CN114518754B - Multi-agent escape problem modeling and trapping strategy generation method - Google Patents

Multi-agent escape problem modeling and trapping strategy generation method Download PDF

Info

Publication number
CN114518754B
CN114518754B CN202210104867.6A CN202210104867A CN114518754B CN 114518754 B CN114518754 B CN 114518754B CN 202210104867 A CN202210104867 A CN 202210104867A CN 114518754 B CN114518754 B CN 114518754B
Authority
CN
China
Prior art keywords
chaser
trapping
agent
escaper
escape
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210104867.6A
Other languages
Chinese (zh)
Other versions
CN114518754A (en
Inventor
董刚奇
邢亚红
黄攀峰
王勇杰
王梓良
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN202210104867.6A priority Critical patent/CN114518754B/en
Publication of CN114518754A publication Critical patent/CN114518754A/en
Application granted granted Critical
Publication of CN114518754B publication Critical patent/CN114518754B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0259—Control of position or course in two dimensions specially adapted to land vehicles using magnetic or electromagnetic means
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • G06N5/042—Backward inferencing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Automation & Control Theory (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Electromagnetism (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a multi-agent pursuit problem modeling and trapping strategy generation method, which aims to solve the defect that the existing trapping strategy cannot truly reflect the actual pursuit situation and solve the problem that the existing trapping strategy considering the obstacle environment is difficult to solve when the intelligent agent is large in scale. When modeling the multi-agent escape problem, the invention comprehensively considers the situations of obstacles and exits in the real scene; each chaser adjusts the trapping target point in real time according to the situation change of the escapers, and can receive the repulsive force from the obstacle at the same time of trapping, and the closer the distance to the obstacle is, the larger the repulsive force is, so that the obstacle can be avoided, the escapers can be trapped, and the method is particularly suitable for trapping tasks in real complex scenes; and the Voronoi partition is carried out on the game environment, each chasing person aims at minimizing the Voronoi unit of the escaper, few factors are required to be considered for decision making, the calculation is carried out in the low-dimensional configuration space of a single agent, and the solution is simple.

Description

Multi-agent escape problem modeling and trapping strategy generation method
Technical Field
The invention relates to a multi-agent escape problem modeling and trapping strategy generation method.
Background
The multi-agent escape-following problem refers to that in a multi-mobile robot-based escape-following complex system, the corresponding motion strategy is applied to each chaser to complete the catching of one escaped or a complex consisting of a plurality of escaped. The behavior between the two groups of the chaser and the escaped is antagonistic, and each intelligent body must be capable of knowing the dynamically changing environment in real time aiming at the continuously changing chase situation, so as to judge the current chase situation, reasonably process the real-time information and finally accurately and timely make a decision. As a typical problem of researching the countering and cooperation of multiple agents, the escape problem is a problem of real-time dynamic system cooperative game, and many key technologies thereof are applied to the industrial field and are receiving a great deal of attention.
In studying the capture of individual escapes by multiple chasers in a confined bounded area, zhengyuan Zhou et al propose a capture strategy based on minimizing the escape's generalized Voronoi cell area, simplifying the high-dimensional problem, wherein individual chasers can share state information, independently calculate their individual strategy inputs, and reduce capture time by improving cooperativity. The strategy is verified, so that the chaser can finish the trapping of the escaped in a limited time, and a new technical scheme is provided for solving the problem of the chaser-escaping game. In recent years, the application scenes of the multi-agent escape-following problem are more and more, and the unmanned aerial vehicle is used for resisting escape to a space vehicle and a spacecraft, so that the existing escape-following algorithm is free from higher requirements, and has better obstacle avoidance performance, high expansibility and flexibility and is more close to the actual environment (with obstacles and exits). For the game problem in the obstacle environment, the prior art mostly combines the target allocation algorithm with the classical differential game algorithm, and searches the optimal track by integrating the terminal condition backwards according to the set performance index function, so as to obtain the optimal trapping strategy of the chaser. Each chasing person needs to know the position information, decision output information, environment information and the like of other agents, and the decision inputs of the chasing persons are mutually coupled.
Disclosure of Invention
The invention provides a multi-agent pursuit problem modeling and trapping strategy generation method, which aims to overcome the defects that the existing trapping strategy does not consider the actual environment with obstacles and exits and cannot truly reflect the actual pursuit situation, and solves the technical problem that the existing trapping strategy considering the environments with obstacles is difficult to solve when the intelligent agent scale is large. The invention expands the existing escape problem to be more close to the actual scene, and considers the situation of existence of an outlet and an obstacle, so that the escape algorithm can be applied to the finer scene, and the actual escape situation can be reflected more truly. The trapping strategy generation method provided can realize the trapping task of the chaser in the real environment.
The technical scheme of the invention is as follows:
the multi-agent escape problem modeling and trapping strategy generation method is characterized by comprising the following steps of:
Step 1: modeling of multi-agent escape problems
Step 1.1: building gaming environments
1.1.1 Defining a bounded non-occluded area Ω
Defining a certain bounded non-closed area omega, wherein n exp outlets are formed in the boundary of the area omega, n bar static barriers are arranged in the area omega, any point in the area omega is taken as a coordinate origin, the horizontal right direction is taken as the positive x-axis direction, the direction vertical to the upper x-axis direction is taken as the positive y-axis direction, and a global coordinate system xOy is established;
The location of the outlets on the boundary of zone Ω Position/>, of each stationary obstacle within region ΩThe width of each outlet on the boundary of the region omega is recorded as { D k|k＝1,···,nexp }, the area of each static obstacle in the region omega is recorded as { S w|w＝1,···,nbar }, the influence radius of each static obstacle is recorded as { rho w|w＝1,···,nbar }, the influence range of each static obstacle is a circle field taking the center of the obstacle as a round point and the influence radius rho w as a radius, the influence radius rho w is artificially set, the value of the influence radius rho w is enough to ensure that each circle field can completely cover the obstacle, the distance from any point on the boundary of the obstacle to the boundary of the circle field is larger than the safety distance r s, the safety distance r s is set according to actual requirements, and meanwhile, the positions of each outlet are not in the influence range of each static obstacle.
1.1.2 Defining parameters of the respective Agents
Defining a plurality of agents, dividing the agents into two types of chasers and escapers, setting the chasers P= { P i |i=1, how, N, evasion e= { E j |j=1, how, M, i.e., the number of chasers is N, the number of escapers is M, the position x p∈Ω,xe epsilon omega of each intelligent agent prescribes that the distance from the initial position of each escaper to any outlet is larger than the escape distance r e, and the difficulty degree of escaping of the escaper can be changed by adjusting the value of the escape distance r e; meanwhile, it is assumed that each agent in the pursuit process has completely solved the position information of the exit and the stationary obstacle in the non-closed area Ω and the position information of each agent, i.e. the process is a game under complete information. The equation of motion of each agent is shown in formula (1):
in the method, in the process of the invention, The initial positions of the chaser and the escaped, u i,uj is the speed control input of the chaser and the escaped, respectively, which have the limitation of the maximum movement velocity v p,max,ve,max and v p,max≥ve,max;
step 1.2: setting a decision mode
The following is provided for the decision of each state in the chase-and-flee game:
When the distance d ij of a certain escaper from any chaser is smaller than the capturing distance d min or the escaper collides with the boundary of the region omega, capturing the escaper successfully; d min is set according to actual requirements;
When a certain escaper reaches any one of the exits of the region omega or passes through the certain exit, the escaper is considered to escape successfully;
if each escaper in the region omega is captured successfully or has escaped successfully, the method is regarded as the end of the chase game;
Step 1.3: setting an evasion strategy
In order to ensure the universality of the trapping strategy of the chaser, namely, no matter how the escaped moves, the trapping strategy of the invention can capture the escaped, so the invention does not make special requirements on the movement of the escaped, and only makes the following regulations:
1) An escaper can identify and avoid obstacles;
2) The escapers should escape the trapping of the chaser as much as possible;
3) On the basis of achieving the two requirements, the escaper should move towards the exit as much as possible to achieve escape;
Step 2: generating a trapping strategy
Step 2.1: distribution of the acquisition tasks
The position coordinates of each agent in the global coordinate system xOy in the area are used as the mother points of the Voronoi graph, the Voronoi units of each agent are generated, and for a certain chaser p i,
If the adjacent Voronoi units have the escapers, the escapers closest to the chaser are the trapping targets;
If no escapers exist in the adjacent Voronoi units, the chaser p i should take the escapers closest to the chaser p i in the region omega as the trapping target;
the individual chasers in the region Ω thus obtained can accordingly capture the target.
Step 2.2: determining a point of enclosure
Calculating the interception coefficient f ij of the chaser p i on the trapping target e j,
When f ij is more than or equal to 0, the target point of the chaser p i is the position of the nearest exit from the trapping target e j;
when f ij <0, the chaser p i target point is determined by the method of Voronoi partitioning;
step 2.3: determining the direction and rate of travel of a chaser
Step 2.4: escape-chasing game
Each chaser moves a time unit in the advancing direction to obtain the position coordinate of the next moment, and returns to the step 2.1 until the end of the chase game is judged according to the judgment mode of the step 1.2.
Further, in the step 2.2, the interception coefficient of the chaser p i to the trapping target e j is calculated according to the following method
Where k is the number of the nearest exit from the current trapping object e j,For the distance of the chaser p i to the kth exit,/>Is the distance from the current trapping object e j to the kth exit.
Further, the method for determining the target point by the Voronoi partition method in the step 2.2 specifically includes: if the pursuer p i has a boundary with the Voronoi cell of the surrounding target e j, the target point of the pursuer p i is the midpoint of the boundary between the two Voronoi cells; if the pursuer p i does not have a boundary with the Voronoi cell of the target e j, the target point of the pursuer p i is the location of the target p j.
Further, the method for determining the travel direction of the chaser in the step 2.3 includes: the resultant force of the attractive force and the repulsive force applied by the chaser p i is calculated, and the resultant force direction is the advancing direction of the chaser p i.
Further, the method for determining the travel direction of the chaser in the step 2.3 specifically includes:
2.3.1 calculating the attractive force from the target point to which the chaser p i is subjected
Fatt(pi)＝ξρ(pi,qgoal) (3)
Where ζ is the gravitational gain coefficient, ρ (p i,qgoal) is the distance between the chaser p i and its target point, and the gravitational direction is directed to the target point from the position where the chaser p i is located.
2.3.2 Calculating the repulsive force from the w-th obstacle received by the chaser p i
Wherein eta is the repulsive force gain coefficient, rho w is the influence radius of the w-th obstacle,The direction of the repulsive force is directed from the position of the w-th obstacle to the chaser p i for the distance between the chaser p i and the w-th obstacle.
2.3.3 Calculating the resultant force of the attractive force and the repulsive force of the chaser p i
Wherein n bar is the number of stationary obstacles, the attraction force and the repulsion force are vector superposition, and the direction of resultant force F (p i) is the travelling direction of the chaser p i.
Further, in step 2.3, each chaser is set to travel at the maximum movement rate, i.e.
The invention has the beneficial effects that:
1. When the multi-agent escape-following problem is modeled, the situations of obstacles and exits in a real scene are comprehensively considered, and compared with the traditional escape-following problem model, the method is closer to reality, and can be used for applying the research on the escape-following algorithm to more refined scenes.
2. According to the method for generating the trapping strategy, each chaser can autonomously determine the trapping task, and the self-programming is dynamically adjusted according to the change of the position information of each agent in the game process, so that the synergy among the chasers is improved, the completion of the whole task is further accelerated, and the method is particularly suitable for the trapping task of multiple chasers to multiple escapers.
3. According to the trapping strategy generation method provided by the invention, each chaser adjusts the trapping target point in real time according to the situation change of the escapers, the escapers are trapped and can receive repulsive force from the obstacle, the closer the escapers are to the obstacle, the larger the repulsive force received is, so that the method can be used for avoiding the obstacle and trapping the escapers, and is particularly suitable for trapping tasks in real complex scenes.
4. According to the method for generating the trapping strategy, the Voronoi partition is carried out on the game environment, each chaser aims at minimizing the Voronoi unit of the escaper, the factors required to be considered for decision are few, a certain chaser only needs to know the position information of each agent and the obstacle and the outlet in the environment, and the required motion strategy can be obtained, so that the calculation is carried out only in the low-dimensional configuration space of a single agent, and is not carried out in the high-dimensional joint state space of all agents, and therefore the solution is simple.
Drawings
FIG. 1 is a flow chart of a modeling and trapping strategy generation method of the present invention.
FIG. 2 shows a multi-agent chase-escaping game process of the present invention.
FIG. 3 is a diagram of a multi-agent chase-escaping game process II according to the present invention.
FIG. 4 is a diagram of a multi-agent chase-escaping game process three according to the present invention.
Fig. 5 is a graph showing the minimum distance change from each escaper to the chaser according to the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The invention provides a multi-agent escape problem modeling and trapping strategy generation method, wherein the modeling method comprises the following steps: constructing a game environment, setting a judgment mode and setting an escape strategy; the method for generating the trapping strategy comprises the following steps: the method comprises the steps of distributing the trapping tasks, determining trapping target points, determining the advancing direction and advancing speed of a chaser, gaming and judging whether the gaming is finished.
Step 1: modeling of multi-agent escape problems
Step 1.1: building gaming environments
Given a square region Ω (in other embodiments, other shaped regions, such as a circular, polygonal, or irregularly shaped region, are possible, and the method steps involved in the following are not changed), the side length is l=3 km, the vertex below and to the left of the region Ω is taken as the origin of coordinates, the horizontal right direction is taken as the positive x-axis direction, the direction in the vertical x-axis direction is taken as the positive y-axis direction, and the global coordinate system is established as xOy. The boundary of the region omega is provided with 4 outlets, 7 static barriers are arranged in the region omega, the positions of the outlets and the static barriers in the region omega are shown in figure 2, five stars on the boundary in figure 2 represent the outlets, black filling regions inside the boundary represent the static barriers, and the area surrounded by a dotted circle around the static barriers is the influence range of the static barriers. The width { D k =0.05km|k=1, ··4} of each exit, the area {S1＝S2＝0.08km2,S3＝S4＝S5＝S6＝S7＝0.02km2}, safety distance r s =0.15 km of each obstacle, the radius of influence { ρ w =0.35 km|w=1, ··7} of each stationary obstacle, and the exit positions are not within the range of influence of the obstacle.
Setting the number of chasers N=4, the number of escapers M=3, i.e. the chasers P= { P i |i=1, & gtof, 4, escapement e= { E j |j=1, ····3, the location of each agent is shown in figure 2, in the figure, X1-X4 represent each chaser, X5-X7 represent each escaper, the motion trail of each intelligent body is marked in the figure, and the distance from the initial position of each escaper to any exit is larger than the escape distance r e =0.2 km. Meanwhile, each intelligent agent completely knows the position information of the outlet on the boundary of the non-closed area omega, the static obstacle in the area omega and the position information of each intelligent agent under the global coordinate system xOy. The equation of motion of each agent is as follows:
in the method, in the process of the invention, The initial positions of the chaser and the escaped, u i,uj is the speed control input of the chaser and the escaped, respectively, which are both limited by the maximum movement velocity v p,max,ve,max, in this embodiment the maximum movement velocity of the chaser and the escaped is v p,max＝0.02km/s,ve,max =0.02 km/s, respectively.
Step 1.2: setting a decision mode
When the distance d ij between a certain escaper and any chaser is smaller than the capturing distance d min =0.04 km, or the escaper collides with the boundary, capturing the escaper successfully;
an escaper is considered to have escaped successfully when it reaches or passes through any one of the exits.
If each escaped player in region Ω is either successfully captured or has successfully escaped, then the chase game is deemed to have ended.
Step 1.3: setting an evasion strategy
The escape strategy meets the following three-point requirement:
1) An escaper can identify and avoid obstacles;
2) The escapers should escape the trapping of the chaser as much as possible;
3) On the basis of achieving the two requirements, the escaper should move towards the exit as much as possible to achieve escape;
The embodiment combines the artificial potential field method to set the escape strategy, and the specific method is as follows:
Step 1.3.1 determining the escape target point
When a certain escapement e j is not surrounded by a chaser, that is, when there is no chaser on one side of the escapement and there is an exit, the escapement should select the exit closest to the chaser as the target point in the direction of no chaser;
When the escapement e j is surrounded by the chaser, the escapement should move in a direction away from the chaser nearest to itself, and the target point at this moment is a point in the direction to which the length of the escapement is equal to the distance between itself and the chaser nearest to the escapement.
Step 1.3.2 determining the direction and rate of travel of the escaper
The direction of travel of the escaper is determined using the following method:
in this embodiment, an escaper e j receives attraction from the target point
Fatt(ej)＝ξρ(ej,qgoal)
Where the gravitational gain coefficient ζ=0.7, ρ (e j,qgoal) is the distance between the escapement e j and its target point, the direction of the gravitational force being directed to the target point by the position where the escapement e j is located.
The repulsive force from the w-th obstacle is received by an escapement e j
Wherein the repulsive force gain coefficient eta=0.3, ρ w is the influence radius of the w-th obstacle,For the distance between the escapement e j and the w-th obstacle, the direction of the repulsive force is directed from the position of the w-th obstacle to the escapement e j.
The escapement e j receives the resultant force of attraction and repulsion
Where n bar =7, each evade can obtain the direction of the resultant force F (e j) by the above equation, and thus can determine the traveling direction thereof.
Setting the rate of travel of the escaper:
Each evacuee was set to travel at the maximum rate of movement, i.e. v e＝ve,max = 0.02km/s.
Step 2: generating a trapping strategy
Step 2.1: distribution of the acquisition tasks
The position coordinates of all the intelligent agents in the region omega are used as the mother points of the Voronoi graph, voronoi units of all the intelligent agents are generated, and if a certain chaser p i exists in the adjacent Voronoi units, the nearest escaper in the vicinity of the chaser p is the trapping target; if there is no escapement in its adjacent Voronoi cell, the chaser p i should target the escapement nearest to itself in the region Ω. The individual chasers in the region Ω thus obtained can accordingly capture the target. As shown in fig. 2, 3, and 4, the chaser X1, X2, and X4 take the escaped person X7 as a trapping target, and the chaser X3 takes the escaped person X5 as a trapping target, as an example in fig. 2.
Step 2.2: determining the target point of the trap
Each chaser calculates the interception coefficient f ij for the captured object by,
Where k is the number of the nearest exit from the current trapping object e j,For the distance of the chaser p i to the kth exit,/>Is the distance from the current trapping object e j to the kth exit.
When the interception coefficient f ij of a certain chaser p i is more than or equal to 0, the target point of the chaser p i should be the nearest exit coordinate from the trapping target e j;
when the interception coefficient f ij of a certain chaser p i is less than 0, the target point of the chaser p i should be determined by the Voronoi partition method, specifically: if the pursuer p i has a boundary with the Voronoi cell of the surrounding target e j, the target point of the pursuer p i is the midpoint of the boundary between the two Voronoi cells; if the pursuer p i does not have a boundary with the Voronoi cell of the target e j, the target point of the pursuer p i is the location of the target e j.
Taking fig. 3 as an example, each chaser is connected with its target point by a straight line, the chaser X1 has a boundary between the Voronoi units of the chaser X1 and the target point of the chaser X5, and the interception coefficient f 15 = -11.3s <0, so that the target point of the chaser X1 is the midpoint of the boundary between the Voronoi units and the Voronoi units of the chaser X5, as indicated in the figure; the interception coefficient f 26 = -10.1s <0 of the pursuer X2 on the captured object X6, and the Voronoi units of the two units are intersected, so that the target point of the pursuer X2 is the midpoint of the intersection of the Voronoi unit and the Voronoi unit of the captured object X6, as indicated in the figure; the interception coefficient f 35 =44.5s >0 of the chaser X3 to the trapping target X5, and the outlet closest to the trapping target X5 is the lower outlet, so the target point of the chaser X3 is the lower outlet; the interception coefficient f 46 =0.5s >0 of the chaser X4 to the trapping object X6, and the exit nearest to the trapping object X6 is the left exit, so the target point of the chaser X4 is the left exit.
Step 2.3: determining the direction and rate of travel of a chaser
The following method is adopted to determine the travel direction of the chaser:
the attractive force from the target point received by a chaser p i can be obtained by the following calculation
Fatt(pi)＝ξρ(pi,qgoal)
In the formula, the gravitational gain coefficient ζ=0.7 is taken, and ρ (p i,qgoal) is the distance between the chaser p i and the target point, and the gravitational direction is directed to the target point from the position of the chaser p i.
The chaser p i receives repulsive force from the w-th obstacle
Wherein, the repulsive force gain coefficient eta=0.3,For the distance between the chaser p i and the w-th obstacle, ρ w is the radius of influence of the w-th obstacle, and the direction of the repulsive force is directed from the position of the w-th obstacle to the chaser p i.
The chaser p i receives the resultant force of attraction and repulsion
Where n bar =7, each chaser can obtain the direction of the resultant force F (p i) by the above formula, and thus can determine the traveling direction thereof.
Setting the travel rate of the chaser:
Each chaser travels at the maximum movement rate during the game, i.e
Step 2.4: each chaser moves a time unit in the advancing direction to obtain the position coordinate of the next moment, and returns to the step 2.1 until the chaser game is finished, and whether the chaser game is finished is judged according to the judgment mode set in the step 1.2.
As can be seen from the motion trail of each chaser in fig. 4, each chaser does not collide with the obstacle in the process of this chaser game, so as to meet the requirement of obstacle avoidance; the escapers can exit the game as long as the capturing conditions are met, and the last section of each curve in fig. 5 shows that the minimum distance from each escaper to the chaser is smaller than the capturing distance d min =0.04 km before the end of the chaser game, so that the capturing conditions set in the step 1.2 can be met, and each chaser can complete the capturing of the escaper under the established chaser game model, which shows that the method provided by the invention can be suitable for capturing tasks in real environments.

Claims (6)

1. A multi-agent escape problem modeling and trapping strategy generation method is characterized by comprising the following steps:
Step 1: modeling of multi-agent escape problems
Step 1.1: building gaming environments
1.1.1 Defining a bounded non-occluded area Ω
Defining a region omega, wherein n exp outlets are formed in the boundary of the region omega, n bar static barriers are arranged in the region omega, and the positions of the outlets are not in the influence range of the static barriers; n exp≥1,nbar is more than or equal to 1;
taking any point in the region omega or on the boundary as a coordinate origin, taking the horizontal rightward direction as the positive direction of the x axis, taking the vertical x axis as the positive direction of the y axis, and establishing a global coordinate system as xOy;
1.1.2 defining parameters of the respective Agents
Defining a plurality of intelligent agents, wherein each intelligent agent is in a region omega or on a boundary, and each intelligent agent knows the position information of each outlet, a static obstacle and each intelligent agent under a global coordinate system xOy in the process of pursuing;
Dividing the intelligent agent into two types of chasers and escapers, wherein the number of the chasers is N, the number of the escapers is M, and the distance from the initial position of each escaper to any exit is larger than the escape distance r e;N≥1,M≥1,re and is set according to actual requirements;
The maximum movement rate of the chaser is greater than or equal to the maximum movement rate of the escaped;
step 1.2: setting a decision mode
The judgment mode is as follows:
When the distance d ij of a certain escaper from any chaser is smaller than the capturing distance d min or the escaper collides with the boundary of the region omega, capturing the escaper successfully; d min is set according to actual requirements;
When a certain escaper reaches or passes through any one of the exits of the region omega, the escaper is considered to escape successfully;
if each escaper in the region omega is captured successfully or has escaped successfully, the method is regarded as the end of the chase game;
Step 1.3: setting an evasion strategy
The escape strategy is as follows:
1) An escaper can identify and avoid obstacles;
2) The escapers should escape the trapping of the chaser as much as possible;
3) On the basis of achieving the two requirements, the escaper should move towards the exit as much as possible to achieve escape;
Step 2: generating a trapping strategy
Step 2.1: distribution of the acquisition tasks
The position coordinates of each agent in the global coordinate system xOy in the area are used as the mother points of the Voronoi graph, the Voronoi units of each agent are generated, and for a certain chaser p i,
If the adjacent Voronoi units have the escapers, the escapers closest to the chaser are the trapping targets;
if no escapers exist in the adjacent Voronoi units, the chaser p i takes the escapers closest to the chaser p i in the region omega as a trapping target;
Step 2.2: determining a point of enclosure
Calculating the interception coefficient f ij of the chaser p i on the trapping target e j,
When f ij is more than or equal to 0, the target point of the chaser p i is the position of the nearest exit from the trapping target e j;
when f ij is less than 0, determining a target point of the chaser p i by a Voronoi partition method;
step 2.3: determining the direction and rate of travel of a chaser
Step 2.4: escape-chasing game
Each chaser moves a time unit in the advancing direction to obtain the position coordinate of the next moment, and returns to the step 2.1 until the end of the chase game is judged according to the judgment mode of the step 1.2.
2. The multi-agent escape problem modeling and trapping strategy generation method according to claim 1, wherein: in step 2.2, the interception coefficient of the chaser p i to the trapping target e j is calculated according to the following method
Where k is the number of the nearest exit from the current trapping object e j,For the distance of the chaser p i to the kth exit,For the distance from the current trapping object e j to the kth exit, v p,max is the maximum rate of motion of the chaser and v e,max is the maximum rate of motion of the escaper.
3. The multi-agent escape problem modeling and trapping strategy generation method according to claim 2, wherein: the method for determining the target point by the Voronoi partition method in the step 2.2 specifically comprises the following steps: if the pursuer p i has a boundary with the Voronoi cell of the surrounding target e j, the target point of the pursuer p i is the midpoint of the boundary between the two Voronoi cells; if the pursuer p i does not have a boundary with the Voronoi cell of the target e j, the target point of the pursuer p i is the location of the target e j.
4. The multi-agent escape problem modeling and trapping strategy generation method according to claim 1, wherein: the method for determining the travel direction of the chaser in the step 2.3 comprises the following steps: the resultant force of the attractive force and the repulsive force applied by the chaser p i is calculated, and the resultant force direction is the advancing direction of the chaser p i.
5. The multi-agent escape problem modeling and trapping strategy generation method according to claim 4, wherein:
The method for determining the travel direction of the chaser in the step 2.3 specifically comprises the following steps:
2.3.1 calculating the attractive force from the target point to which the chaser p i is subjected
Fatt(pi)＝ξρ(pi,qgoal)
Where ζ is the gravitational gain coefficient, ρ (p i,qgoal) is the distance between the chaser p i and its target point, and the gravitational direction is directed to the target point from the position where the chaser p i is located;
2.3.2 calculating the repulsive force from the w-th obstacle received by the chaser p i
Wherein eta is the repulsive force gain coefficient, rho w is the influence radius of the w-th obstacle,For the distance between the chaser p i and the w-th obstacle, the direction of the repulsive force is directed from the position of the w-th obstacle to the chaser p i;
2.3.3 calculating the resultant force of the attractive force and the repulsive force of the chaser p i
Wherein n bar is the number of stationary obstacles, the attraction force and the repulsion force are vector superposition, and the direction of resultant force F (p i) is the travelling direction of the chaser p i.
6. The multi-agent escape problem modeling and trapping strategy generation method according to claim 1, wherein: in step 2.3, each chaser is set to travel at the maximum movement rate.
CN202210104867.6A 2022-01-28 2022-01-28 Multi-agent escape problem modeling and trapping strategy generation method Active CN114518754B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210104867.6A CN114518754B (en) 2022-01-28 2022-01-28 Multi-agent escape problem modeling and trapping strategy generation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210104867.6A CN114518754B (en) 2022-01-28 2022-01-28 Multi-agent escape problem modeling and trapping strategy generation method

Publications (2)

Publication Number Publication Date
CN114518754A CN114518754A (en) 2022-05-20
CN114518754B true CN114518754B (en) 2024-04-23

Family

ID=81596521

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210104867.6A Active CN114518754B (en) 2022-01-28 2022-01-28 Multi-agent escape problem modeling and trapping strategy generation method

Country Status (1)

Country Link
CN (1) CN114518754B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116430865A (en) * 2023-04-17 2023-07-14 北方工业大学 Multi-machine collaborative trapping method under uncertain probability framework

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109085754A (en) * 2018-07-25 2018-12-25 西北工业大学 A kind of spacecraft neural network based is pursued and captured an escaped prisoner game method
WO2019194628A1 (en) * 2018-04-06 2019-10-10 엘지전자 주식회사 Mobile robot and control method for same
CN113552872A (en) * 2021-01-28 2021-10-26 北京理工大学 Pursuit and escape game decision method for chaser at different speeds

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6917855B2 (en) * 2002-05-10 2005-07-12 Honda Motor Co., Ltd. Real-time target tracking of an unpredictable target amid unknown obstacles

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019194628A1 (en) * 2018-04-06 2019-10-10 엘지전자 주식회사 Mobile robot and control method for same
CN109085754A (en) * 2018-07-25 2018-12-25 西北工业大学 A kind of spacecraft neural network based is pursued and captured an escaped prisoner game method
CN113552872A (en) * 2021-01-28 2021-10-26 北京理工大学 Pursuit and escape game decision method for chaser at different speeds

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
关于多机器人围捕协调路径策略研究;张旭;贾磊磊;陈群;;计算机仿真;20160615(第06期);362-366 *

Also Published As

Publication number Publication date
CN114518754A (en) 2022-05-20

Similar Documents

Publication Publication Date Title
CN108549407B (en) Control algorithm for multi-unmanned aerial vehicle cooperative formation obstacle avoidance
CN113589842B (en) Unmanned cluster task cooperation method based on multi-agent reinforcement learning
CN111880563B (en) Multi-unmanned aerial vehicle task decision method based on MADDPG
CN108827312B (en) Cooperative game path planning method based on neural network and artificial potential field
CN108680063B (en) A kind of decision-making technique for extensive unmanned plane cluster dynamic confrontation
CN108196583B (en) Unmanned aerial vehicle cluster control method
CN109144102A (en) A kind of Path Planning for UAV based on improvement bat algorithm
WO2007033101A2 (en) Hybrid control device
CN114518754B (en) Multi-agent escape problem modeling and trapping strategy generation method
CN113655810B (en) Unmanned aerial vehicle obstacle avoidance method and system based on speed potential field
CN112666976B (en) Consistency-based multi-unmanned aerial vehicle cluster collision avoidance method
CN113900449B (en) Multi-unmanned aerial vehicle track planning method and device, unmanned aerial vehicle and storage medium
CN113741525A (en) Strategy set based MADDPG multi-unmanned aerial vehicle cooperative attack and defense countermeasure method
CN113359756B (en) Method for realizing real-time planning of obstacle avoidance path of omnidirectional mobile robot based on grid method
CN114779828B (en) Unmanned aerial vehicle cluster topological control and intelligent anti-collision method based on heterogeneous formation datum points
CN106716282B (en) Method, control device and the control equipment that a kind of pair of target is controlled
CN112651486A (en) Method for improving convergence rate of MADDPG algorithm and application thereof
Sun et al. Cooperative strategy for pursuit-evasion problem in the presence of static and dynamic obstacles
CN113759935B (en) Intelligent group formation mobile control method based on fuzzy logic
CN114442628A (en) Mobile robot path planning method, device and system based on artificial potential field method
Li et al. Coordinated multi-robot target hunting based on extended cooperative game
CN113485454A (en) Unmanned aerial vehicle obstacle avoidance maneuvering method and system for geometrically improved artificial potential field
CN115061495B (en) Unmanned aerial vehicle group countermeasure autonomous control method based on eagle pigeon game
CN116203989A (en) Multi-unmanned aerial vehicle cooperative target searching method and system based on particle swarm optimization
CN115454136B (en) Unmanned aerial vehicle cluster cooperative attack and defense countermeasure decision-making method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant