A target source localization and obstacle avoidance method based on tabu search and particle swarm optimization

Abstract

The present invention relates to a target source positioning and obstacle avoidance method based on taboo search and particle swarm algorithm. The method comprises: establishing a fitness function model that characterizes the target source signal strength, treating an intelligent agent as a particle, initializing the particle swarm algorithm parameters, and clearing the taboo list; determining whether to update the taboo list based on the weighted average velocity of the particles, and combining taboo search to perform movement and obstacle avoidance; updating the fitness value and global and local optimal positions, and executing speed and position update strategies; and iterating repeatedly until the intelligent agent locates the target source or an iteration termination condition is met, ultimately locating the optimal target source. Compared with existing technologies, this invention solves the problem of intelligent agents becoming trapped in obstacles and unable to escape without prior knowledge of the environment.

Description

Technical Field

[0001] The present invention relates to the field of target signal source search, and in particular to a target source positioning and obstacle avoidance method based on taboo search and particle swarm algorithm. Background Art

[0002] The task of locating the source of a signal by measuring signal strength within a search area is known as the target signal source localization problem. The signal source localization problem has attracted widespread attention due to its various potential applications. This problem requires multiple agents to self-organize and search for the location of the signal source, i.e., the point of maximum signal strength, in an unknown environment, subject to certain constraints. The agents possess certain perceptual capabilities, capable of detecting the signal strength at their location and locating the source through group collaboration. This problem arises in a variety of scenarios, such as environmental monitoring, disaster search and rescue operations, and chemical leak investigations. Because such tasks are likely to occur in hazardous environments and pose a threat to human life, it is imperative to develop an unmanned, autonomous, multi-agent collaborative search method.

[0003] In recent years, numerous researchers have conducted a series of studies on the source localization problem, proposing methods such as gradient descent / ascent, mechanism-based methods, and biomimetic behavior. For example, a multi-agent system can achieve a better gradient estimate through collaboration. Bee swarm algorithms can be used to guide agents to gather around a signal source. Particle swarm optimizers (PSOs) are widely used to solve source localization problems due to their excellent performance and compatibility with intelligent agents.

[0004] Most obstacle avoidance methods perform well when obstacles in unknown environments are convex or known. However, if concave obstacles appear in the environment, or if prior knowledge of the obstacles is unknown, such as U-shaped obstacles, the agent may become trapped and unable to escape.

[0005] Therefore, there is an urgent need to invent a collaborative and efficient obstacle avoidance algorithm that does not require prior information about the environment. Summary of the Invention

[0006] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a target source positioning and obstacle avoidance method based on taboo search and particle swarm algorithm without the need for prior information of the environment.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] The present invention provides a target source positioning and obstacle avoidance method based on tabu search and particle swarm algorithm, the method comprising:

[0009] Establish a fitness function model that characterizes the signal strength of the target source, treat the agent as a particle, initialize the particle swarm algorithm parameters and clear the taboo list;

[0010] Based on the weighted average velocity of the particle, determine whether the particle falls into the taboo area and update the taboo list, and combine the taboo search to move and avoid obstacles;

[0011] Update the fitness value and global and local optimal positions, and execute speed and position update strategies;

[0012] The iteration is repeated until the agent locates the target source or the iteration termination condition is met, thereby enabling the agent to locate the optimal target source.

[0013] Preferably, the initialization particle swarm parameters are specifically:

[0014] Initialize the population size N, acceleration factors c1 and c2, problem dimension D, number of iterations T and clear the taboo list; randomly initialize the particle speed v i =(v i1 ,...,v ij ,...,v iD ) and position x i =(x i1 ,...,x ij ,...,x iD ), where i is the i-th particle.

[0015] Preferably, the updating formulas for the local optimal position and the global optimal position are as follows:

[0016]

[0017]

[0018] Among them, p i =(p i1 ,...,p id ,...,p iD ) represents the optimal position of the i-th particle, g=(g1,g2,...,g D ) represents the global optimal position determined by the particle swarm, and k represents the current number of iterations.

[0019] Preferably, the speed and position update strategy is as follows:

[0020]

[0021]

[0022] in, is the velocity of the i-th particle in the k-th iteration, p i =(p i1 ,...,p id ,...,piD ) represents the optimal position of the i-th particle, g=(g1,g2,...,g D ) represents the global optimal position determined by the particle swarm, k represents the current number of iterations; c1 and c2 are acceleration factors used to determine the optimal position p of the i-th particle. i and the relative importance of the global optimal position g determined by the particle swarm; ∈1 and ∈2 are two random numbers uniformly distributed in the interval [0,1]; ω is the adjustment weight of the particle search for the target source area.

[0023] Preferably, the weighted average velocity of the particle is used to determine whether the particle falls into the taboo area and to update the taboo list, and the taboo search is combined to move and avoid obstacles. The specific process is as follows:

[0024] If the weighted average velocity of the particle in k iterations If the value is less than the preset threshold γ and is not close to its global optimal position, the particle is considered to be trapped near an obstacle. For trapped particles, the area near the current trapped particle is added to the taboo list as a new taboo area and stored. All particles in the taboo area are given a velocity along the line connecting the particle and the center of the taboo area to make them leave the taboo area.

[0025] Preferably, the weighted average speed expression is:

[0026]

[0027]

[0028] in, is the weighted average velocity of the i-th particle in the k-th iteration, is the velocity of the i-th particle in the k+1-th iteration; α is the weight coefficient, representing the time state of the current iteration.

[0029] Preferably, the speed expression of the particle leaving the taboo area is:

[0030]

[0031] in, is the position of the i-th particle in the k-th iteration, c is the center of the taboo area, V max Represents the velocity limit of the particle.

[0032] Preferably, the taboo area is centered at the midpoint between the particle's previous iteration position and the current iteration position, with a side length of Rectangular area, where V max Represents the velocity limit of the particle.

[0033] Preferably, the preset threshold γ satisfies:

[0034] If the agent is trapped near an obstacle, it will fall into a deadlock state and can only move between two points. As the number of iterations increases, the proportion of the weighted average speed of the most recent iteration also increases; when the proportion of the weighted average speed increases to a preset value, the upper bound of the weighted average expectation is obtained, which is the reference value of the preset threshold γ.

[0035] Preferably, the taboo list is stored using R-Tree.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] 1) This paper proposes a target source localization and obstacle avoidance method based on tabu search and particle swarm optimization, which solves the problem of an intelligent agent being trapped in a U-shaped obstacle and unable to escape without prior knowledge of the environment;

[0038] 2) This invention uses tabu search to organize multiple agents to move and avoid obstacles, improving the ability of agents to escape from U-shaped obstacle areas while ensuring search efficiency;

[0039] 3) This invention innovatively sets taboo areas as taboo objects and uses weighted average speed as the criterion for determining whether an object has fallen into a local obstacle area, thus avoiding repeated exploration of taboo areas. It also uses R-Tree to store taboo areas to speed up the taboo list search efficiency, improve the agent's ability to avoid U-shaped obstacles, and enhance the obstacle avoidance efficiency of the particle swarm algorithm.

[0040] 4) The particle swarm algorithm of the method of the present invention is applied to the target source positioning problem, and combined with the tabu search to make up for the shortcomings of most current optimization algorithms in terms of poor obstacle avoidance capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of the method of the present invention;

[0042] Figure 2 Schematic diagram of the taboo area in the present invention;

[0043] Figure 3 This is a comparison chart of the convergence of the single-source algorithm with and without the taboo search strategy; Figure 3 a is the convergence comparison diagram of A-RPSO algorithm with and without taboo search strategy, Figure 3 b is the convergence comparison diagram of RbRDPSO algorithm with and without taboo search strategy;

[0044] Figure 4 This is a comparison chart of the convergence of the multi-source algorithm with and without the taboo search strategy; Figure 4 a is the convergence comparison diagram of SAP-CFPSO algorithm with and without taboo search strategy, Figure 4b is the convergence comparison diagram of the Multi-MPSO algorithm with and without the taboo search strategy. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0046] The present invention provides a target source positioning and obstacle avoidance method based on taboo search and particle swarm algorithm without the need for prior information about the environment. This method is based on the assumption that the agent has no prior information about the environment in a limited search range, and applies the particle swarm optimization algorithm to the multi-agent system. Each agent is regarded as a particle, and the learning strategy of the particle swarm algorithm is used to guide the evolution of the multi-agent. In the multi-agent system, each agent can detect the signal strength of its position. The current position of the particle in the particle swarm algorithm is the coordinate of each agent; the fitness value of the particle in the intensity particle swarm algorithm corresponds to the signal detected by the agent. The update of the particle position guides the moving direction and distance of the agent. Figure 1 As shown, the embodiment adopted by the present invention includes the following steps:

[0047] Establish the fitness function model f(x), initialize the particle swarm parameters: initial population size N, acceleration factors c1 and c2, problem dimension D, number of iterations T and clear the taboo list and R-Tree;

[0048] Step 1: Randomly initialize the particle's velocity v i =(v i1 ,...,v ij ,...,v iD ) and position x i =(x i1 ,...,x ij ,...,x iD );

[0049] Where D represents the dimension of the solution space, j∈{1,...,D}, x i represents the position of the i-th particle, v i represents the velocity of the i-th particle;

[0050] Step 2: Calculate the fitness of each particle f(x i );

[0051] Step 3: According to particle x i The fitness value updates the k-th iteration local optimal position pi And the global optimal position g, the update formula is as follows:

[0052]

[0053]

[0054] Among them, p i =(p i1 ,...,p id ,...,p iD ) represents the optimal position of the i-th particle, g=(g1,g2,...,g D ) represents the global optimal position determined by the particle swarm, and k represents the current number of iterations;

[0055] Step 4: Update the particle's velocity and position. The update formula is as follows:

[0056]

[0057]

[0058] Among them, p i =(p i1 ,...,p id ,...,p iD ) represents the optimal position of particle i, g=(g1,g2,...,g D ) represents the optimal position determined by the particle swarm, k represents the current number of iterations, c1 and c1 are acceleration factors used to determine p i and the relative importance of g; ∈1 and ∈2 are two random numbers in [0,1]; ω is the adjustment weight of the particle's search for the target source area. By adjusting the ω value, the particle has a greater tendency to explore the target source area and conducts a refined search in this area.

[0059] Step 5: Calculate the weighted average velocity of the particle to determine whether it is trapped in the obstacle area. The weighted average velocity calculation formula is as follows:

[0060]

[0061]

[0062] in, is the weighted average velocity of the i-th particle in the k-th iteration, is the velocity of the i-th particle in the k+1-th iteration; α is the weight coefficient, representing the time state of the current iteration;

[0063] Step 6: For the trapped particles, add the area near the current particle as a new taboo area to the taboo list and store it in the R-Tree;

[0064] Step 7: Give all particles in the taboo area a velocity along the line connecting the particles and the center of the taboo area to make them leave the area. The velocity formula is:

[0065]

[0066] Among them, V max represents the speed limit of the particle; is the position of the i-th particle in the k-th iteration, and c is the center of the taboo area;

[0067] Step 8: Update the particle swarm state;

[0068] Step 9: Determine whether the target source is located or the iteration termination condition is met. If so, jump to step 10; otherwise, jump to step 2.

[0069] Step 10: End the search.

[0070] In step 5, compared with the average velocity, the weighted average velocity can better reflect the state of the particle in the most recent time period. By expanding the weighted average velocity formula, the following conclusions are drawn:

[0071]

[0072] in, On each The weight of depends on the number of iterations; when k is large enough, the early iterations have little effect on The speed weight decays exponentially according to (1-α). If the value of α is large, It is mainly determined by the speed of the last few iterations. In the extreme case, α=1, that is

[0073] If the agent's weighted average speed in iteration k is If the particle velocity is less than the preset threshold γ, then it means that the position of the agent has hardly changed over time. If an agent is not close to its global optimal position, that is, the particle's velocity is moving in the opposite direction of the global optimal position, then it may be trapped near an obstacle.

[0074] In order to quantify the above phenomenon, the parameter η is introduced, taking the last n iterations as the main consideration, that is, the weighted average speed of the first n iterations is less than β. The formula is as follows:

[0075]

[0076] If the agent is stuck in a deadlock, it can only move between two points. As the number of iterations increases, the proportion of the weighted average speed of the most recent iteration also increases. When the proportion increases to a certain extent, the upper bound of the expected weighted average speed can be obtained. This upper bound can be used as a reference for the preset threshold value γ. Suppose an agent is trapped in a small range and makes a periodic motion. The speed can be considered as v i+1 =-v i ;

[0077] When k is large enough, if we mainly consider the last n iterations, the parameters satisfy:

[0078]

[0079]

[0080] In step 6, the taboo area is set as a taboo object and a suitable size is selected for it, and R-Tree is used to store it. Figure 2 As shown in Figure 2, if an agent moves from point 1 to point 2, it means that the agent has detected the shaded area and warned itself not to go there. Considering the maximum value of the speed constraint, the distance between any position in the shaded area and point 1 is less than the maximum value. Therefore, the midpoint between points 1 and 2 is the center, and the side length is The rectangle is set as the taboo area.

[0081] To more intuitively demonstrate the performance of the present invention in the signal source search problem, this embodiment embeds the proposed strategy into multiple source localization algorithms based on the particle swarm optimization algorithm. To verify the effectiveness of the algorithms, these algorithms are executed in two scenarios (with and without the taboo strategy). The test algorithms are single-source localization and multi-source localization algorithms, both in environments with many concave obstacles that can easily cause the robot to get stuck.

[0082] Two state-of-the-art single-source search methods are compared:

[0083] Adaptive robot PSO (A-RPSO) 4 (M.Dadgar, S.Jafari, and A.Hamzeh, "A pso-basedmulti-robot cooperation method for target searching in unknown environments," Neurocomputing, vol.177, no.C, pp.62–74, 2015.) and repulsion-based RDPSO (RbRDPSO) (M.Dadgar, M.Couceiro, and A. Hamzeh, "Rbrdpso: Repulsion-based rdpso for robotic target searching," Iranian Journal of Science and Technology, Transactions of Electrical Engineering, vol. 44, no. 1, pp. 551–563, 2019.).

[0084] The convergence diagram of the algorithm is as follows Figure 3 It is clear that by embedding the proposed tabu search strategy, their performance is significantly improved. The improvement of the tabu search strategy has good robustness under different numbers of agents. In A-RPSO and RDPSO, the original strategy may fall into local optimality in the early iterations. The tabu search strategy helps to escape the local optimality and converge to a smaller value.

[0085] For multi-source search, two algorithms are considered for comparison. One is the two-stage algorithm SAP-CFPSO (Q. Tang, L. Ding, F. Yu, Y. Zhang, Y. Li, and H. Tu, “Swarm robots search for multiple targets based on an improved grouping strategy,” IEEE / ACM Transactions on Computational Biology and Bioinformatics, vol. 15, no. 6, pp. 1943–1950, 2018.), which performs dynamic grouping search based on search auxiliary points. The other algorithm is the multi-source mechanical particle swarm algorithm (Multi-MPSO) (Q. Tang, F. Yu, Z. Xu, and P. Eberhard, “Swarm robots search for multiple targets,” IEEE Access, vol. 8, pp. 92 814–92 826, 2020.). This algorithm increases or decreases the number of robots in the neighborhood based on the difference between the individual performance of a robot and the average performance of all robots, achieving the goal of grouping and multi-source search.

[0086] The convergence diagram of the algorithm is as follows Figure 4 The proposed tabu search algorithm achieves better performance when the tabu search strategy is embedded into the deterministic obstacle avoidance strategy.

[0087] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A target source positioning and obstacle avoidance method based on tabu search and particle swarm algorithm, characterized in that: The method includes: Establish a fitness function model that characterizes the signal strength of the target source, regard the intelligent agent as a particle, initialize the particle swarm algorithm parameters and clear the taboo list; the intelligent agent is a robot; Based on the weighted average velocity of the particle, determine whether the particle falls into the taboo area and update the taboo list, and combine the taboo search to move and avoid obstacles; Update the fitness value and global and local optimal positions, and execute speed and position update strategies; Repeat the iteration until the agent locates the target source or the iteration termination condition is met, so that the agent locates the optimal target source; Based on the weighted average velocity of the particle, it is determined whether the particle falls into the taboo area and the taboo list is updated. The taboo search is combined to move and avoid obstacles. The specific process is as follows: If the weighted average velocity of the particle in k iterations If the value is less than the preset threshold γ and is not close to its global optimal position, the particle is considered to be trapped near an obstacle. For trapped particles, the area near the current trapped particle is added to the taboo list as a new taboo area and stored. All particles in the taboo area are given a velocity along the line connecting the particle and the center of the taboo area to make them leave the taboo area. The speed expression of the particle leaving the taboo area is: in, is the position of the i-th particle in the k-th iteration, c is the center of the taboo area, V max represents the speed limit of the particle; The taboo area is centered at the midpoint between the particle's previous iteration position and the current iteration position, and its side length is Rectangular area, where V max represents the speed limit of the particle; The preset threshold γ satisfies: If the agent is trapped near an obstacle, it will fall into a deadlock state and can only move between two points. As the number of iterations increases, the proportion of the weighted average speed of the most recent iteration also increases. When the proportion of the weighted average speed increases to a preset value, the upper bound of the weighted average expectation is obtained, which is the reference value of the preset threshold γ. The taboo list is stored using R-Tree.

2. The target source positioning and obstacle avoidance method based on tabu search and particle swarm algorithm according to claim 1, characterized in that: The initialization particle swarm parameters are specifically: Initialize the population size N, acceleration factors c1 and c2, problem dimension D, number of iterations T and clear the taboo list; randomly initialize the particle speed v i =(v i1 ,...,v ij ,...,v iD ) and position x i =(x i1 ,...,x ij ,...,x iD ), where i is the i-th particle.

3. The target source positioning and obstacle avoidance method based on tabu search and particle swarm algorithm according to claim 1, characterized in that: The update formulas for the local optimal position and the global optimal position are as follows: Among them, p i =(p i1 ,...,p id ,...,p iD ) represents the optimal position of the i-th particle, g=(g1,g2,...,g D ) represents the global optimal position determined by the particle swarm, and k represents the current number of iterations.

4. The target source positioning and obstacle avoidance method based on tabu search and particle swarm algorithm according to claim 3, characterized in that: The speed and position update strategy is as follows: in, is the velocity of the i-th particle in the k-th iteration, p i =(p i1 ,...,p id ,...,p iD ) represents the optimal position of the i-th particle, g=(g1,g2,...,g D ) represents the global optimal position determined by the particle swarm, k represents the current number of iterations; c1 and c2 are acceleration factors used to determine the optimal position p of the i-th particle. i and the relative importance of the global optimal position g determined by the particle swarm; ∈1 and ∈2 are two random numbers uniformly distributed in the interval [0,1]; ω is the adjustment weight of the particle search for the target source area.

5. The target source positioning and obstacle avoidance method based on tabu search and particle swarm algorithm according to claim 1, characterized in that: The weighted average speed expression is: in, is the weighted average velocity of the i-th particle in the k-th iteration, is the velocity of the i-th particle in the k+1-th iteration; α is the weight coefficient, representing the time state of the current iteration.

Images