in the prior art.

[0005] To achieve the above object, the present invention provides a swarm robot capture simulation verification platform, comprising:

[0006] The virtual simulation environment layer is used to build a virtual capture simulation platform environment to realize the visualization of multi-robot collaborative capture in different scenarios;

[0007] The control algorithm layer is responsible for the motion control calculation of the robots in the cluster capture scene and realizes the control of the motion trajectory of all robot characters;

[0008] The communication layer is used to provide an information transmission channel between the virtual simulation environment and the control algorithm layer, and realize data communication between the control algorithm layer and the virtual simulation environment layer.

[0009] Furthermore, the virtual simulation environment layer includes a user login interface module, a role selection module, a capture scene generation module, a capture scene matching and synchronization module, and a role control and visualization module; wherein:

[0010] The user login interface is used to collect the user name and password information input by the user, and send this information to the server through a script for verification;

[0011] The role selection module includes the role selection of the capturer or escaper; the escaper controls the movement trajectory of the escaper manually through an external device, and the capturer controls the capture strategy and movement state through algorithm design;

[0012] The capture scene generation module includes map creation and map addition selection, wherein map creation is used to generate a new random map, and map addition is used to add a pre-set map; in the map creation selection interface, the capture scene is controlled by custom options; the custom options include weather selection, terrain setting, map size, obstacle setting, map naming and save path selection;

[0013] The capture scene matching and synchronization module is used to set the initial position of the character, including manually inputting the coordinates of the initial position of the character or randomly generating the initial position of the character in the map; detecting the type and number of the character through the server, and when all the characters are ready, synchronously entering the map and realizing real-time visualization on the map;

[0014] The character control and visualization module are used to select manual control or automatic control of the robot to achieve changes in its motion state; each robot character includes a first-person perspective, a third-person perspective, and a free perspective; and global map information is simultaneously visualized above the map, including map boundaries, obstacle location information, and location information of all robots in the map.

[0015] Furthermore, in the capture scene generation module, the weather selection includes sunny, cloudy, foggy and rainy days; the terrain setting includes the highest terrain height and the lowest terrain height setting; the obstacle setting includes setting the number and position of obstacles, and the obstacles include trees and stones;

[0016] Set the control method of the role in the roundup scene generation module, including manual control and automatic control; manual control means that the role's movement direction is controlled by an external keyboard, and automatic control means that the role's movement direction is controlled by the control algorithm layer.

[0017] Furthermore, in the character control and visualization module, selecting manual control or automatic control of a robot includes: selecting a manually controlled robot to control its movement direction and speed through the keyboard, and controlling the rotation of its viewing angle through the mouse; selecting an automatically controlled robot to change its movement state through control data transmitted from the communication layer.

[0018] Furthermore, the control algorithm layer controls the motion trajectories of all robot characters, and plans the next motion posture through information on the robot's own position, other robots' positions, obstacle positions, target positions, and map boundaries.

[0019] Furthermore, the communication layer is implemented based on the message queue telemetry transmission protocol, communicates with the server by subscribing to the map and robot position topics, and decodes the data format of the subscribed topics through the callback function to realize data communication between the control algorithm layer and the virtual simulation environment layer.

[0020] To achieve the above object, the present invention also provides a method for collaborative planning of a capture path based on the above cluster robot capture simulation verification platform, comprising the following steps:

[0021] S1: Initialize the map and tree: Build a map based on the communication information of the simulation verification platform, including the boundary and obstacle information of the map; build multiple tree nodes of the capture robot; among them, create the root node of the tree as the starting position of all robots, and then initialize an empty tree Ti for each capture robot i, and use the tree node as the starting position of the capture robot i;

[0022] S2: loop to build tree;

[0023] S2.1: Randomly select a point on the map as X_rand, and find the node X_nearest closest to X_rand in the tree Ti;

[0024] S2.2: Based on the search step limit L_limit, determine the position of the new tree node X_new between X_nearest and X_rand;

[0025] S2.3: Boundary detection and collision detection: Set the obstacle list to the static environment obstacle and dynamic obstacle set, check whether the line from X_nearest to X_new has exceeded the boundary limit or collided with an obstacle, if not, add X_new to the tree T and make X_nearesr the parent node of X_new;

[0026] S3: Target judgment: Check whether the connection between X_new and the target satisfies the boundary detection and collision detection. If the connection is successful, the path finding is successful.

[0027] S4: Path extraction and smoothing:

[0028] S4.1: Starting from the target pose, traverse the tree backwards and store the node sequence to the root node as the planned path;

[0029] S4.2: Smoothing the stored node sequence, using a spline curve to smooth the path;

[0030] S4.3: Returns the interpolated smoothed path point sequence.

[0031] Furthermore, in the sub-step S2.3, the collision detection includes: detecting the collision between the robot and the obstacle, and performing synchronous collision detection on other capturing robots as dynamic obstacles.

[0032] Furthermore, in step S3, the target position and posture of each capture robot is determined by the position and posture of the target robot and the number of robots.

[0033] Furthermore, in the sub-step S4.2, a B-spline curve is used to achieve path smoothing, thereby ensuring the kinematic constraints of the robot while obtaining continuous posture changes.

[0034] The beneficial effects of the present invention are:

[0035] 1. Rapid theoretical verification: Before applying theoretical algorithms to practice, the simulation verification platform provided by the present invention can be used as a tool to verify and test new theories, algorithms, and control strategies. By simulating the hunting behavior of swarm robots in a simulation environment, possible problems can be foreseen without actual physical risks, thus reducing the risks of actual deployment.

[0036] 2. Cost saving and fast algorithm iteration: The simulation verification platform does not need to actually deploy the robot system, effectively saving expensive hardware and maintenance costs. It can also test multiple configurations in a short time and quickly modify and adjust the strategy before getting the ideal results.

[0037] 3. Algorithm performance evaluation: By reproducing the cluster capture status in different cases and scenarios, the simulation verification platform provided by the present invention can quickly review and verify the task completion time, energy consumption, and algorithm accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is the overall architecture diagram of the simulation verification platform in the present invention;

[0039] Figure 2 It is an operation flow chart of the simulation verification platform in the present invention;

[0040] Figure 3 It is a flow chart of the collaborative planning method of the roundup path in the present invention. DETAILED DESCRIPTION

[0041] The present invention is described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the features of the following embodiments and implementations can be combined with each other.

[0042] See also Figure 1 and Figure 2 The present invention provides a cluster robot roundup simulation verification platform, including a virtual simulation environment layer, a control algorithm layer, and a communication layer.

[0043] The virtual simulation environment layer is used to build a virtual capture simulation platform environment to realize the visualization of multi-robot collaborative capture in different scenarios;

[0044] The control algorithm layer is responsible for the motion control calculation of the robot in the cluster capture scenario;

[0045] The communication layer provides an information transmission channel between the virtual simulation environment and the control algorithm layer.

[0046] In the virtual simulation environment layer, the virtual roundup simulation environment is developed based on Unity, including a user login interface module, a role selection module, a roundup scene generation module, a roundup scene matching and synchronization module, and a role control and visualization module; among which:

[0047] The user login interface module collects the username and password information entered by the user; when the user clicks the login button, this information will be sent to the server through a script for verification; the server will check whether this information matches the correct username and password stored in advance.

[0048] This simulation verification platform supports seven clients logging in at the same time. Taking unmanned vehicles as an example, six unmanned vehicles can capture the target vehicle.

[0049] After the user logs in successfully, he / she enters the role selection module; it includes two role settings, namely the capturer and the escaper; the escaper's movement trajectory is controlled by external devices, while the capturer controls his / her capture strategy and movement status through algorithm design; the two are distinguished by color.

[0050] In the capture scene generation module, it includes map creation and map addition selection, where map creation is used to generate a new random map, and map addition is used to add a map that has been set and saved; in the map creation selection interface, the capture scene is controlled by custom options; the custom options mainly include: weather selection (four types: sunny, cloudy, foggy, and rainy, different weather mainly affects visibility), terrain settings (including the highest terrain height and the lowest terrain height settings), map size, obstacle settings (the field scene contains trees and stones as obstacles, and the number of trees and stones needs to be manually set, and their position settings can include random definition or manual definition; random definition means that trees and stones will be randomly generated on the map, and manual definition means that trees and stones will be generated on the map according to the specific coordinates set), map naming and save path selection. In addition, the capture scene generation can also set the control method of the character, including manual control and automatic control. Among them, the manual control method means that the movement of the character is controlled by the external keyboard, and the automatic control means that the movement of the character is controlled by the control algorithm layer.

[0051] In the roundup scene matching and synchronization module, the initial position of the character can be set. It supports manual input of the character's initial position coordinates or random generation on the map. It supports setting whether video recording is supported at the moment. The server detects the role category and the number of participants. All characters can enter the map synchronously by clicking ready, and realize synchronous visualization in the same map.

[0052] In the character control and visualization module, manually controlled unmanned vehicles can control their direction of movement through the keyboard W / A / S / D, their speed through the up and down arrow keys, and their view rotation through mouse sliding; automatically controlled unmanned vehicles can change their motion state through control data transmitted from the communication layer. Each character can set the view attributes, which include first-person, third-person and free view. At the same time, the global map information is visualized above the map, including map boundaries, location information of obstacles such as trees and stones, and location information of all unmanned vehicles in the map.

[0053] In the capture simulation verification platform provided by the present invention, the control of the motion trajectories of all unmanned vehicle characters is achieved through the control algorithm layer. In each loop iteration, the next movement posture of the unmanned vehicle is planned through information such as the unmanned vehicle's own position, the positions of other unmanned vehicles, the obstacle positions of objects in the environment, the target position, and the map boundary. On the one hand, obstacle avoidance and collision avoidance of unmanned vehicles in a highly dense environment are achieved, and on the other hand, coordinated capture of a target by multiple capture unmanned vehicles is achieved.

[0054] Specifically, in the capture simulation verification platform, the map information is expressed as the boundary coordinates (0,0) and (xmax,ymax) in the global coordinate system of the map. The obstacles include trees and stones, which are expressed as circles with the center (x0,y0) in the control algorithm layer, and the radii of the trees and stones are r_tree and r_store, respectively. Secondly, the robot's position and posture are expressed by the center of mass coordinates (xi, yi) and the orientation angle (thetai) in the global coordinate system of the map, where i represents the ID number of the i-th robot. The above information is used as the input data of the decision-making of the multi-machine collaborative capture method. The output data of the multi-machine collaborative capture method is the speed direction and size of each capture unmanned vehicle at the next moment. The type of control algorithm can adopt a traditional path planning algorithm or an algorithm based on deep reinforcement learning, etc. As long as the input and output ends are consistent, the capture simulation verification platform of the present invention can be used to verify the capture decision-making method.

[0055] In the roundup simulation verification platform provided by the present invention, the communication layer is implemented based on the MQTT (Message Queuing Telemetry Transport) protocol to realize the timely communication of data between the control algorithm layer and the virtual simulation environment layer, and the communication data is in the jason format (JavaScript Object Notation, lightweight data exchange format). The client is created in the control algorithm layer and the virtual simulation environment layer respectively, and the control algorithm layer and the virtual simulation environment layer code communicate with the server by subscribing to the map and robot position topics, and the jason format data of the subscription topic is decoded through the callback function, and the control algorithm layer and the virtual simulation environment layer communicate data at a frequency of 20~200 HZ.

[0056] Within the single-step communication frame rate, the data transmitted by the virtual roundup simulation environment to the control algorithm layer include: map boundaries, the location and size of obstacles, the numbers, postures, and speed information of all unmanned vehicles synchronized in the map; after receiving these data, the control algorithm layer calculates through its core algorithm module, and transmits to the virtual roundup simulation environment the numbers of all unmanned vehicles synchronized in the map, the target posture of the unmanned vehicle in the next time step, and the movement speed information of the unmanned vehicle. The movement speed is obtained through frame rate solution (that is, calculated based on the distance between the time step and the target posture).

[0057] In order to achieve multi-machine collaborative capture, compared with single-machine path planning, multi-machine collaborative capture path planning requires two additional algorithm improvements: first, in order to achieve capture from multiple directions, the posture angles of the end points of the path planning of multiple capture robots are multiple angle directions that surround the posture of the escape robot, and the specific angle values ​​are determined by the posture of the escape robot and the number of robots; second, when planning the path of a single machine, the remaining collaborative capture robots are added as dynamic obstacles to the input module of the path planning algorithm to avoid collision avoidance between multiple machines.

[0058] The simulation verification platform of the present invention is a cluster robot capture strategy verification simulation platform with good simulation effect and reliable and stable communication. It provides interfaces for multiple scene types, state selections, and multiple robot control mode selections, and provides a visual verification interface for capture algorithm verification.

[0059] Considering the real-time problem, the present invention also provides a method for collaborative planning of roundup paths executed on the above-mentioned roundup simulation verification platform, that is, a collaborative path planning algorithm for realizing multi-machine cluster roundups. The method adopts a sampling-based RRT (Rapid Random Tree) path planning algorithm for single-machine path planning; secondly, a smoothness function is added to the output end of the RRT algorithm to optimize the smoothness index of the output path. Figure 3 , the method specifically comprises the following steps:

[0060] S1: Initialize the map and tree: Build a map according to the communication information of the simulation verification platform, including the boundary and obstacle information of the map; Build tree nodes for multiple capture robots: Create a root node of the tree, whose position is the common starting position of all robots; then, initialize an empty tree Ti for each capture robot i, and use the tree node as the starting position of the capture robot i.

[0061] S2: Loop to build the tree:

[0062] S2.1: Randomly select a point on the map as X_rand, and find the node X_nearest closest to X_rand in the tree Ti;

[0063] S2.2: Based on the search step limit L_limit, determine the position of the new tree node X_new between X_nearest and X_rand;

[0064] S2.3: Boundary detection and collision detection: Set the obstacle list to a collection of static environmental obstacles (trees, stones, etc.) and dynamic obstacles (other capture robots), check whether the line from X_nearest to X_new has exceeded the boundary limit or has collided with an obstacle. If not, add X_new to the tree T and use X_nearesr as the parent node of X_new.

[0065] S3: Target judgment: Check whether the connection between X_new and the target satisfies boundary detection and collision detection. If the connection is successful, the path finding is successful.

[0066] S4: Path extraction and smoothing:

[0067] S4.1: Starting from the target pose, traverse the tree backwards and store the node sequence to the root node as the planned path.

[0068] S4.2: Smooth the stored node sequence; use a spline curve to smooth the path.

[0069] S4.3: Returns the interpolated smoothed path point sequence.

[0070] In the above-mentioned collaborative planning method for the capture path, the collision detection in sub-step S2.3 not only needs to detect the collision between the robot and the obstacle, but also needs to perform synchronous collision detection on other capture robots as dynamic obstacles.

[0071] The target posture in step S3 is different from that in single-machine planning. The target posture of each capture robot is determined by the posture of the target robot and the number of robots. That is, if there are n capture robots, the target posture of the i-th robot is (target robot position, angle + 360 / n∈(0, 360)).

[0072] In sub-step S4.2, B-spline curve is used to achieve path smoothing, ensuring the robot's kinematic constraints while obtaining continuous posture changes.

[0073] The capture path planning method provided by the present invention can realize collaborative capture operations by multiple robots without colliding with the environment or avoiding collisions with other capture robots. The capture path planning method relies on the simulation verification platform for algorithm verification, providing a case illustration for the verification of capture strategy development in the future.

[0074] Implementation steps of the simulation verification platform: First, open the capture algorithm to run and monitor, open the server and client, and log in to the client by setting the account-password; enter the role selection interface, select the capturer or escaper, their appearance is distinguished by color; enter the map selection interface, at this time you can choose to create a map or join a map, generally the first user to log in needs to create a map, and the rest of the users can directly choose to join the map; the map creator needs to set the map information, including map size, terrain information, the number of obstacles such as trees and stones, etc., and the map information is sent to the control algorithm layer through the MQTT protocol; after all users enter the map, they enter the waiting interface, determine the role initialization position and control method, and send it to the control algorithm layer through the MQTT protocol; if the role is selected as manual control, the virtual simulation environment layer controls the robot movement by reading the peripheral control signal; if the role is selected as automatic control, the simulation verification platform starts to monitor the topic sent by the control algorithm layer to control the robot movement; when the escaper is successfully captured or the esc exit signal is detected, the simulation is exited.

[0075] The scenarios in the simulation verification platform provided by the present invention include: scenario 1: the capture capability in sunny days, undulating terrain, and sparse obstacle scenarios (focusing on verifying the collaborative capture algorithm under conditions of high visibility); scenario 2: the capture capability in cloudy days, flat terrain, and large obstacle scenarios (focusing on verifying the collaborative capture algorithm under conditions of general visibility and a narrow accessible range); scenario 3: the capture capability in rainy days, flat terrain, and dense obstacle scenarios (focusing on verifying the collaborative capture algorithm under conditions of general visibility and in dense obstacle environments); scenario 4: the capture capability in foggy days, flat terrain, and sparse obstacle scenarios (focusing on verifying the collaborative capture algorithm capability under conditions of poor visibility).

[0076] Algorithm test: Taking sunny day, undulating terrain, and sparse obstacle scene maps as examples (map size: terrain height: number of trees: tree size: number of stones: stone size), 2 to 6 robots are used as the number of robots for encirclement and capture, and the relative speed is set to level 1 to 3. The test is conducted three times, in which the target robot is completed manually to determine the average encirclement and capture time. Table 1 shows the test results of the encirclement and capture time of different speeds and different numbers of encirclement and capture robots.

[0077] Table 1 Test results of the capture time of different speeds and different numbers of capture robots:

[0078] Relative speed Average capture time when the number of capture robots is 2 Average capture time when the number of capture robots is 3 Average capture time when the number of capture robots is 4 Average capture time when the number of capture robots is 5 Average capture time when the number of capture robots is 6 1 40.23s 33.27s 18.62s 10.88s 5.93s 2 35.78s 28.94s 10.34s 6.55s 3.26s 3 28.90s 15.29s 8.88s 3.22s 1.08s

[0079] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0080] The above embodiments are only used to illustrate the design ideas and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, any equivalent changes or modifications made based on the principles and design ideas disclosed by the present invention are within the protection scope of the present invention.

Claims

1. A swarm robot roundup simulation verification platform, characterized in that: include: The virtual simulation environment layer is used to build a virtual capture simulation platform environment to realize the visualization of multi-robot collaborative capture in different scenarios; The virtual simulation environment layer includes a user login interface module, a role selection module, a capture scene generation module, a capture scene matching and synchronization module, and a role control and visualization module; wherein: The user login interface is used to collect the user name and password information input by the user, and send this information to the server through a script for verification; The role selection module includes the role selection of the capturer or escaper; the escaper controls the movement trajectory of the escaper manually through an external device, and the capturer controls the capture strategy and movement state through algorithm design; The capture scene generation module includes map creation and map addition selection, wherein map creation is used to generate a new random map, and map addition is used to add a pre-set map; in the map creation selection interface, the capture scene is controlled by custom options; the custom options include weather selection, terrain setting, map size, obstacle setting, map naming and save path selection; The capture scene matching and synchronization module is used to set the initial position of the character, including manually inputting the coordinates of the initial position of the character or randomly generating the initial position of the character in the map; detecting the type and number of the character through the server, and when all the characters are ready, synchronously entering the map and realizing real-time visualization on the map; The character control and visualization module are used to select manual control or automatic control of the robot to achieve the change of its motion state; each robot character includes a first-person perspective, a third-person perspective and a free perspective; the global map information is visualized above the map at the same time, including map boundaries, obstacle location information, and location information of all robots in the map; The control algorithm layer is responsible for the motion control calculation of the robots in the cluster capture scene and realizes the control of the motion trajectory of all robot characters; The communication layer is used to provide an information transmission channel between the virtual simulation environment and the control algorithm layer, and realize data communication between the control algorithm layer and the virtual simulation environment layer.

2. The swarm robot roundup simulation verification platform according to claim 1 is characterized in that: In the capture scene generation module, the weather options include sunny, cloudy, foggy and rainy days; the terrain settings include the highest terrain height and the lowest terrain height settings; the obstacle settings include the number and location of obstacles, and the obstacles include trees and stones; Set the control method of the role in the roundup scene generation module, including manual control and automatic control; manual control means that the role's movement direction is controlled by an external keyboard, and automatic control means that the role's movement direction is controlled by the control algorithm layer.

3. The swarm robot roundup simulation verification platform according to claim 1 is characterized in that: In the character control and visualization module, selecting manual control or automatic control of a robot includes: selecting a manually controlled robot to control its movement direction and speed through the keyboard, and controlling the rotation of its viewing angle through the mouse; selecting an automatically controlled robot to change its movement state through control data transmitted from the communication layer.

4. The swarm robot roundup simulation verification platform according to claim 1, characterized in that: The control algorithm layer controls the motion trajectory of all robot characters and plans the next motion posture through the information of the robot's own position, other robots' positions, obstacle positions, target positions, and map boundaries.

5. The swarm robot roundup simulation verification platform according to claim 1, characterized in that: The communication layer is implemented based on the message queue telemetry transmission protocol. It communicates with the server by subscribing to the map and robot position topics, and decodes the data format of the subscribed topics through the callback function to realize data communication between the control algorithm layer and the virtual simulation environment layer.

6. A method for collaborative planning of a capture path based on the cluster robot capture simulation verification platform of claim 1, characterized in that: The following steps are involved: S1: Initialize the map and tree: Build the map based on the communication information of the simulation verification platform, including the map boundary and obstacle information; Construct multiple tree nodes for capturing robots; wherein the root node of the tree is created as the starting position of all robots, and then an empty tree Ti is initialized for each capturing robot i, and the tree node is used as the starting position of the capturing robot i; S2: loop to build tree; S2.1: Randomly select a point on the map as X_rand, and find the node X_nearest closest to X_rand in the tree Ti; S2.2: Based on the search step limit L_limit, determine the position of the new tree node X_new between X_nearest and X_rand; S2.3: Boundary detection and collision detection: Set the obstacle list to the static environment obstacle and dynamic obstacle set, check whether the line from X_nearest to X_new has exceeded the boundary limit or collided with an obstacle, if not, add X_new to the tree T and make X_nearesr the parent node of X_new; S3: Target judgment: Check whether the connection between X_new and the target satisfies the boundary detection and collision detection. If the connection is successful, the path finding is successful. S4: Path extraction and smoothing: S4.1: Starting from the target pose, traverse the tree backwards and store the node sequence to the root node as the planned path; S4.2: Smoothing the stored node sequence, using a spline curve to smooth the path; S4.3: Returns the interpolated smoothed path point sequence.

7. The method for collaborative planning of roundup paths according to claim 6, characterized in that: In the sub-step S2.3, the collision detection includes: detecting the collision between the robot and the obstacle, and performing synchronous collision detection on other capture robots as dynamic obstacles.

8. The method for collaborative planning of roundup paths according to claim 6, characterized in that: In step S3, the target position of each capture robot is determined by the position of the target robot and the number of robots.

9. The method for collaborative planning of a capture path according to claim 6, characterized in that: In the sub-step S4.2, a B-spline curve is used to achieve path smoothing, thereby ensuring the kinematic constraints of the robot while obtaining continuous posture changes.

Citation Information

Patent Citations

  • Multi-agent deep reinforcement learning-based method for enforcing hunting by using obstacles

    CN117400248A

  • Multi-agent cluster obstacle avoidance method based on rolling fast expansion random tree algorithm

    CN117850476A

  • Multi-agent hunting method based on double-layer graph attention reinforcement learning

    CN118153621A

  • Mobile robot moving autonomously under environment with obstruction, and control method for mobile robot

    JP2007249632A

  • Obstacle recognition method for autonomous robots

    US20200225673A1