Multi-agent cluster crossing path planning method and device based on social force model

Abstract

The present application relates to a multi-agent cluster crossing path planning method and device based on a social force model, wherein the method includes: obtaining the environmental depth observation information of each target agent and the neighbor posture observation information of the corresponding neighbor agent; determining the interaction force between each target agent, the repulsion of the preset obstacle on each target agent, and the gravitational force of the preset target point on each target agent, and building a social force model in combination with a preset finite state machine; calculating the interaction force, repulsion and gravitational force of the social force model, and performing a state conversion operation in the finite state machine according to the interaction force, repulsion and gravitational force to obtain a state conversion result, so as to generate a target crossing path for multiple target agents through the state conversion result. Thus, the problem that the existing technology cannot coordinate the convergence and divergence behaviors of robots through distributed control at the same time and is difficult to efficiently cross complex unknown environments is solved.

Description

Technical Field

[0001] The present application relates to the technical field of multi-agent path planning, and in particular to a multi-agent cluster crossing path planning method and device based on a social force model. Background Art

[0002] Multi-agent swarms can cope with more complex tasks in search and rescue, transportation, exploration, etc. Multi-agent path planning is a basic problem in the application of multi-agent swarms, especially the problem of crossing in complex unknown environments. A good system path needs to reflect self-organized collaborative behavior. However, existing technologies cannot coordinate the convergence and divergence behaviors of robots through distributed control at the same time, which needs to be solved urgently. Summary of the invention

[0003] The present application provides a multi-agent cluster crossing path planning method and device based on a social force model to solve the problems that the existing technology cannot simultaneously coordinate the convergence and divergence behaviors of robots through distributed control, and is difficult to efficiently cross complex unknown environments.

[0004] The first aspect of the present application provides a multi-agent cluster crossing path planning method based on a social force model, comprising the following steps: obtaining environmental depth observation information of each target agent among multiple target agents and neighbor posture observation information of neighbor agents corresponding to each target agent; determining the interaction force between each target agent, the repulsion of a preset obstacle on each target agent, and the gravitational force of a preset target point on each target agent, and constructing a social force model based on the interaction force, the repulsion, the gravitational force and a preset finite state machine; calculating the interaction force, the repulsion and the gravitational force of the social force model based on the environmental depth observation information and the neighbor posture observation information, and performing a state transition operation in the finite state machine based on the interaction force, the repulsion and the gravitational force to obtain a state transition result, so as to generate the target crossing paths of the multiple target agents through the state transition result.

[0005] Optionally, in one embodiment of the present application, the interaction force between each target intelligent agent, the repulsive force of a preset obstacle on each target intelligent agent, and the gravitational force of a preset target point on each target intelligent agent are determined, and a social force model is constructed based on the interaction force, the repulsive force, the gravitational force, and a preset finite state machine, including: determining the target directions and preset gravitational weight parameters of the multiple target intelligent agents, and calculating the left turn probability of the multiple target intelligent agents, so as to calculate the gravitational force based on the left turn probability, the target direction, and the preset gravitational weight parameters; calculating the distance vectors between the multiple target intelligent agents and the preset obstacles based on the environmental depth observation information, and determining the repulsive force based on the distance vector, the unit vector corresponding to the distance vector, and a preset repulsive potential energy field; calculating the actual distance between any two target intelligent agents among the multiple target intelligent agents, and determining the interaction force based on the actual distance and the preset reference distance, and in combination with the environmental depth observation information of each target intelligent agent.

[0006] Optionally, in one embodiment of the present application, the state transition operation is performed in the finite state machine according to the interaction force, the repulsive force and the gravitational force, including: setting the initial current mode of the target intelligent body to the forward mode, and determining the field of view angle of the target intelligent body; when the preset obstacle exists within the observation range of the field of view angle, and the repulsive force of the preset obstacle on the target intelligent body in the target direction and the gravitational force of the target intelligent body satisfy a first preset balance condition, converting the target intelligent body from the forward mode to the decision mode; based on the interaction force, the repulsive force and the gravitational force, calculating the resultant force of the target intelligent body, and determining the left turn obstacle avoidance probability of the target intelligent body according to the resultant force and a preset obstacle avoidance decision expression, so as to switch the target intelligent body from the decision mode to the left turn obstacle avoidance probability. The target intelligent agent is converted from the obstacle avoidance mode to the obstacle avoidance mode; based on the left turn obstacle avoidance probability and the target direction, the target intelligent agent is determined to move along the target obstacle avoidance direction, and when the repulsive force and the attractive force of the target intelligent agent meet a preset offset condition, the target intelligent agent is converted from the obstacle avoidance mode to the forward mode, or when the repulsive force and the attractive force between the target intelligent agent and the neighbor intelligent agent corresponding to the target intelligent agent in the target direction meet a second preset balance condition, the target intelligent agent is converted from the obstacle avoidance mode to the avoidance mode; when at least two of the multiple target intelligent agents simultaneously enter the avoidance mode, based on the priority order between the at least two target intelligent agents, the at least two target intelligent agents are controlled to sequentially convert from the avoidance mode to the forward mode.

[0007] Optionally, in one embodiment of the present application, the mathematical expression of the first preset balance condition is:

[0008]

[0009] in, represents the repulsive force of the preset obstacle on the target intelligent body; a unit vector representing the direction of the target; represents the gravity; a represents the number of the target intelligent agent; B represents the preset obstacle; Θ s Represents the field of view angle.

[0010] Optionally, in one embodiment of the present application, the mathematical expression of the second preset balance condition is:

[0011]

[0012] in, Represents the interaction force between each target intelligent agent; a unit vector representing the direction of the target; represents the gravity; a represents the number of the target agent; i represents the number of the neighbor agent corresponding to the target agent.

[0013] The second aspect of the present application provides a multi-agent cluster crossing path planning device based on a social force model, including: an acquisition module, used to obtain the environmental depth observation information of each target intelligent agent among multiple target intelligent agents and the neighbor posture observation information of the neighbor intelligent agents corresponding to each target intelligent agent; a modeling module, used to determine the interaction force between each target intelligent agent, the repulsion of a preset obstacle on each target intelligent agent, and the gravitational force of a preset target point on each target intelligent agent, and construct a social force model based on the interaction force, the repulsion, the gravitational force and a preset finite state machine; a generation module, used to calculate the interaction force, the repulsion and the gravitational force of the social force model based on the environmental depth observation information and the neighbor posture observation information, and perform a state transition operation in the finite state machine according to the interaction force, the repulsion and the gravitational force to obtain a state transition result, so as to generate the target crossing path of the multiple target intelligent agents through the state transition result.

[0014] Optionally, in one embodiment of the present application, the modeling module includes: a first calculation unit, used to determine the target direction and preset gravity weight parameters of the multiple target intelligent agents, and calculate the left turn detour probability of the multiple target intelligent agents, so as to calculate the gravity based on the left turn detour probability, the target direction and the preset gravity weight parameters; a determination unit, used to calculate the distance vector between the multiple target intelligent agents and the preset obstacle according to the environmental depth observation information, and determine the repulsive force based on the distance vector, the unit vector corresponding to the distance vector and the preset repulsive potential energy field; a second calculation unit, used to calculate the actual distance between any two target intelligent agents among the multiple target intelligent agents, and determine the interaction force based on the actual distance and the preset reference distance, and in combination with the environmental depth observation information of each target intelligent agent.

[0015] Optionally, in one embodiment of the present application, the generation module includes: a setting unit, which is used to set the initial current mode of the target intelligent agent to the forward mode and determine the field of view angle of the target intelligent agent; a first conversion unit, which is used to convert the target intelligent agent from the forward mode to the decision mode when the preset obstacle exists within the observation range of the field of view angle and the repulsive force of the preset obstacle on the target intelligent agent in the target direction and the gravitational force of the target intelligent agent satisfy a first preset balance condition; a second conversion unit, which is used to calculate the resultant force of the target intelligent agent based on the interaction force, the repulsive force and the gravitational force, and determine the left turn obstacle avoidance probability of the target intelligent agent according to the resultant force and a preset obstacle avoidance decision expression, so as to convert the target intelligent agent from the decision mode to the obstacle avoidance mode through the left turn obstacle avoidance probability; a third conversion unit, which is used to convert the target intelligent agent from the decision mode to the obstacle avoidance mode through the left turn obstacle avoidance probability. A unit for determining the target obstacle avoidance direction of the target intelligent agent based on the left turn obstacle avoidance probability and the target direction, and controlling the target intelligent agent to move along the target obstacle avoidance direction, and converting the target intelligent agent from the obstacle avoidance mode to the forward mode when the repulsive force and the attractive force of the target intelligent agent satisfy a preset offset condition, or converting the target intelligent agent from the obstacle avoidance mode to the avoidance mode when the repulsive force and the attractive force between the target intelligent agent and the neighbor intelligent agent corresponding to the target intelligent agent in the target direction satisfy a second preset balance condition; a fourth conversion unit for controlling the at least two target intelligent agents among the multiple target intelligent agents to convert from the avoidance mode to the forward mode in sequence based on the priority order between the at least two target intelligent agents when at least two target intelligent agents among the multiple target intelligent agents enter the avoidance mode at the same time.

[0016] Optionally, in one embodiment of the present application, the mathematical expression of the first preset balance condition is:

[0017]

[0018] in, represents the repulsive force of the preset obstacle on the target intelligent body; a unit vector representing the direction of the target; represents the gravity; a represents the number of the target intelligent agent; B represents the preset obstacle; Θ s Represents the field of view angle.

[0019] Optionally, in one embodiment of the present application, the mathematical expression of the second preset balance condition is:

[0020]

[0021] in, Represents the interaction force between each target intelligent agent; a unit vector representing the direction of the target; represents the gravity; a represents the number of the target agent; i represents the number of the neighbor agent corresponding to the target agent.

[0022] The third aspect of the present application provides an electronic device, comprising: 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 multi-agent cluster crossing path planning method based on the social force model as described in the above embodiment.

[0023] The fourth aspect of the present application provides a computer-readable storage medium, which stores a computer program. When the program is executed by a processor, it implements the multi-agent cluster crossing path planning method based on the social force model as described above.

[0024] The fifth aspect of the present application provides a computer program product, including a computer program, which is executed to implement the above-mentioned multi-agent cluster crossing path planning method based on the social force model.

[0025] Therefore, the embodiments of the present application have the following beneficial effects:

[0026] The embodiment of the present application can obtain the environmental depth observation information of each target agent in multiple target agents and the neighbor posture observation information of the neighbor agent corresponding to each target agent; determine the interaction force between each target agent, the repulsion of the preset obstacle on each target agent, and the gravitational force of the preset target point on each target agent, and construct a social force model based on the interaction force, repulsion, gravitation and preset finite state machine; based on the environmental depth observation information and the neighbor posture observation information, calculate the interaction force, repulsion and gravitation of the social force model, and perform state conversion operations in the finite state machine according to the interaction force, repulsion and gravitation to obtain the state conversion result, so as to generate the target crossing path of multiple target agents through the state conversion result. The computing cost and communication requirements of the present application are low, the calculation can be completed in real time, and it can be extended to large-scale clusters and deployed in small, low-power robots or complex environments. Thus, the problems that the existing technology cannot coordinate the convergence and divergence behaviors of robots through distributed control at the same time and it is difficult to efficiently cross complex unknown environments are solved.

[0027] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0029] Figure 1 A flowchart of a multi-agent cluster crossing path planning method based on a social force model provided according to an embodiment of the present application;

[0030] Figure 2 A schematic diagram of an intelligent agent observation model provided for one embodiment of the present application;

[0031] Figure 3 A schematic diagram of the logical architecture of a multi-agent cluster traversal path planning method based on a social force model provided for one embodiment of the present application;

[0032] Figure 4 A schematic diagram of a finite state machine provided for one embodiment of the present application;

[0033] Figure 5 A schematic diagram of test results of the present application in a real-world environment provided for an embodiment of the present application;

[0034] Figure 6 This is an example diagram of a multi-agent cluster crossing path planning device based on a social force model according to an embodiment of the present application;

[0035] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0036] Among them, 10-a multi-agent cluster crossing path planning device based on social force model; 100-acquisition module, 200-modeling module, 300-generation module; 701-memory, 702-processor, 703-communication interface. DETAILED DESCRIPTION

[0037] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0038] The following describes the multi-agent cluster crossing path planning method and device based on the social force model of the embodiment of the present application with reference to the accompanying drawings. In view of the problems mentioned in the above background technology, the present application provides a multi-agent cluster crossing path planning method based on the social force model, in which the environmental depth observation information of each target agent in multiple target agents and the neighbor posture observation information of the neighbor agent corresponding to each target agent are obtained; the interaction force between each target agent, the repulsion of the preset obstacle to each target agent, and the gravitational force of the preset target point to each target agent are determined, and the social force model is constructed according to the interaction force, repulsion, gravitation and preset finite state machine; based on the environmental depth observation information and the neighbor posture observation information, the interaction force, repulsion and gravitation of the social force model are calculated, and the state conversion operation is performed in the finite state machine according to the interaction force, repulsion and gravitation, and the state conversion result is obtained, so as to generate the target crossing path of multiple target agents through the state conversion result. The computing cost and communication requirements of the present application are low, the calculation can be completed in real time, and it can be extended to large-scale clusters and deployed in small, low-power robots or complex environments. This solves the problems that the existing technology is unable to coordinate the convergence and divergence behaviors of robots through distributed control, and is difficult to efficiently traverse complex and unknown environments.

[0039] Specifically, Figure 1 A flowchart of a multi-agent cluster crossing path planning method based on a social force model provided in an embodiment of the present application.

[0040] like Figure 1 As shown, the multi-agent cluster crossing path planning method based on the social force model includes the following steps:

[0041] In step S101, the environment depth observation information of each target agent among multiple target agents and the neighbor posture observation information of the neighbor agents corresponding to each target agent are obtained.

[0042] It should be noted that the observation data of the agent in the embodiment of the present application includes the environmental depth observation data (i.e., environmental depth observation information) and the neighbor posture observation data (i.e., neighbor posture observation information) of the neighboring agents corresponding to each agent. For an agent a in the embodiment of the present application, its environmental depth observation data can be represented by the vector x a It is used to store the environment depth data corresponding to the azimuth centered on the agent, and the environment depth observation data of the agent can be transmitted between all neighboring agents within the field of view of the agent using a wireless communication network.

[0043] In addition, if Figure 2 As shown, based on the agent observation model, the embodiment of the present application can a It is divided into four areas, where areas I and II are used to judge the traffic conditions in front of the agent, and areas III and IV are the agent's peripheral vision. The vector composed of the free space areas in the four areas of the agent itself is recorded as x0 = [s 01 ,s 02 ,s 03 ,s 04 ].

[0044] The observation data received by agent a from neighbor agent i is recorded as x i , the observation has been rotated to the local coordinate system of agent a, x i The longest traversable distance in regions I and II is denoted by x ni =[d 01 ,d 02 ].

[0045] For a neighboring agent i of agent a, the relative pose observation includes the distance vector between the two and relative attitude angle

[0046] Therefore, the embodiments of the present application provide reliable data support for the construction of subsequent social force models by obtaining the environmental depth observation information of each intelligent agent and the neighbor posture observation information of neighboring intelligent agents.

[0047] In step S102, the interaction force between each target intelligent agent, the repulsive force of the preset obstacle on each target intelligent agent, and the gravitational force of the preset target point on each target intelligent agent are determined, and a social force model is constructed based on the interaction force, repulsive force, gravitational force and the preset finite state machine.

[0048] Furthermore, the embodiments of the present application also need to determine the gravitational force of the target point on the agent, the repulsive force of the obstacle on the agent, the interaction force between agents, and a finite state machine for executing path planning, such as Figure 3 As shown, a social force model is constructed.

[0049] It can be understood that, based on the crowd behavior model, the embodiment of the present application constructs a social force model of a multi-agent cluster from the mechanism level of group collaboration, which is used to solve the problem of crossing in a complex unknown environment. By introducing the social force model, the multi-agent cluster spontaneously exhibits the effects of aggregation, dispersion and avoidance; in addition, the model has the characteristics of being lightweight, with low computing cost and low communication requirements, and can complete calculations in real time. The model can be expanded to large-scale clusters and deployed on small, low-power robots or complex environments.

[0050] Optionally, in one embodiment of the present application, the interaction force between each target intelligent agent, the repulsion of a preset obstacle on each target intelligent agent, and the gravitational force of a preset target point on each target intelligent agent are determined, and a social force model is constructed based on the interaction force, repulsion, gravitational force and a preset finite state machine, including: determining the target directions and preset gravitational weight parameters of multiple target intelligent agents, and calculating the left turn probability of multiple target intelligent agents, so as to calculate the gravitational force based on the left turn probability, the target direction and the preset gravitational weight parameters; calculating the distance vectors between multiple target intelligent agents and the preset obstacles based on the environmental depth observation information, and determining the repulsion based on the distance vector, the unit vector corresponding to the distance vector and the preset repulsive potential energy field; calculating the actual distance between any two target intelligent agents among the multiple target intelligent agents, and determining the interaction force based on the actual distance and the preset reference distance, and in combination with the environmental depth observation information of each target intelligent agent.

[0051] It should be noted that the gravitational force of the target point on the agent in the embodiment of the present application is: the target direction of the given multi-agent cluster The force that the agent needs to exert in the direction of the target point to reach it.

[0052] This gravitational force is not a real force in the physical sense, but a force modeled for path planning. The same is true for the repulsive force of subsequent obstacles on the agent and the interaction force between agents. Therefore, the gravitational force of the target point on the agent is expressed as:

[0053]

[0054] Among them, w a represents the weight parameter of the force. In the embodiment of the present application, the gravitational weight of the target on the intelligent agent can be increased by adjusting the parameter; P sTo calculate the probability of turning left when encountering an obstacle, the embodiment of the present application may provide a sigmoid function to map environmental observations to a probability domain:

[0055]

[0056] Among them, the sigmoid function W0 is a weight parameter vector, which represents the weight of the agent in each observation quadrant.

[0057] Secondly, in the embodiment of the present application, the repulsive force of the obstacle on the agent is: the repulsive force of the obstacle observed in the agent's observation area I and II on itself, to avoid collision with the obstacle during the agent's forward movement. The distance vector observed by the agent to the obstacle B is The unit vector in this direction is The embodiment of the present application can use the force field to repel obstacles to the intelligent body. The mathematical model is:

[0058]

[0059]

[0060] in, is the repulsive potential energy field, V 0B is the proportional coefficient, and the planning distance of the agent is l th .

[0061] When the distance of obstacles observed in regions I and II is greater than l th , the agent can move forward safely, otherwise the repulsive and gravitational forces Balance, enter the decision mode, set the sum of the observation angles of area I and II as α, then we can calculate:

[0062]

[0063] In addition, the interaction force between intelligent agents in the embodiment of the present application is: the attraction or repulsion between two intelligent agents in the cluster.

[0064] When the distance between the two robots is very close, the force is repulsive. The increase in With further increase, the gravity decays slowly from steady to zero; the embodiment of the present application can also use the force field V ai The interaction force between agents is modeled as follows:

[0065]

[0066]

[0067] Among them, σ a is the reference distance, when the distance between two agents is σ a hour, The force just transitions from repulsion to attraction. According to the different modes of the intelligent agent, the embodiment of the present application constructs a piecewise function Use different calculation expressions in different modes:

[0068]

[0069] Among them, P ni is the probability of the agent following the neighbor. In a two-dimensional environment, the embodiment of the present application divides the neighbor agents into left neighbor agents and right neighbor agents according to the direction of the agent. Correspondingly, P ni The expressions in both cases are:

[0070]

[0071] Among them, 1 is a vector whose elements are all 1; w n is the threshold parameter vector of the two observation areas, which is used to adjust the probability value of following the neighbor direction.

[0072] Therefore, the embodiments of the present application calculate the gravitational force of the target on the agent based on the target point direction of each agent in the multi-agent cluster, and calculate the repulsion of the obstacle on the agent based on the observation of the obstacles in the environment by each agent in the cluster; in addition, the embodiments of the present application calculate the interaction force between agents based on the mutual posture observations between the agents in the cluster, thereby effectively ensuring the reliability of the social force model and providing solid technical support for the conversion of states within the finite state machine.

[0073] In step S103, based on the environment depth observation information and the neighbor posture observation information, the interaction force, repulsion and attraction of the social force model are calculated, and the state transition operation is performed in the finite state machine according to the interaction force, repulsion and attraction to obtain the state transition result, so as to generate the target crossing path of multiple target intelligent agents through the state transition result.

[0074] Furthermore, the finite state machine in the embodiment of the present application includes four execution states: forward mode, decision mode, obstacle avoidance mode and avoidance mode. The intelligent agent calculates the interaction force, repulsion and attraction of the social force model according to its own target point requirements, observation of the environment and mutual observation between intelligent agents, and then completes the state transition in the finite state machine according to the force conditions, ultimately realizing multi-agent collaborative crossing path planning.

[0075] Therefore, the embodiments of the present application can achieve the coordination and avoidance effect of the intelligent agent cluster by relying on the local environment observation of the intelligent agent cluster when the intelligent agents are relatively dense, the perception is limited, and there is no external positioning and environment map, thereby reducing the path cost of the crossing task.

[0076] Optionally, in one embodiment of the present application, a state transition operation is performed in a finite state machine according to the interaction force, repulsion and gravity, including: setting the initial current mode of the target intelligent body to the forward mode, and determining the field of view angle of the target intelligent body; when there is a preset obstacle within the observation range of the field of view angle, and the repulsion of the preset obstacle on the target intelligent body in the target direction and the gravity of the target intelligent body meet the first preset balance condition, the target intelligent body is converted from the forward mode to the decision mode; based on the interaction force, repulsion and gravity, the resultant force of the target intelligent body is calculated, and the left turn obstacle avoidance probability of the target intelligent body is determined according to the resultant force and the preset obstacle avoidance decision expression, so as to switch the target intelligent body from the decision mode to the left turn obstacle avoidance probability through the left turn obstacle avoidance probability. The target intelligent agent is converted to obstacle avoidance mode; based on the left turn obstacle avoidance probability and the target direction, the target intelligent agent is determined to move in the target obstacle avoidance direction, and the target intelligent agent is controlled to move along the target obstacle avoidance direction, and when the repulsion and attraction of the target intelligent agent meet a preset offset condition, the target intelligent agent is converted from the obstacle avoidance mode to the forward mode, or when the repulsion and attraction between the target intelligent agent and the neighbor intelligent agent corresponding to the target intelligent agent in the target direction meet a second preset balance condition, the target intelligent agent is converted from the obstacle avoidance mode to the avoidance mode; when at least two target intelligent agents among multiple target intelligent agents simultaneously enter the avoidance mode, based on the priority order between the at least two target intelligent agents, the at least two target intelligent agents are controlled to convert from the avoidance mode to the forward mode in turn.

[0077] In the actual implementation process, Figure 4 As shown, the finite state machine of the embodiment of the present application includes four execution states: forward mode, decision mode, obstacle avoidance mode and avoidance mode.

[0078] The embodiment of the present application can generally set the initial current mode of the agent to the forward mode. After that, the embodiment of the present application can determine the field of view of the agent. The agent moves towards the target direction under the action of s , when the agent observes the repulsive force of any obstacle B in its field of view Direction and When the balance (i.e., the first preset balance condition) is reached, the agent can switch to the decision-making mode.

[0079] Afterwards, the agent can determine the direction of the obstacle (turn left or turn right) according to the combined force of the social forces it receives. In the embodiment of the present application, the probability of turning left to avoid the obstacle can be calculated using the following calculation formula:

[0080]

[0081]

[0082] The agent selects an obstacle-avoiding direction according to the probability value of P(left) and moves around the obstacle.

[0083] Then, the agent selects a distance along the obstacle avoidance direction. Move in the nearest safe direction and adjust the direction according to real-time environmental observations, keeping a certain distance from obstacles while getting as close as possible. Direction. When the repulsive force of the obstacle cannot offset the gravitational force of the agent toward the goal, as shown in the following equation, the agent returns to the forward mode.

[0084]

[0085] When the repulsive force between the agent and its neighbors is Direction and When the balance is reached (i.e., the second preset balance condition), the agent switches to the avoidance mode; in the avoidance mode, the agent with a higher priority among its neighbors returns to the forward mode first, and the other agents avoid and wait in place. As the relative positions between the agents change, when the second preset balance condition is no longer met, the agent exits the avoidance mode and returns to the forward mode.

[0086] Therefore, the embodiment of the present application allows the multi-agent cluster to spontaneously exhibit the effects of aggregation, dispersion and avoidance in the crossing scenario in a two-dimensional complex unknown environment, thereby optimizing the crossing path of the multi-agent system.

[0087] Optionally, in one embodiment of the present application, the mathematical expression of the first preset balance condition is:

[0088]

[0089] in, Represents the repulsive force of the preset obstacle on the target agent; A unit vector representing the target direction; represents gravity; a represents the number of the target agent; B represents the preset obstacle; Θ s Indicates the field of view angle.

[0090] It should be noted that the embodiment of the present application can observe the repulsive force of any obstacle B in the field of vision of the intelligent agent. Direction and When the balance (i.e., the first preset balance condition) is reached, the agent can switch to the decision-making mode, and the mathematical expression of its switching condition (i.e., the first preset balance condition) is:

[0091]

[0092] in, Represents the repulsive force of the preset obstacle on the target agent; A unit vector representing the target direction; represents gravity; a represents the number of the target agent; B represents the preset obstacle; Θ s Indicates the field of view angle.

[0093] Therefore, the embodiment of the present application effectively ensures the efficiency and reliability of the agent's conversion to the decision-making mode by setting a suitable first preset balance condition.

[0094] Optionally, in one embodiment of the present application, the mathematical expression of the second preset balance condition is:

[0095]

[0096] in, Represents the interaction force between each target agent; A unit vector representing the target direction; represents gravity; a represents the number of the target agent; i represents the number of the neighbor agent corresponding to the target agent.

[0097] The embodiment of the present application can be used when the repulsive force between the intelligent agent and its neighbors is Direction and When the balance is reached (i.e., the second preset balance condition), the agent can switch to the avoidance mode, and the mathematical expression of the switching condition (i.e., the second preset balance condition) is:

[0098]

[0099] in, Represents the interaction force between each target agent; A unit vector representing the target direction; represents gravity; a represents the number of the target agent; i represents the number of the neighbor agent corresponding to the target agent.

[0100] Therefore, the embodiment of the present application improves the accuracy of the agent's selection of the conversion mode by setting an appropriate second preset balance condition.

[0101] The following application conducts a physical verification of a multi-agent cluster crossing path planning method based on a social force model. The specific process is as follows:

[0102] This application first deploys the social force model in a multi-agent system consisting of 5 intelligent vehicles to traverse an unknown and cluttered environment. The environment size is 7m*10m, and the point cloud of the obstacle distribution is as follows Figure 5As shown, the model parameters used are:

[0103] w a =1,w i =2,l th =0.83m,Θ s =360°,α=72°,

[0104] W0=[0.5,0.5,-0.5,-0.5],w n =[-2.15m,-2.15m]

[0105] As a feasible method, the algorithm used for comparison and verification in this application is the DH-Bug algorithm, and w i The baseline algorithm is set to 0. This application will i Setting it to 0 can effectively verify the optimization effect of the interaction between neighbors on the cluster path; the long exposure trajectory and external positioning record trajectory of the experimental operation results are shown in Figure 5 As shown, the external positioning here is only used to record the motion trajectory of the intelligent agent in order to show the experimental results.

[0106] The experimental results prove that the multi-agent cluster crossing path planning method based on the social force model of this application can optimize the cluster path through the action of social forces; the average cluster path of the multi-agent cluster crossing path planning method based on the social force model of this application is 10.77m, compared with the DH-Bug baseline and w i = 0, the baseline algorithm is 11.29m and 11.33m respectively.

[0107] Specifically, Figure 5 The optimization effects of points A, B, and C in the figure are very obvious. Points A and C represent the optimization effects of attraction due to social forces, and point B represents the optimization effect of repulsion. i = 0 has trajectory interference at point D due to the lack of avoidance effect. A detour phenomenon occurs at point E due to the lack of attraction between neighbors, which proves the optimization effect of the interaction between neighbors on the cluster path. In addition, the DH-Bug baseline method also has a large number of detour phenomena, so the execution path in the statistical results is longer.

[0108] In summary, the multi-agent cluster crossing path planning method based on the social force model of the present application can enable the multi-agent cluster to spontaneously exhibit the effects of aggregation, dispersion and avoidance, thereby realizing the optimization of the crossing path of the multi-agent system.

[0109] According to the multi-agent cluster crossing path planning method based on the social force model proposed in the embodiment of the present application, by obtaining the environmental depth observation information of each target agent in multiple target agents and the neighbor posture observation information of the neighbor agents corresponding to each target agent; determining the interaction force between each target agent, the repulsion of the preset obstacle on each target agent, and the gravitational force of the preset target point on each target agent, and constructing a social force model based on the interaction force, repulsion, gravitation and preset finite state machine; based on the environmental depth observation information and the neighbor posture observation information, calculating the interaction force, repulsion and gravitation of the social force model, and performing state transition operations in the finite state machine according to the interaction force, repulsion and gravitation, obtaining the state transition result, so as to generate the target crossing path of multiple target agents through the state transition result. The computing cost and communication requirements of the present application are low, the calculation can be completed in real time, and it can be extended to large-scale clusters and deployed in small, low-power robots or complex environments.

[0110] Secondly, the multi-agent cluster crossing path planning device based on the social force model proposed in accordance with the embodiment of the present application is described with reference to the accompanying drawings.

[0111] Figure 6 It is a block diagram of a multi-agent cluster crossing path planning device based on a social force model in an embodiment of the present application.

[0112] like Figure 6 As shown, the multi-agent cluster crossing path planning device 10 based on the social force model includes: an acquisition module 100, a modeling module 200 and a generation module 300.

[0113] The acquisition module 100 is used to acquire the environment depth observation information of each target intelligent agent among multiple target intelligent agents and the neighbor posture observation information of the neighbor intelligent agent corresponding to each target intelligent agent.

[0114] The modeling module 200 is used to determine the interaction force between each target intelligent agent, the repulsive force of the preset obstacle on each target intelligent agent, and the attractive force of the preset target point on each target intelligent agent, and to construct a social force model based on the interaction force, repulsive force, attractive force and preset finite state machine.

[0115] The generation module 300 is used to calculate the interaction force, repulsion and attraction of the social force model based on the environment depth observation information and the neighbor posture observation information, and perform state transition operations in the finite state machine according to the interaction force, repulsion and attraction to obtain the state transition results, so as to generate target crossing paths for multiple target intelligent agents through the state transition results.

[0116] Optionally, in one embodiment of the present application, the modeling module 200 includes: a first computing unit, a determining unit and a second computing unit.

[0117] Among them, the first calculation unit is used to determine the target direction and preset gravity weight parameters of multiple target intelligent bodies, and calculate the left turn detour probability of multiple target intelligent bodies, so as to calculate gravity according to the left turn detour probability, target direction and preset gravity weight parameters.

[0118] A determination unit is used to calculate the distance vectors between multiple target intelligent agents and preset obstacles according to the environmental depth observation information, and determine the repulsive force based on the distance vector, the unit vector corresponding to the distance vector and the preset repulsive potential energy field.

[0119] The second calculation unit is used to calculate the actual distance between any two target intelligent agents among the multiple target intelligent agents, and determine the interaction force based on the actual distance and the preset reference distance and in combination with the environmental depth observation information of each target intelligent agent.

[0120] Optionally, in one embodiment of the present application, the generation module 300 includes: a setting unit, a first conversion unit, a second conversion unit, a third conversion unit and a fourth conversion unit.

[0121] The setting unit is used to set the initial current mode of the target intelligent body to the forward mode and determine the field of view angle of the target intelligent body.

[0122] The first conversion unit is used to convert the target intelligent body from the forward mode to the decision mode when there is a preset obstacle within the observation range of the field of view angle, and the repulsive force of the preset obstacle on the target intelligent body in the target direction and the gravitational force of the target intelligent body meet the first preset balance condition.

[0123] The second conversion unit is used to calculate the resultant force of the target intelligent agent based on the interaction force, repulsion and attraction, and determine the left turn obstacle avoidance probability of the target intelligent agent according to the resultant force and a preset obstacle avoidance decision expression, so as to convert the target intelligent agent from the decision mode to the obstacle avoidance mode through the left turn obstacle avoidance probability.

[0124] The third conversion unit is used to determine the target obstacle avoidance direction of the target intelligent body based on the left turn obstacle avoidance probability and the target direction, and control the target intelligent body to move along the target obstacle avoidance direction, and when the repulsion and attraction of the target intelligent body meet a preset offset condition, the target intelligent body is converted from the obstacle avoidance mode to the forward mode, or when the repulsion and attraction between the target intelligent body and the neighbor intelligent body corresponding to the target intelligent body in the target direction meet a second preset balance condition, the target intelligent body is converted from the obstacle avoidance mode to the avoidance mode.

[0125] The fourth conversion unit is used to control at least two target agents among the multiple target agents to convert from the avoidance mode to the forward mode in sequence based on the priority order between the at least two target agents when at least two target agents among the multiple target agents enter the avoidance mode at the same time.

[0126] Optionally, in one embodiment of the present application, the mathematical expression of the first preset balance condition is:

[0127]

[0128] in, Represents the repulsive force of the preset obstacle on the target agent; A unit vector representing the target direction; represents gravity; a represents the number of the target agent; B represents the preset obstacle; Θ s Indicates the field of view angle.

[0129] Optionally, in one embodiment of the present application, the mathematical expression of the second preset balance condition is:

[0130]

[0131] in, Represents the interaction force between each target agent; A unit vector representing the target direction; represents gravity; a represents the number of the target agent; i represents the number of the neighbor agent corresponding to the target agent.

[0132] It should be noted that the aforementioned explanation of the embodiment of the multi-agent cluster crossing path planning method based on the social force model is also applicable to the multi-agent cluster crossing path planning device based on the social force model of this embodiment, and will not be repeated here.

[0133] According to the multi-agent cluster crossing path planning device based on the social force model proposed in the embodiment of the present application, it includes an acquisition module for acquiring the environmental depth observation information of each target agent in multiple target agents and the neighbor posture observation information of the neighbor agent corresponding to each target agent; a modeling module for determining the interaction force between each target agent, the repulsion of the preset obstacle on each target agent, and the gravitational force of the preset target point on each target agent, and building a social force model based on the interaction force, repulsion, gravitation and preset finite state machine; a generation module for calculating the interaction force, repulsion and gravitation of the social force model based on the environmental depth observation information and the neighbor posture observation information, and performing state conversion operations in the finite state machine according to the interaction force, repulsion and gravitation to obtain the state conversion result, so as to generate the target crossing path of multiple target agents through the state conversion result. The computing cost and communication requirements of the present application are low, the calculation can be completed in real time, and it can be extended to large-scale clusters and deployed in small, low-power robots or complex environments.

[0134] Figure 7A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include:

[0135] A memory 701 , a processor 702 , and a computer program stored in the memory 701 and executable on the processor 702 .

[0136] When the processor 702 executes the program, the multi-agent cluster crossing path planning method based on the social force model provided in the above embodiment is implemented.

[0137] Furthermore, the electronic device further comprises:

[0138] The communication interface 703 is used for communication between the memory 701 and the processor 702 .

[0139] The memory 701 is used to store computer programs that can be executed on the processor 702 .

[0140] The memory 701 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0141] If the memory 701, the processor 702 and the communication interface 703 are implemented independently, the communication interface 703, the memory 701 and the processor 702 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0142] Optionally, in a specific implementation, if the memory 701, the processor 702 and the communication interface 703 are integrated on a chip, the memory 701, the processor 702 and the communication interface 703 can communicate with each other through an internal interface.

[0143] The processor 702 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0144] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-agent cluster crossing path planning method based on the social force model as described above.

[0145] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed, is used to implement the above-mentioned multi-agent cluster crossing path planning method based on the social force model.

[0146] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0147] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0148] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0149] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or N wirings (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways as necessary and then storing it in a computer memory.

[0150] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0151] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0152] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0153] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A multi-agent cluster crossing path planning method based on a social force model, characterized in that: The following steps are involved: Obtaining environmental depth observation information of each target intelligent agent among multiple target intelligent agents and neighbor position observation information of neighbor intelligent agents corresponding to each target intelligent agent; Determine the interaction force between each target intelligent agent, the repulsive force of a preset obstacle on each target intelligent agent, and the gravitational force of a preset target point on each target intelligent agent, and construct a social force model according to the interaction force, the repulsive force, the gravitational force, and a preset finite state machine; Based on the environment depth observation information and the neighbor posture observation information, the interaction force, the repulsion and the gravitational force of the social force model are calculated, and a state transition operation is performed in the finite state machine according to the interaction force, the repulsion and the gravitational force to obtain a state transition result, so as to generate a target crossing path of the multiple target intelligent agents through the state transition result; The determining of the interaction force between each target intelligent agent, the repulsive force of a preset obstacle on each target intelligent agent, and the gravitational force of a preset target point on each target intelligent agent, and constructing a social force model according to the interaction force, the repulsive force, the gravitational force, and a preset finite state machine, includes: Determine the target directions and preset gravity weight parameters of the multiple target intelligent bodies, and calculate the left turn detour probabilities of the multiple target intelligent bodies, so as to calculate the gravity according to the left turn detour probabilities, the target directions and the preset gravity weight parameters; Calculating distance vectors between the multiple target agents and preset obstacles according to the environmental depth observation information, and determining the repulsive force based on the distance vector, a unit vector corresponding to the distance vector, and a preset repulsive potential energy field; Calculating an actual distance between any two target agents among the plurality of target agents, and determining the interaction force based on the actual distance and a preset reference distance and in combination with environmental depth observation information of each target agent; The mathematical expression of the gravitational force of the preset target point on each target intelligent agent is: In the formula, represents the preset gravity weight parameter; is the probability of turning left when encountering an obstacle, and the mathematical expression of the probability of turning left is: In the formula, the sigmoid function ; is the weight parameter vector.

2. The method according to claim 1, characterized in that The performing a state conversion operation in the finite state machine according to the interaction force, the repulsive force and the attractive force comprises: Setting the initial current mode of the target agent to the forward mode and determining the field of view angle of the target agent; When the preset obstacle exists within the observation range of the field of view angle, and the repulsive force of the preset obstacle on the target intelligent body in the target direction and the gravitational force of the target intelligent body satisfy a first preset balance condition, converting the target intelligent body from the forward mode to the decision mode; Based on the interaction force, the repulsive force and the gravitational force, a resultant force of the target intelligent agent is calculated, and a left turn obstacle avoidance probability of the target intelligent agent is determined according to the resultant force and a preset obstacle avoidance decision expression, so as to convert the target intelligent agent from the decision mode to the obstacle avoidance mode according to the left turn obstacle avoidance probability; Based on the left turn obstacle avoidance probability and the target direction, determine the target obstacle avoidance direction of the target intelligent body, control the target intelligent body to move along the target obstacle avoidance direction, and when the repulsive force and the attractive force of the target intelligent body meet a preset offset condition, convert the target intelligent body from the obstacle avoidance mode to the forward mode, or when the repulsive force and the attractive force between the target intelligent body and the neighbor intelligent body corresponding to the target intelligent body in the target direction meet a second preset balance condition, convert the target intelligent body from the obstacle avoidance mode to the avoidance mode; When at least two of the multiple target agents enter the avoidance mode at the same time, based on the priority order between the at least two target agents, the at least two target agents are controlled to switch from the avoidance mode to the forward mode in sequence.

3. The method according to claim 2, characterized in that The mathematical expression of the first preset balance condition is: in, represents the repulsive force of the preset obstacle on the target intelligent body; a unit vector representing the direction of the target; represents said gravitational force; The number of the target agent; Indicates the preset obstacle; Represents the field of view angle.

4. The method according to claim 2, characterized in that: The mathematical expression of the second preset balance condition is: in, Represents the interaction force between each target intelligent agent; a unit vector representing the direction of the target; represents said gravitational force; The number of the target agent; Indicates the number of the neighbor agent corresponding to the target agent.

5. A multi-agent cluster crossing path planning device based on a social force model, characterized in that: include: An acquisition module, used to acquire environmental depth observation information of each target intelligent body among multiple target intelligent bodies and neighbor position observation information of neighbor intelligent bodies corresponding to each target intelligent body; A modeling module, used to determine the interaction force between each target intelligent agent, the repulsive force of a preset obstacle on each target intelligent agent, and the gravitational force of a preset target point on each target intelligent agent, and to construct a social force model according to the interaction force, the repulsive force, the gravitational force, and a preset finite state machine; A generation module, used to calculate the interaction force, the repulsion and the gravitational force of the social force model based on the environment depth observation information and the neighbor posture observation information, and perform a state conversion operation in the finite state machine according to the interaction force, the repulsion and the gravitational force to obtain a state conversion result, so as to generate a target crossing path of the multiple target intelligent agents through the state conversion result; Wherein, the modeling module includes: A first calculation unit is used to determine the target directions and preset gravity weight parameters of the multiple target agents, and calculate the left turn detour probabilities of the multiple target agents, so as to calculate the gravity according to the left turn detour probabilities, the target directions and the preset gravity weight parameters; a determination unit, configured to calculate the distance vectors between the plurality of target agents and the preset obstacles according to the environmental depth observation information, and determine the repulsive force based on the distance vectors, the unit vectors corresponding to the distance vectors, and a preset repulsive potential energy field; A second calculation unit, configured to calculate an actual distance between any two target agents among the plurality of target agents, and determine the interaction force based on the actual distance and a preset reference distance and in combination with environmental depth observation information of each target agent; The mathematical expression of the gravitational force of the preset target point on each target intelligent agent is: In the formula, represents the preset gravity weight parameter; is the probability of turning left when encountering an obstacle, and the mathematical expression of the probability of turning left is: In the formula, the sigmoid function ; is the weight parameter vector.

6. An electronic device, characterized in that: include: 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 a multi-agent cluster crossing path planning method based on a social force model as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement a multi-agent cluster crossing path planning method based on a social force model as described in any one of claims 1 to 4.

8. A computer program product, comprising a computer program, characterized in that The computer program is executed to implement the multi-agent cluster crossing path planning method based on the social force model as described in any one of claims 1-4.

Images