Multi-robot cooperative pursuit method and system based on elliptic robustness control
By employing a multi-robot cooperative pursuit method with elliptic robust control, the high-cost training and local optima problems in existing technologies are solved, achieving efficient multi-robot pursuit and adapting to pursuit tasks in complex environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SUN YAT SEN UNIV
- Filing Date
- 2023-02-20
- Publication Date
- 2026-05-29
AI Technical Summary
Existing multi-robot pursuit methods require a large number of samples for training, which is costly. Furthermore, the neural network parameters trained in simulation environments do not perform well in practice. They lack cooperation with other intelligent agents, are prone to getting trapped in local optima, and do not consider the existence of no-go zones, which limits their application scenarios.
A multi-robot cooperative pursuit method based on elliptic robust control is adopted. By defining the pursuit space, constructing kinematic equations, meshing and searching, and optimizing and updating the path, dynamic target positions are generated, enabling collision avoidance and cooperative pursuit among robots.
It improves the efficiency of multi-robot pursuit, solves the problem of pursuit dead spots, avoids high-cost training, enhances the cooperation ability between robots, and adapts to pursuit tasks in complex environments.
Smart Images
Figure CN116165893B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of pursuit and capture technology, and in particular to a multi-robot cooperative pursuit method and system based on elliptic robustness control. Background Technology
[0002] Swarm robot systems are mobile distributed systems characterized by high density, robustness, scalability, and flexibility. Therefore, swarm robots face numerous challenges in dynamic multi-target pursuit in complex environments. Before addressing the problem of multi-robot cooperative pursuit, the planning and control of individual robots must first be solved. This includes achieving local localization and path planning for individual robots. For example, obtaining the relative positions of pursuers and escapees to serve the global localization and path planning of the multi-robot system; obtaining the relative positions of pursuers to prevent collisions and maximize the capture of escapees; and obtaining the relative positions of robots and restricted areas to prevent robots from entering restricted areas. From the local localization and path planning of individual robots, the global localization and path planning of the multi-robot system are derived. Finally, a relevant allocation mechanism is used for task decomposition and allocation to achieve the planning and control of individual robots in complex environments. Existing technologies include robot pursuit methods based on multi-agent reinforcement learning. These methods, through steps such as constructing a two-pursuit-one-escape environment, building a Markov model, obtaining a two-pursuit-one-escape network model, and extending multi-pursuit-multi-escape strategies, make the pursuit process more efficient and reliable. The multi-agent distributed encirclement method based on the uncertainty of the escapee's location considers a scenario where multiple pursuers surround multiple escapees. The design purpose of each pursuer's strategy is to continuously reduce the area of the escapee's safe reach over time, ultimately leaving the escapee with nowhere to hide and thus achieving capture—that is, the area minimization strategy. However, the above method still has the following problems: it requires a large number of samples for learning, which increases the cost of use. Secondly, when training the pursuit strategy, the escapee's strategy must be given accordingly to fully construct the simulation environment. The quality of the escapee's decision greatly affects the upper limit of the pursuit robot's decision-making ability, which leads to the neural network parameters trained in the simulation environment performing poorly in experiments. It is necessary to collect real data in experiments to retrain and correct each parameter, which is too time-consuming. In some cases, the robot will make passive decisions, which can also be called a lazy state. It lacks cooperation with other agents and gets stuck in a local optimum that it cannot escape. Most existing methods do not consider the existence of no-entry zones, which greatly limits the application scenarios of the algorithm in real-world scenarios. Summary of the Invention
[0003] To address the aforementioned technical problems, the present invention aims to provide a multi-robot cooperative pursuit method and system based on elliptic robustness control, which can improve the efficiency of robot cooperative pursuit by solving the problem of pursuit dead points and achieving collision avoidance between pursuers.
[0004] The first technical solution adopted in this invention is: a multi-robot cooperative pursuit method based on elliptic robustness control, comprising the following steps:
[0005] Define the pursuit space and set the corresponding variable parameters, and construct the pursuit kinematic equations;
[0006] The pursuit space is sequentially divided into grids and searched to obtain the shortest path to the spatial nodes.
[0007] The pursuit kinematics equations are optimized and updated based on an elliptic robust control strategy, and combined with the shortest path of spatial nodes, to generate dynamic target positions.
[0008] The pursuers coordinate to capture the escapee based on the dynamic location of the target.
[0009] Furthermore, the step of defining the pursuit space and setting the corresponding variable parameters to construct the pursuit kinematic equations specifically includes:
[0010] Define the pursuit space and set the corresponding variable parameters, including the escapee's position, the pursuer's position, the speed range of the escapee and the pursuer, and the conditions for determining a successful pursuit.
[0011] The conditions for determining a successful pursuit are that the distance between the escapee and the nearest pursuer is less than a preset distance and the pursuit time is less than a preset pursuit time.
[0012] The kinematic equations for the pursuit are constructed based on the positions of the escapee and the pursuer, as well as the speed ranges of the escapee and the pursuer.
[0013] Furthermore, the expression for the pursuit kinematic equations is as follows:
[0014]
[0015]
[0016] In the above formula, The kinematic equations representing the escapee, Let u represent the kinematic equation of the i-th pursuer. e u represents the maximum speed of the fleeing person. pi Indicates the maximum speed of the pursuer. Indicates the initial position of the escapee. Let N represent the initial position of the i-th pursuer, and N represent the total number of pursuers.
[0017] Furthermore, the step of sequentially performing grid division and retrieval processing on the pursuit space to obtain the shortest path for spatial nodes specifically includes:
[0018] The space for pursuit is divided into spatial grids to obtain a spatial grid matrix.
[0019] The spatial grid matrix is searched to obtain the shortest path between each node position in the spatial grid matrix.
[0020] Furthermore, the step of optimizing and updating the pursuit kinematic equations based on the elliptic robustness control strategy and combining the shortest paths of spatial nodes to generate the dynamic target position specifically includes:
[0021] The target location area of the escapee is obtained based on the kinematic equations of pursuit.
[0022] Based on the final target location area of the escapee, a target point is assigned to each pursuer according to the pursuit principle, which is to determine whether the distance between the pursuer and the target point is less than the shortest path of the spatial node.
[0023] Calculate the distance between each pursuer and the corresponding target point and normalize it to obtain the pursuit range of each pursuer;
[0024] Based on the pursuit range of each pursuer, target points that meet the pursuit principles are assigned to each pursuer, generating dynamic target locations.
[0025] Furthermore, the expression for the pursuit range of each pursuer is as follows:
[0026]
[0027] In the above formula, A(δ1, δ2) represents the pursuit range of the pursuer, δ1 and δ2 represent the positional parameters within the pursuit range, x represents the predicted location of the escapee, and path(M, x) e (t) represents the time t in which the escapee moves from position x. e The area d traversed on the way to location x M (x e Let x) represent the location value function of the escapee, and P represent the predicted destination area of the escapee in time t, i.e., the final target area.
[0028] Furthermore, the step of obtaining the target location area of the escapee based on the pursuit kinematic equations specifically includes:
[0029] Based on the pursuit space, the time required for pursuit between two points in the pursuit space is obtained according to the pursuit kinematic equations.
[0030] Introduce space constraints for escapees and define the search space for escapees;
[0031] The escapee's danger space is predicted based on the escapee's search space. The escapee's danger space is the set of all spatial nodes in which the time it takes for the pursuer to reach a certain spatial node is less than the time it takes for the escapee to reach that spatial node.
[0032] The search space and danger space of the escapee are intersected to obtain the prediction space of the escapee;
[0033] By introducing a location value function to select the prediction space of the escapee, the target location area of the escapee can be obtained.
[0034] Furthermore, the expression for the target location region of the escapee is as follows:
[0035]
[0036]
[0037]
[0038] In the above formula, V(x) represents the location value function, S1 represents the prediction space of the escapee, and V min V represents the minimum value of the escapee's position value function. max Let x represent the maximum value of the value function, x represent a node in the pursuit space, and P represent the target location region of the escapee.
[0039] Furthermore, the step of the pursuers coordinating the pursuit of the escapee based on the dynamic target location specifically includes:
[0040] The distance between the first pursuer and the escapee is set to be smaller than the distance between the other pursuers and the escapee;
[0041] Assign the dynamic target location of the escapee to the first pursuer;
[0042] The first pursuer tracks the escapee's dynamic target location and updates the distance between the pursuer and the escapee in real time.
[0043] Until the distance between the second pursuer and the escapee is less than the distance between the first pursuer and the escapee;
[0044] The dynamic target location of the escapee is assigned to the second pursuer;
[0045] The second pursuer tracks the escapee's dynamic target location and updates the distance between the pursuer and the escapee in real time.
[0046] Repeat the above collaborative pursuit steps until the conditions for successful pursuit are met or the preset maximum pursuit time is exceeded, at which point the pursuit stops.
[0047] The second technical solution adopted in this invention is: a multi-robot cooperative pursuit system based on elliptic robustness control, comprising:
[0048] The building module is used to define the pursuit space and set the corresponding variable parameters, and to construct the pursuit kinematic equations;
[0049] The acquisition module is used to sequentially perform grid division and retrieval processing on the pursuit space to obtain the shortest path of the spatial nodes;
[0050] The generation module optimizes and updates the pursuit kinematic equations based on the elliptic robustness control strategy and combines the shortest paths of spatial nodes to generate dynamic target positions.
[0051] The pursuit module is used by pursuers to coordinate the pursuit of escapees based on the dynamic location of the target.
[0052] The beneficial effects of the method and system of this invention are as follows: This invention, through the constructed pursuit kinematic equations, observes the self-perceived pose, speed, direction angle, and relative distance and azimuth angle between each robot, achieving mutual positioning. Furthermore, it performs grid partitioning of the pursuit space, dividing multi-robot path planning into global path planning and local path planning. Based on feasibility and path length, it plans the global shortest path, and then performs local path planning, solving the dead-point problem and achieving collision avoidance. Using elliptical robust control avoids the huge training costs and other problems associated with reinforcement learning and other methods. For the potential escape behavior of the target during pursuit, it considers the range of the Apollonius circle, designs action strategies for the pursuit robots, and proposes a method based on dynamic virtual range, further improving pursuit efficiency. Attached Figure Description
[0053] Figure 1 This is a flowchart of the steps of the multi-robot cooperative pursuit method based on elliptic robustness control of the present invention;
[0054] Figure 2 This is a structural block diagram of the multi-robot cooperative pursuit system based on elliptic robustness control of the present invention;
[0055] Figure 3 This is a schematic diagram of the steps involved in the coordinated pursuit and capture of escapees according to the present invention;
[0056] Figure 4 This is a schematic diagram of the structure of the present invention for solving the shortest node path based on spatial grid matrix retrieval;
[0057] Figure 5 This is a simulation diagram of the pursuit scenario of the present invention;
[0058] Figure 6 This is a simulation diagram of the enclosed simulation experiment of the present invention. Detailed Implementation
[0059] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. The step numbers in the following embodiments are only for ease of explanation and do not limit the order of the steps. The execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.
[0060] To capture a high-speed, intelligent escapee, a successful pursuit strategy should accurately predict the escapee's further location and leverage numerical superiority to reduce the escapee's predictable reachability set. Therefore, this invention first proposes a reachability set prediction method from the escapee's perspective, and then designs a cooperative pursuit strategy to assign reasonable targets to each pursuer. The strategy of this invention is based on game theory, considering the interaction between the escapee's and pursuer's decisions. First, it attempts to assess the impact of the pursuer's location on the escapee's decision; this invention uses a region to describe the pursuer's influence on the escapee. Second, this invention further attempts to predict what will happen to the escapee in the next period, trying to think from the escapee's perspective about what it would do in this situation. In this process, it assumes that the escapee possesses a certain level of intelligence and is capable of making rational decisions. Third, based on the predictions made, this invention generates targets for each pursuer and guides them to appropriate locations.
[0061] Reference Figure 1 and Figure 3 This invention provides a multi-robot cooperative pursuit method based on elliptic robustness control, which includes the following steps:
[0062] S1. Set variable parameters, including the location of the escapee, the location of the pursuer, the speed range of the escapee and the pursuer, and the conditions for judging the success of the pursuit.
[0063] Specifically, first let x pi (t) and x e (t) represents the positions of the i-th pursuer and the escapee at time t, respectively, and the maximum speeds of the escapee and the pursuer are u and u, respectively. e and u pi The kinematic equations can be expressed as:
[0064]
[0065]
[0066] In the above formula, The kinematic equations representing the escapee, Let u represent the kinematic equation of the i-th pursuer. e u represents the maximum speed of the fleeing person.pi Indicates the maximum speed of the pursuer. Indicates the initial position of the escapee. Let N represent the initial position of the i-th pursuer, and N represent the total number of pursuers.
[0067] Let r c Indicates the pursuit radius, t max d represents the upper limit of the duration of the pursuit process. min (t) represents the distance between the escapee and the nearest pursuer at time t, so a successful pursuit can be represented as:
[0068] d min (t c )<r c , t c <t max
[0069] In the above formula, d min (t c () represents the distance between the escapee and the nearest pursuer at time t, and r is the distance between them. c Indicates the pursuit radius, t max t represents the upper limit of the duration of the pursuit process. c This indicates the actual time spent in the pursuit.
[0070] S2, Single robot path planning;
[0071] Specifically, spatial partitioning is based on a grid. Referring to the ensemble learning concept in machine learning, the spatial region is divided into a grid and a graph structure is constructed. The distance between two points in space is defined as the number of grid cells traversed. Figure 4 As shown;
[0072] To more closely approximate the optimal path in real-world conditions, this invention employs the Floyd algorithm and obtains different spatial grid matrices by gradually changing the angle of grid generation. For a given pair of starting and ending points, the method of this invention can search among these multiple matrices and select the optimal path, thereby significantly reducing the negative impact of using grids.
[0073] S3, Set the pursuit strategy for the pursuers.
[0074] Specifically, the pursuit strategy is designed to leverage the limited number of pursuers and space to reduce the reachable area of the escapees. The strategy first requires predicting the reachable area of the escapees and assumes that the escapees have a certain level of intelligence and will make rational decisions, so they will go to the location that is most valuable to them. Therefore, the most likely location of the escapees can be predicted, and a target location can be generated for each pursuer, and the pursuers can move towards the target location.
[0075] S31. Obtain the target location area of the escapee;
[0076] Specifically, for a given pursuit space M, the distance between x1 and x2 is first defined as d. M (x1, x2), and let it be equal to the time taken to move from x1 to x2, as shown in the equation:
[0077] d M (x1, x2) = cost(M, x1, x2)
[0078] In the above formula, d M (x1, x2) represents the distance between two nodes in the pursuit space;
[0079] The search is conducted within a limited area, therefore the present invention adds additional restrictions, making T predict This can be represented as a configurable constant used to control the size of the S0 region, i.e., the number of time steps to be predicted. Therefore, the current position x of the escapee can be defined. e The search region S0 is shown in the following formula:
[0080] S0={x|d M (x e x)≤T predict}
[0081] In the above formula, S0 represents the search region, and T predict d represents the prediction time steps. M (x e (x) represents the escapee starting from the current position x. e The distance to move to the next position x;
[0082] To predict the areas a fugitive might go to, we first exclude areas the fugitive is unlikely to go to, i.e., areas that are dangerous for the fugitive. If a pursuer arrives at location x in a shorter time than the fugitive arrives, then location x is dangerous for the fugitive. The criteria for determining a dangerous area are:
[0083]
[0084] In the above formula, d M (x pi (x) represents the current location of the pursuer. pi The distance to the escapee's next position x;
[0085] Further define d MP (x) represents the minimum distance from the hunter group to position x, and its expression is as follows:
[0086]
[0087] In the above formula, d MP (x) represents the minimum distance from the group of pursuers to position x;
[0088] In summary, the expression for the danger zone of an escapee is as follows:
[0089] H={x|d MP (x)≤d M (x e ,x)}
[0090] In the above formula, H represents the danger zone for the escapee;
[0091] Assuming the escapee possesses some intelligence, it is unlikely to travel to a dangerous coordinate or a coordinate that would pass through region H. Let path(M, x) e ,x) represents starting from position x e Since the area traversed by the escapee to position x is known, we can first define the predicted region S1 for the escapee, as shown in the following expression:
[0092]
[0093] In the above formula, S1 represents the predicted region for the escapee, and path(M, x) e ,x) represents starting from position x e Head to the area traversed by location x;
[0094] Because space is limited, assuming the escapee prefers a more distant point, the escapee's positional value function can be defined as:
[0095] V(x)=d M (x e ,x)
[0096] In the above formula, V(x) represents the positional value function of the escapee;
[0097] Because there may be multiple x positions that maximize the value function, and other positions, although not the maximum, still have some reference value, we use a normalization method to obtain these relatively good positions x. Let θ∈[0,1] be a screening threshold set by this invention (0.5 is selected in this invention). Finally, we can obtain the final prediction region, that is, the escapee's position at time T. predict The expression for the area P, which the escapee is highly likely to go to, i.e., the target location area, is as follows:
[0098]
[0099]
[0100]
[0101] In the above formula, V(x) represents the location value function, S1 represents the prediction space of the escapee, and V min V represents the minimum value of the escapee's position value function. max Let x represent the maximum value of the value function, x represent a node in the pursuit space, and P represent the target location region of the escapee.
[0102] S32. Obtain the pursuit range of the pursuers;
[0103] Specifically, given the intelligence of the escapee, it is clear that the farther away the escapee is from the pursuer, the more likely the escapee is to go. Therefore, after obtaining the predicted area of the escapee, target points can be assigned to the pursuer based on the distance between the points in the predicted area and the pursuer. The pursuit method of this invention will follow a principle: the pursuer who has a higher cost to reach the escapee will take the more distant predicted point as the target. First, the escapees are sorted from smallest to largest according to the distance of each pursuer to the escapee, resulting in escapees pk(1), pk(2)...pk(n), whose expressions are as follows:
[0104] d M (x pk(1) x e )≤d M (x pk(2) x e )≤…≤d M (x pk(n) x e )
[0105] Further define a location region A, which accepts two parameters δ1 and δ2 [0, 1] to control the distance range between locations in region A and the escapee's current location. That is, using normalization, assume d max It is the distance to the farthest position in P, A(δ1, δ2) is the distance from the current position of the escapee selected from set P. max *δ1,d max Points within the range of *δ2) are defined, thus planning different pursuit ranges for each pursuer and ensuring the basis for their cooperation. The expression is as follows:
[0106]
[0107] In the above formula, A(δ1, δ2) represents the pursuit range of the pursuer, δ1 and δ2 represent the positional parameters within the pursuit range, x represents the predicted location of the escapee, and path(M, x) e (t) represents the time t in which the escapee moves from position x. e The area d traversed on the way to location x M (x eLet x) represent the location value function of the escapee, and P represent the predicted destination area of the escapee in time t, i.e., the final target area.
[0108] S33. Generate the dynamic target location information of the escapee and coordinate the pursuit by the pursuers.
[0109] Specifically, using the aforementioned region A, a target location can be assigned to each pursuer. In the case of n pursuers and one escapee, for the pursuer pk(i) closest to the escapee, the region A is assigned to the target location. The nearest position to it is assigned as its target position a. k(i) Its expression is as follows:
[0110]
[0111] Finally, after each time step, a new position is assigned to each pursuer based on their target location. That is, the increment dx(M, x1, x2) of position x1 after one time step dt on the path from the escapee x1 to the target location x2 is obtained, and its expression is as follows:
[0112] dx(M,x1,x2)={x-x1|x∈path(M,x1,x2),d M (x1, x) = dt}
[0113] In the above formula, x1 represents the current position of the escapee, x2 represents the target position of the escapee, dx(M, x1, x2) represents the position increment of the escapee, and dt represents a time step.
[0114] The method described in this invention can be applied to practical scenarios such as maritime search and rescue using drones, in order to address situations where the location of victims is difficult to predict due to random drifting at sea, which affects search and rescue efforts.
[0115] Reference Figure 2 A multi-robot cooperative pursuit system based on elliptic robust control includes:
[0116] The building module is used to define the pursuit space and set the corresponding variable parameters, and to construct the pursuit kinematic equations;
[0117] The acquisition module is used to sequentially perform grid division and retrieval processing on the pursuit space to obtain the shortest path of the spatial nodes;
[0118] The generation module optimizes and updates the pursuit kinematic equations based on the elliptic robustness control strategy and combines the shortest paths of spatial nodes to generate dynamic target positions.
[0119] The pursuit module is used by pursuers to coordinate the pursuit of escapees based on the dynamic location of the target.
[0120] The simulation experiment of this invention is as follows:
[0121] Reference Figure 5 ,in Figure 5 (a) indicates the initial state of the pursuit. The three dots below represent the pursuer, the dots above represent the escapee, the white area represents the escapee's safe zone, the gray area represents the danger zone, and the black area represents obstacles. Figure 5 (b) Increase the prediction of escape routes for escapees; Figure 5 (c) Add a division of escape routes by distance; Figure 5 (d) indicates that a target is assigned to the pursuers;
[0122] Reference Figure 6 Four encirclement scenarios are presented under different contexts: Figure 6 (a) indicates enclosed. Figure 6 (b) indicates flanking around an obstacle. Figure 6 (c) indicates cooperative approximation. Figure 6 (d) indicates splitting up to act separately.
[0123] In summary, this invention, in the area of multi-robot cooperative localization, employs a cooperative localization method based on relative observations. It utilizes each robot's perceived pose, velocity, direction angle, and the relative distance and azimuth between robots to achieve mutual localization. Furthermore, based on this, the target can be located according to the geometric relationship between the robot and the target. In the robot localization problem, the method based on relative observations uses the Kalman principle to guess the target's position with a certain probability, and the final result is obtained after convergence. During the pursuit process, this problem can be extended to guessing the target's position as being distributed within a certain range with a certain probability, and then studying the convergence of this range to determine whether the pursuit is complete.
[0124] In multi-robot path planning, it is divided into global path planning and local path planning. The fitness function is designed based on feasibility, path length and smoothness. The genetic algorithm is used to plan the global shortest path, and then the rolling window method is used for local path planning to solve the dead point problem and achieve the collision avoidance effect.
[0125] In terms of multi-robot pursuit algorithms, this approach avoids a series of problems such as the huge training costs associated with methods like reinforcement learning. Considering the potential escape behavior of targets during pursuit, the range of the Apollonius circle is taken into account, and action strategies are designed for the pursuit robots. A method based on a dynamic virtual range is proposed. Simulation experiments demonstrate that the pursuit efficiency is improved through the statistical analysis of the number of pursuit steps. The key technologies of multi-robot collaboration mentioned above are verified using a multi-robot pursuit system as an application background.
[0126] The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.
[0127] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.
Claims
1. A multi-robot cooperative pursuit method based on elliptic robust control, characterized in that, Includes the following steps: Define the pursuit space and set the corresponding variable parameters, and construct the pursuit kinematic equations; The pursuit space is sequentially divided into grids and searched to obtain the shortest path to the spatial nodes. The pursuit kinematics equations are optimized and updated based on an elliptic robust control strategy, and combined with the shortest path of spatial nodes, to generate dynamic target positions. The pursuers coordinate their efforts to apprehend the escapee based on the dynamic location of the target. The step of optimizing and updating the pursuit kinematic equations based on the elliptic robust control strategy and generating the dynamic target position by combining the shortest paths of spatial nodes specifically includes: The target location area of the escapee is obtained based on the kinematic equations of pursuit. Based on the final target location area of the escapee, a target point is assigned to each pursuer according to the pursuit principle, which is to determine whether the distance between the pursuer and the target point is less than the shortest path of the spatial node. Calculate the distance between each pursuer and the corresponding target point and normalize it to obtain the pursuit range of each pursuer; Based on the pursuit range of each pursuer, target points that meet the pursuit principles are assigned to each pursuer, and dynamic target locations are generated. The expression for the pursuit range of each pursuer is as follows: In the above formula, Indicates the area of pursuit by the pursuers. , This represents the location parameters within the pursuit range. This indicates the predicted location of the escapee. Indicates the time of the escapee From the position Go to location The areas traversed The location value function represents the position of the escapee. Indicates the time of the escapee The predicted arrival area within the area is the final target area; The step of obtaining the target location area of the escapee based on the pursuit kinematic equations specifically includes: Based on the pursuit space, the time required for pursuit between two points in the pursuit space is obtained according to the pursuit kinematic equations. Introduce space constraints for escapees and define the search space for escapees; The escapee's danger space is predicted based on the escapee's search space. The escapee's danger space is the set of all spatial nodes in which the time it takes for the pursuer to reach a certain spatial node is less than the time it takes for the escapee to reach that spatial node. The search space and danger space of the escapee are intersected to obtain the prediction space of the escapee; By introducing a location value function to select the prediction space of the escapee, the target location area of the escapee can be obtained; The expression for the target location region of the escapee is as follows: In the above formula, Represents the location value function. Represents the prediction space for escapees. This represents the minimum value of the escapee's location-value function. This represents the maximum value of the value function. This represents a node within the pursuit space. This indicates the target location area of the escapee.
2. The multi-robot cooperative pursuit method based on elliptic robust control according to claim 1, characterized in that, The step of defining the pursuit space and setting the corresponding variable parameters to construct the pursuit kinematic equations specifically includes: Define the pursuit space and set the corresponding variable parameters, including the escapee's position, the pursuer's position, the speed range of the escapee and the pursuer, and the conditions for determining a successful pursuit. The conditions for determining a successful pursuit are that the distance between the escapee and the nearest pursuer is less than a preset distance and the pursuit time is less than a preset pursuit time. The kinematic equations for the pursuit are constructed based on the positions of the escapee and the pursuer, as well as the speed ranges of the escapee and the pursuer.
3. The multi-robot cooperative pursuit method based on elliptic robust control according to claim 2, characterized in that, The expression for the pursuit kinematic equations is as follows: In the above formula, The kinematic equations representing the escapee, Indicates the first The kinematic equations of the pursuers Indicates the maximum speed of the fleeing person. Indicates the maximum speed of the pursuer. Indicates the initial position of the escapee. Indicates the first The initial location of the pursuers This indicates the total number of pursuers.
4. The multi-robot cooperative pursuit method based on elliptic robust control according to claim 3, characterized in that, The step of sequentially dividing and searching the pursuit space to obtain the shortest path to the spatial nodes specifically includes: The space for pursuit is divided into spatial grids to obtain a spatial grid matrix. The spatial grid matrix is searched to obtain the shortest path between each node position in the spatial grid matrix.
5. The multi-robot cooperative pursuit method based on elliptic robust control according to claim 1, characterized in that, The step of the pursuers coordinating the pursuit of the escapee based on the dynamic target location specifically includes: The distance between the first pursuer and the escapee is set to be smaller than the distance between the other pursuers and the escapee; Assign the dynamic target location of the escapee to the first pursuer; The first pursuer tracks the escapee's dynamic target location and updates the distance between the pursuer and the escapee in real time. Until the distance between the second pursuer and the escapee is less than the distance between the first pursuer and the escapee; The dynamic target location of the escapee is assigned to the second pursuer; The second pursuer tracks the escapee's dynamic target location and updates the distance between the pursuer and the escapee in real time. Repeat the above collaborative pursuit steps until the conditions for successful pursuit are met or the preset maximum pursuit time is exceeded, at which point the pursuit stops.
6. A multi-robot cooperative pursuit system based on elliptic robust control, characterized in that, The multi-robot cooperative pursuit method based on elliptic robust control as described in claim 1 includes the following modules: The building module is used to define the pursuit space and set the corresponding variable parameters, and to construct the pursuit kinematic equations; The acquisition module is used to sequentially perform grid division and retrieval processing on the pursuit space to obtain the shortest path of the spatial nodes; The generation module optimizes and updates the pursuit kinematic equations based on the elliptic robustness control strategy and combines the shortest paths of spatial nodes to generate dynamic target positions. The pursuit module is used by pursuers to coordinate the pursuit of escapees based on the dynamic location of the target.