Mobile robot path planning method based on artificial mouse travel algorithm

Abstract

The invention discloses a robot path planning method based on an artificial mouse travel algorithm. The method comprises the following steps: constructing a map; the method comprises the steps of parameter setting, initialization, initial path fitness evaluation, global search, local search, new path evaluation and local enhancement search, a dynamic adaptive search radius strategy is used for dynamically adjusting a search range C, and when an exploration-development switching threshold value E is larger than 1, namely Egt; the method comprises the following steps: 1, starting global search by adopting a dual-mode hybrid strategy, Elt; the method comprises the following steps: 1, starting local search, adopting an adaptive probability Levy flight or small-range disturbance strategy, introducing a local search enhancement module, carrying out refined search on a current global optimal path Xbest through a simplex method Nelder-Mead, and evaluating the fitness of the optimal path of iteration each time through the algorithm; and outputting an optimal path sequence according to the fitness value of the recorded optimal path. The method has the characteristics of improving the search capability, obtaining the optimal path and the like.

Description

Technical Field

[0001] The present invention relates to the field of robotics and artificial intelligence technology, and in particular to a robot path planning method based on an artificial lemming algorithm. Background Art

[0002] Mobile robot path planning is the process of finding the optimal or feasible path from a starting point to a destination in a given environment through algorithms. This requires comprehensive consideration of obstacle avoidance, motion constraints, and mission requirements. Current research on mobile robot path planning focuses on adaptability to dynamic environments, multi-objective optimization, and the integration of intelligent and biomimetic algorithms.

[0003] The Artificial Lemmings Algorithm (ALA) is a biomimetic algorithm that simulates the intelligent behavior of lemming colonies. ALA primarily emulates three typical behaviors exhibited by lemmings: over-reproduction leading to food shortages and subsequent long-distance migration; burrowing for survival and reproduction; and using their keen senses of smell and hearing to locate food during foraging. The algorithm is simple in structure, easy to implement, and computationally efficient. It integrates strategies such as Brownian motion, random walks, and Lévy flight to balance exploration and exploitation. During the iterative process, adaptive parameter adjustments gradually narrow the exploration range as the number of iterations increases. However, this algorithm has certain limitations. The search radius and jump probability parameters are fixed, and there is a lack of adaptive mechanisms that dynamically adjust based on population diversity or iteration progress. Furthermore, there is a lack of a local fine-tuning strategy for the current optimal solution, which can lead to stuck in local optima and poor path search capabilities. Summary of the Invention

[0004] The purpose of the present invention is to overcome the above shortcomings and propose a robot path planning method based on the artificial lemming algorithm that can improve the search capability and obtain the optimal path.

[0005] The present invention is a robot path planning method based on the artificial lemming algorithm, wherein: the method specifically comprises the following steps: Step 1: Map construction: Construct a grid map for mobile robot path planning. Discretize the environment map into a grid map, including the obstacle part, the path planning starting point, the path planning end point, and the traversable map part. Each grid coordinate (x, y) corresponds to a node. Step 2: Path planning based on the artificial lemming algorithm: including parameter setting, initialization, initial path fitness evaluation, global search, local search, new path evaluation, and local enhanced search. The detailed steps are as follows: Step 2.1: Parameter setting: Set the global fitness optimal value Score and initialize it to infinity Score = inf; the initial global optimal position X bestis a 1×dim zero vector X best = zeros(1, dim), where dim is the dimension of the artificial lemmings algorithm, i.e., the number of nodes in the corresponding path planning. The initial fitness value of each individual is a 1×N zero vector, fitness = zeros(1, N), where N is the number of random paths initially generated, which will serve as the population size for the algorithm to search. The initial convergence curve storage parameter Convergence is empty, Convergence= [], which is used to store the current optimal fitness. The parameters UB and LB are the upper and lower bounds of the variable, respectively, used to limit the path nodes to within the boundaries of the constructed map, ensuring that the planned path is feasible in the actual environment. Step 2.2: Initialization: After inputting the parameters dim, N, UB, and LB, the algorithm generates an initial path solution space vector X of size N×dim: add the lower bound of the variable LB to the difference between UB and LB, and then multiply it by the random number rand; each random initial path X i,j The node coordinates are randomly distributed in the valid area of ​​the constructed map; the robot's traversable path, that is, the initial path solution space vector X is expressed as: Among them: UB j and LB j are the upper and lower bounds of the j-th variable, i=1,2,…,N, j=1,2,…,dim; the value range of the random number rand is (0, 1).

[0006] Step 2.3: Initial fitness evaluation: First, initialize the evaluation counter Fes = 0 and the global optimal path cost Score; after decoding each path in the population, calculate its comprehensive performance through the fitness function fobj. fobj uses path length as the optimization target and comprehensively considers obstacle collision penalty and smoothness constraints; in mobile robot path planning, the shorter the path length, the shorter the robot travels, the less energy and time it consumes. Avoiding collisions with obstacles is a basic requirement for path feasibility, while the smoothness of the path affects the stability and efficiency of the robot's motion; a greedy strategy is used to update the global optimal path X best The fitness value Score and the evaluation times Fes increase with iteration. When Fes reaches the preset threshold, the optimization is terminated. The preset threshold is the maximum evaluation times Maxfes. Finally, the coordinate sequence of the global optimal path of the robot is output, that is, the global optimal path X best ; Step 2.4: Establish a dynamic adaptive adjustment mechanism: Use the dynamic adaptive search radius strategy to dynamically adjust the search range C, expressed as: Where: Maximum search radius r max is the Euclidean norm, denoted as r max =norm(UB-LB), minimum search radius r min =0.01*r max ; Radius attenuation coefficient λ=5, Fes is the number of evaluations, Maxfes is the maximum number of evaluations; The search range C decays exponentially with the number of evaluations Fes. Initially, the maximum search radius r is the length of the map diagonal. max Perform a large-scale scan of the environment, and later use r min Precision optimization path; In order to prevent the search range from falling into the local optimum, a small-scale perturbation factor p is set, which is expressed as: Where: p max is the maximum jump probability, p min is the minimum jump probability; to reflect the degree of dispersion of the population in the solution space and to determine whether the artificial lemming algorithm converges prematurely, calculate the standard deviation of each column of X, diversity = std(X). The larger the standard deviation, the greater the individual differences in that dimension, that is, the better the population diversity; calculate the average value of the standard deviation of each dimension, d = mean(diversity), which represents the average diversity of the entire population in all dimensions; Step 2.5: Set the exploration-exploitation switching threshold: The path planning based on the artificial lemming algorithm intelligently switches between exploration and exploitation modes through the exploration-exploitation switching threshold E, where E is expressed as: Where: the evaluation progress parameter θ = 2*arctan(1-Fes / Maxfes) decreases monotonically with iteration, achieving a natural transition from global scanning to local optimization; log(1 / rand) follows a standard exponential distribution; Step 2.6: Global Search: When the exploration-exploitation switching threshold E is greater than 1, that is, E>1, global search begins. This stage adopts a dual-mode hybrid strategy, specifically including: Step 2.6.1: When the random number rand is less than 0.3, that is, rand < 0.3, start a hybrid search strategy based on all optimal and random individuals to advance the current path towards the historical optimal direction. At the same time, introduce Brownian motion noise to avoid regular oscillation, which is expressed as: Where: X i(Fes + 1) is the path of the i-th search individual at the (Fes + 1)-th evaluation, which is the possible traveling path of the mobile robot at this stage; X best (Fes) is the current optimal path; X i (Fes) is the path of the i-th search individual at the Fes-th evaluation; X a (Fes) is a path randomly selected from the population individuals, a ∈ 1, 2, …, N; r = 2 * rand - 1 is a random number uniformly distributed within the range [-1, 1]; B = exp(-x 2 / 2) / (2 * π)^½ is the Brownian motion noise; F is the random switching flag of the search direction; Step 2.6.2: When the random number rand is greater than or equal to 0.3, that is, rand ≥ 0.3, start the random walk strategy with an adaptive step size. This strategy allows the mobile robot to explore new paths in the global space through random walks, expressed as: where: the step size r2 = rand * (1 + sin(0.5 * Fes)) * C is the random walk factor, which is dynamically adjusted with the number of iterations; Step 2.7: Local search: When the exploration - exploitation switching threshold E is less than 1, that is, E < 1, start local search, specifically including: Step 2.7.1: When the random number rand is less than 0.5, that is, rand < 0.5, start the spiral search around the current optimal path X best expressed as: where: the spiral search coefficient S = rad * (sin(2 * π * r3) + cos(2 * π * r3)), r3 is a random number within (0, 1); the spiral search radius rad = sqrt(sum(X best (Fes) - X i (Fes))^2), sqrt is the square root function; Step 2.7.2: When the random number rand is greater than or equal to 0.5, that is, rand ≥ 0.5, if the random number rand is less than the small - range perturbation factor p, that is, rand < p, start the adaptive probabilistic Lévy flight. By randomly perturbing the current optimal path X best , it makes it possible for the mobile robot to jump out of the current local optimal path and explore a wider space, expressed as: Where: the jump intensity coefficient G=2*(sign(rand-0.5))*(1-Fes / Maxfes), sign is the sign function, and its value changes dynamically between [-2*(Fes / Maxfes), 2*(Fes / Maxfes)] to control the disturbance amplitude; in the Levy flight, beta=1.5 is the characteristic exponent of the Levy distribution, which controls the heavy-tail characteristic of the step length; gamma is the gamma function, Lt=1+beta, sigma is the scale parameter of the Levy step length, which controls the overall amplitude of the step length, sigma=(gamma*Lt*sin(π*beta / 2) / (gamma*(Lt / 2)*beta*2^((beta1) / 2)))^(1 / beta), Levy=randn(1,dim)*sigma / abs(randn(1,dim))^(1 / beta), rand(1,dim) is a Gaussian distributed random number of dim dimension; Step 2.7.3: When the random number rand is greater than or equal to 0.5, that is, rand ≥ 0.5, if the random number rand is greater than or equal to the small-scale perturbation factor p, that is, rand ≥ p, start the small-scale perturbation strategy. In the mobile robot path planning, this strategy uses Gaussian distributed random numbers to calculate the current optimal path X. best Perform a small-scale perturbation, expressed as: Among them: p1=0.05 is the basic scaling factor, which controls the amplitude of the perturbation; randn(1,dim) is a Gaussian distributed random number of dim dimension. Since Gaussian perturbation has a strong local development ability, it has a strong effect on X. best Small-scale perturbations can effectively prevent the path from falling into local narrow areas; Step 2.8: New path evaluation: The new path obtained in the global search phase and the local search phase is used to calculate the fitness value of the new path through the objective function fobj, which is expressed as: Among them: newfitness is the fitness value of the new path; The relationship between the fitness value of the new path and the fitness value of the current path is compared by the greedy strategy. If the new path X i When the fitness value newfitness of (Fes+1) is better than the historical value fitness(i) of the individual, the path data is updated immediately to ensure the continuous evolution of the population; if the new path fitness value is also better than the global optimal Score, the path planning coordinates X are updated. best, the fitness value fitness(i) of the path is updated to the optimal path fitness value Score; Step 2.9: Local enhancement search: Introduce a local search enhancement module to improve the quality and convergence accuracy of the path planned by the mobile robot. Use the Nelder-Mead simplex method to perform refined search on the current global optimal path X best to enhance the exploration ability of the algorithm, so as to improve the convergence accuracy and the quality of the solved path. At the same time, maintain the population diversity by replacing the worst path in the population; When the ratio of the current evaluation number Fes to the maximum evaluation number Maxfes is greater than 0.7, that is, Fes / Maxfes > 0.7, it means that the algorithm has reached the late stage of iteration. mod(Fes, 5) = 0 means that the remainder of the current iteration number Fes divided by 5 is 0. When Fes / Maxfes > 0.7 and mod(Fes, 5) = 0, it means that when more than 70% of the maximum evaluation number is reached, perform gradient-free Nelder-Mead local search every 5 evaluations. While ensuring the quality of path planning, greatly reduce the computational cost of the algorithm; The specific steps of this local enhancement search are as follows: First, when performing local search, use the function Fminsearch, which takes the objective function fobj, the current global optimal path X best and the search option options as inputs. After local search by the Fminsearch function, the fitness value newfit of the new path is output, which reflects the quality of the new path; If the fitness value newfit of the new path obtained by local enhancement search is better than the fitness value Score of the current global optimal path, that is, newfit < Score, then update the global optimal path X best and its fitness value Score, which means that a better path for the mobile robot to travel has been found, and it is used as the new global optimal path; Then, replace the worst path in the population with the new path P and the fitness value newfit obtained by Fminsearch: Use the function max to calculate the fitness values of each path in the population, so as to obtain the index value worst_id of the worst path; Then, replace the coordinate values and fitness values of the worst path with the new path P and the fitness value newfit, and eliminate the inferior path solutions to improve the quality of the entire population; Finally, the fitness value of the optimal path from each iteration of the algorithm is recorded in the Convergence data. By recording these values, you can visually observe the algorithm's convergence during the iteration process, helping to analyze the algorithm's performance and optimize the path planning. By continuously optimizing the path through local enhanced search, the mobile robot can find a more optimal path in complex environments. Step 3: Based on the fitness value of the recorded optimal path, output the optimal path sequence to obtain the final planned path.

[0007] The above-mentioned robot path planning method based on the artificial lemming algorithm, wherein: in step 1, the environment map is discretized into three grid maps of different sizes: 10*10, 20*20 and 30*30.

[0008] The above-mentioned robot path planning method based on the artificial lemming algorithm, wherein: in step 2.4, in the initial search stage of the algorithm, the evaluation number Fes tends to 0, that is, Fes→0, and the evaluation number index tends to 1, that is, exp(Fes)→1, the search range C is approximately equal to r max , that is, C≈r max When the robot is moving, it can search for possible paths in a larger area.

[0009] The above-mentioned robot path planning method based on the artificial lemming algorithm, wherein: in step 2.4, in the late search stage of the algorithm, the number of evaluations Fes tends to Maxfes, that is, Fes→Maxfes, the ratio of the number of evaluations Fes to Maxfes tends to 1, that is, (Fes / Maxfes)→1, the exponent of the radius attenuation coefficient λ tends to 0, that is, when exp(-λ) →0, the search range C tends to r min , that is, C≈r min When , the algorithm search tends to local fine search, which enables the robot to make fine adjustments near the discovered better path and converge to the optimal solution.

[0010] The above-mentioned robot path planning method based on the artificial lemming algorithm, wherein: in step 2.4, the maximum jump probability p max =0.5, minimum jump probability p min =0.1, to ensure that jitter is eliminated during the convergence phase.

[0011] The above-mentioned robot path planning method based on the artificial lemming algorithm, wherein: in step 2.4, the degree of dispersion of the population in the solution space, when the population distribution is dispersed, that is, when the path distribution is dispersed, the average value d of the standard deviation of each dimension is approximately equal to r max , the small-scale perturbation factor p tends to the maximum jump probability p max , that is, d≈r max , p→pmax , maintain 50% disturbance intensity to achieve high jump probability.

[0012] The above-mentioned robot path planning method based on the artificial lemming algorithm, wherein: in step 2.4, the degree of dispersion of the population in the solution space, when the population distribution is concentrated, that is, when the path distribution is concentrated, the average value d of the standard deviation of each dimension is approximately close to 0, and the small-scale disturbance factor p is close to the maximum jump probability p min , that is, d→0, p→p min , reduce the jump probability, and retain 10% disturbance intensity to prevent the algorithm from falling into the local optimum.

[0013] In the above-mentioned robot path planning method based on the artificial lemming algorithm, in step 2.6.1, F is a flag for randomly switching the search direction, which further increases the randomness and diversity of the search, helping the mobile robot to more comprehensively explore navigable paths on a global scale. It is expressed as: .

[0014] The above-mentioned robot path planning method based on the artificial lemming algorithm, wherein: in step 2.6.2, in the initial search stage of the algorithm, the step length is maintained at 1.5 times the reference step length, so that the mobile robot can explore in a larger range and increase the possibility of discovering new paths.

[0015] In the above-mentioned robot path planning method based on the artificial lemming algorithm, in step 2.6.2, in the later stage of the search of the algorithm, the step size is reduced to 0.6 times the reference step size, which helps to improve the stability of the path and avoid excessive fluctuations.

[0016] Compared with the prior art, the present invention has obvious beneficial effects. As can be seen from the above scheme, the present invention addresses the shortcomings of the ALA algorithm in lacking adaptive mechanism and local search ability, and integrates a dynamic adaptive mechanism, a hybrid Nelder-Mead method and a small-scale perturbation strategy. The dynamic adaptive mechanism enables the DMSALA algorithm to explore the global path in a large range in the early stage of the search, and finely optimize the local trajectory in the later stage, and dynamically adjust the search step size according to the obstacle density; the hybrid Nelder-Mead local enhancement search strategy uses the simplex method to optimize the current global optimal path X. bestA refined search enhances the algorithm's development capabilities, improving convergence accuracy and the quality of the solution path. It also maintains population diversity by replacing the worst paths in the population. A small-scale perturbation strategy monitors the dispersion of path distributions by measuring the average value d of the standard deviations of each dimension. When paths are concentrated, a small-scale perturbation is applied, enabling the algorithm to conduct a refined search near the current optimal solution, avoiding local optima and enhancing the algorithm's robustness in complex scenarios. These strategies form a comprehensive optimization framework of "global exploration-local optimization-fine-tuning enhancement," improving the search performance of robot path planning.

[0017] The beneficial effects of the present invention are further illustrated below through specific implementation methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of the present invention; Figure 2 This is a comparison of planned paths on a 10*10 scale map in an embodiment of the present invention; Figure 3 This is a comparison of planned paths on a 20*20 scale map in an embodiment of the present invention; Figure 4 This is a comparison of planned paths on a 30*30 scale map in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following, in conjunction with the accompanying drawings and preferred embodiments, details on the implementation, features and efficacy of a robot path planning method based on an artificial lemming algorithm proposed in the present invention are described in detail.

[0020] See also Figure 1 The present invention is a robot path planning method based on artificial lemming algorithm (DMSALA), wherein: the method specifically comprises the following steps: Step 1: Grid Map Construction and Path Encoding: Construct a grid map for mobile robot path planning. Discretize the environment map into three grid maps of different sizes: 10*10, 20*20, and 30*30. These maps include the obstacle area, the path planning starting point, the path planning end point, and the traversable map area. Each grid coordinate (x, y) corresponds to a dimension of the algorithm solution space. Step 2: Path planning based on the artificial lemming algorithm: including initial parameter setting, initialization, initial path fitness evaluation, global search, local search, new path evaluation, and hybrid Nelder-Mead local enhancement search.

[0021] Step 2.1: Before starting DMSALA, the parameters need to be initialized. The initial global fitness optimal value Score is set to infinity (Score = inf), which means that in the initial state, since no feasible path has been found, the Score is set to infinity, reflecting the unknown and complexity of path planning. The initial global optimal position X best is a 1×dim zero vector (X best = zeros(1, dim)), dim corresponds to the dimension or number of nodes of the path planning problem. Initially, the fitness value fitness of each individual is a 1×N zero vector (fitness = zeros(1, N)), where N represents the number of random paths initially generated, which will serve as the starting point for subsequent searches. The initial convergence curve storage parameter Convergence is empty (Convergence = []). It is used to store the current optimal fitness, which can be used to analyze the convergence of the algorithm and understand whether the algorithm gradually finds a better path during the iteration process. In addition, the parameters UB and LB are the upper and lower bounds of the variables, respectively, which are used to limit the path nodes to not exceed the map boundary, ensuring that the planned path is feasible in the actual environment. The initial parameter setting reserves computing space in advance, which helps to reduce the subsequent computing burden and improve the efficiency of the algorithm.

[0022] Step 2.2: Initialization: After inputting the parameters N, dim, UB, and LB, an initial path solution space vector X of size N×dim will be generated. The specific method is to add the lower bound LB of the variable to the difference between UB and LB and then multiply it by the random number rand (the value range is (0,1)). Each X(i, :) represents a random initial path, and its node coordinates are randomly distributed in the valid area of ​​the map. These random initial paths provide diverse inputs for the subsequent calculation of fitness values ​​through feasible paths. In the mobile robot path planning scenario, multiple possible travel routes for the robot are randomly generated, which lays the foundation for the subsequent algorithm to find a better path based on these routes. The multiple possible travel paths X of the robot are expressed as: Among them: UB j and LB j are the upper and lower bounds of the j-th variable, i=1,2,…,N, j=1,2,…,dim, and the value range of the random number rand is (0,1).

[0023] Step 2.3: Initial Fitness Evaluation: First, initialize the evaluation counter Fes = 0 and the global optimal path cost Score. After decoding each path in the population, calculate its overall performance using the fitness function fobj. fobj optimizes path length and takes into account obstacle collision penalties and smoothness constraints. In mobile robot path planning, a shorter path length means a shorter robot travel distance, consuming less energy and time. Avoiding collisions with obstacles is a fundamental requirement for path feasibility. Path smoothness affects the stability and efficiency of the robot's motion. A greedy strategy is used to update the global optimal path X. best And the fitness value Score, the evaluation number Fes increases with the iteration. When Fes reaches the preset threshold Maxfes, the optimization is terminated and the final output X best This is the coordinate sequence of the robot's global optimal path. It can be expressed as: Step 2.4: Establish a dynamic adaptive adjustment mechanism to adapt to the dynamic changes in path planning requirements of mobile robots in complex environments: Use a dynamic adaptive search radius strategy to dynamically adjust the search range C, expressed as: Where: Maximum search radius r max is the Euclidean norm, denoted as r max =norm(UB-LB), minimum search radius r min =0.01*r max , radius attenuation coefficient λ=5, current objective function evaluation times Fes, maximum objective function evaluation times Maxfes; in the initial stage, Fes→0, exp(Fes)→1, C≈r max , allowing the mobile robot to find possible paths in a larger area. In the later stage of the algorithm search, Fes→Maxfes, (Fes / Maxfes)→1, exp(-λ) →0, C≈r min The algorithm tends to search locally and finely, so that the robot can make fine adjustments near the found better path and converge to the optimal solution. The search radius C decays exponentially with the number of evaluations Fes. In the initial stage, the length of the map diagonal r max Scan the environment in a large area, and then use r min r min Precision optimization path.

[0024] In order to prevent the search range from falling into the local optimum, a small-scale perturbation factor p is set, which is expressed as: This mechanism ensures that the mobile robot can quickly locate a feasible path in a complex environment and eliminates unnecessary jitter during the convergence phase.max =0.5, minimum jump probability p min =0.1, which reflects the degree of dispersion of the population in the solution space and determines whether the artificial lemming algorithm converges prematurely. The standard deviation of each column of X is calculated as diversity=std(X). The larger the standard deviation, the greater the individual differences in that dimension, that is, the better the population diversity. The average of the standard deviations of each dimension is calculated as d=mean(diversity), which represents the average diversity of the entire population in all dimensions. When the path distribution is dispersed, that is, d≈r max , p→p max , maintain 50% perturbation intensity to achieve high jump probability. When the path distribution is concentrated, that is, d→0, p→p min , reducing the jump probability and retaining 10% of the base perturbation to prevent the algorithm from falling into a local optimum. This mechanism ensures that the mobile robot can quickly locate a feasible path in complex environments and eliminate unnecessary jitter during the convergence phase.

[0025] Step 2.5: The DMSALA method intelligently switches between exploration and exploitation modes through a dynamic threshold E, which is expressed as: The evaluation progress parameter θ=2*arctan(1-Fes / Maxfes) decreases monotonically with iteration, achieving a natural transition from global scanning to local optimization. Log(1 / rand) follows a standard exponential distribution λ=1, generating positive random numbers with most values ​​close to 0 and a few larger values, belonging to a long-tail distribution. When E>1, the robot performs large-scale environmental exploration, and when E≤1, it switches to localized path refinement. This mechanism significantly improves path discovery efficiency in complex obstacle environments.

[0026] Step 2.6: Global Search: When the exploration-exploitation switching threshold E is greater than 1, that is, E>1, global search begins; this means that a large-scale exploration of feasible paths for the mobile robot is required, and the algorithm enters the global search phase. This phase adopts a dual-mode hybrid strategy, and both use a real-time grid map detection matrix to ensure that the planned path is feasible in the actual environment. Specifically, it includes: Step 2.6.1: When the random number rand is less than 0.3, that is, rand < 0.3, start a hybrid search strategy based on all optimal and random individuals. In the context of mobile robot path planning, this strategy aims to advance the current path towards the historical optimal direction while introducing Brownian motion noise to avoid regular oscillations. It can be expressed as: Where: X i (Fes+1) is the path evaluated by the i-th search individual at the (Fes+1)th time, that is, the possible travel path of the mobile robot at this stage.best (Fes) is the current optimal path, representing the best path solution found by the algorithm for the mobile robot so far. i (Fes) is the path of the i-th search individual at the Fes-th evaluation, X a (Fes) is a randomly selected path among individuals in the population, a∈1,2,…,N, r=2*rand-1 is a random number uniformly distributed in the range [-1,1], which provides a certain degree of randomness for the adjustment of the path. B=exp(-x 2 / 2) / (2*π)½ is Brownian motion noise, which can prevent regular oscillations in the path during adjustment, making path exploration more flexible. F is the random switching flag for the search direction, which further increases the randomness and diversity of the search, helping the mobile robot to more comprehensively explore possible paths globally. It can be expressed as: Step 2.6.2: When the random number rand is greater than or equal to 0.3, that is, rand ≥ 0.3, start the random walk strategy with adaptive step size. In mobile robot path planning, this strategy allows the mobile robot to explore new paths in the global space through random walks. It can be expressed as: Where r2 = rand*(1+sin(0.5*Fes))*C is the random walk factor. The step size r2 is dynamically adjusted with the number of iterations. In the early stages of the algorithm, the step size is maintained at 1.5 times the baseline step size, allowing the mobile robot to explore a wider range and increasing the probability of discovering new paths. In the later stages, the step size is reduced to 0.6 times the baseline step size, helping to improve path stability and avoid excessive fluctuations.

[0027] These two strategies in the global search phase allow the mobile robot to explore possible paths globally in different ways. At the same time, they combine the real-time grid map detection matrix to ensure the feasibility of the path, laying the foundation for the mobile robot to plan a better travel path.

[0028] Step 2.7: Local search: When the exploration-exploitation switching threshold E is less than 1, that is, E<1, local search begins, which includes the following three strategies: Step 2.7.1: When the random number rand is less than 0.5, that is, rand < 0.5, this strategy will revolve around the current optimal path X within the limit. best Start a spiral search, expressed as: where: r3 = rand (r3 is a random number within (0, 1)) introduces a phase random term to ensure the diversity of the search direction, and S = rad * (sin(2 * π * r3) + cos(2 * π * r3)) is the spiral search coefficient. rad = sqrt(sum(X best (Fes)-X i (Fes))^2) is the spiral search radius. This spiral search helps the mobile robot to more carefully search for possible better paths near the current optimal path, further optimizing the path planning result.

[0029] Step 2.7.2: When the random number rand is greater than or equal to 0.5, that is, rand ≥ 0.5, if rand < p, start the adaptive probability Levy flight. In the path planning of the mobile robot, the core purpose of this strategy is to enhance the exploration ability of the algorithm and avoid the algorithm falling into a local optimal solution. By randomly perturbing the current optimal path X best , it enables the mobile robot to possibly jump out of the current local optimal path and explore a broader space. It is expressed as: where: G = 2 * (sign(rand - 0.5)) * (1 - Fes / Maxfes) is the jump direction and intensity coefficient, and its value dynamically changes between [-2 * (Fes / Maxfes), 2 * (Fes / Maxfes)], controlling the perturbation amplitude and enhancing the development ability in the later stage of the algorithm search. In the Levy flight, beta = 1.5 is the characteristic exponent of the Levy distribution, controlling the heavy-tailed characteristic of the step size. gamma is the gamma function, Lt = 1 + beta, sigma = (gamma * Lt * sin(π * beta / 2) / (gamma * (Lt / 2) * beta * 2^((beta1) / 2)))^(1 / beta), sigma is the scale parameter of the Levy step size, controlling the overall amplitude of the step size. Levy = randn(1, dim) * sigma / abs(randn(1, dim))^(1 / beta), randn(1, dim) is a Gaussian distribution random number in dim dimensions. Most of the step sizes in the Levy flight are small, but occasionally there are large jumps, generating breakthrough path mutations with a heavy-tailed distribution of β = 1.5, which enables the mobile robot to have the opportunity to discover better paths far from the current local optimal path, thus effectively escaping from the local optimum.

[0030] Step 2.7.3: When the random number rand is greater than or equal to 0.5, that is, rand ≥ 0.5, if rand ≥ p, start the small-range perturbation strategy. In the path planning of the mobile robot, this strategy uses Gaussian distribution random numbers to perform small-range perturbation on the current optimal path X best as follows: Where: p1=0.05 is the basic scaling factor, which controls the amplitude of the perturbation. randn(1,dim) is a Gaussian distributed random number of dim dimension. Since Gaussian perturbation has a strong local development ability, it has a strong effect on X. best Small-scale perturbations can effectively prevent the path from falling into local narrow areas and prevent the algorithm from falling into local optimality, thereby making the path planning of the mobile robot more flexible and optimized.

[0031] These strategies in the local search phase adjust and explore the current optimal path in different ways, which helps to find a better travel path in the mobile robot path planning and enhances the algorithm's ability to improve the quality and efficiency of path planning.

[0032] Step 2.8: New path evaluation part: Calculate the new path in the global search phase and the local search phase, and calculate the fitness value of the new path through the objective function fobj, which is expressed as: Among them: newfitness is the fitness value of the new path calculated by the algorithm; The relationship between the fitness value of the new path and the fitness value of the current path is compared by the greedy strategy. If the new path X i When the fitness value newfitness of (Fes+1) is better than the historical value fitness(i) of the individual, the path data is updated immediately to ensure the continuous evolution of the population; if the new path fitness value is also better than the global optimal Score, the path planning coordinates X are updated. best , the fitness value of the path fitness(i) is updated to the optimal path fitness value Score. It can be expressed as: Where: Score represents the fitness value of the current optimal path. Greedy selection ensures that the algorithm will not degenerate and always evolve towards a better path. best Updates guide the search direction of the entire population to avoid blind randomness.

[0033] Step 2.9: Local Enhanced Search: The local search enhancement module is introduced in the later stage of the method. Its core purpose is to improve the quality and convergence accuracy of the path planned by the mobile robot. Specifically, the current global optimal path X is obtained by the Nelder-Mead simplex method. best Perform refined searches and enhance the algorithm's development capabilities to improve convergence accuracy and the quality of solution paths, while maintaining population diversity by replacing the worst paths in the population.

[0034] When the ratio of the current number of evaluations to the maximum number of evaluations is greater than 0.7, i.e., Fes / Maxfes > 0.7, it indicates that the algorithm has reached the later stage of iteration. mod(Fes, 5) = 0 means that the remainder of the current iteration number Fes divided by 5 is 0. When Fes / Maxfes > 0.7 and mod(Fes, 5) = 0, it means that when more than 70% of the maximum number of evaluations is reached, a gradient-free Nelder-Mead local search is performed every 5 evaluations, which can significantly reduce the computational cost of the algorithm while ensuring the quality of path planning.

[0035] When performing local search, the Fminsearch function is used. This function takes the objective function fobj, the current global optimal path X best and the search option options('Display', 'off') as inputs. After local search by the Fminsearch function, the fitness value newfit of the new path is output, which reflects the quality of the new path. If the fitness value of the new path obtained by local enhanced search is better than the fitness value of the current global optimal path, i.e., newfit < Score, then the global optimal path X best and its fitness value Score are updated. This means that a better moving robot travel path has been found and is used as the new global optimal path.

[0036] Next, the worst path in the population is replaced with the new path P and the fitness value newfit obtained by Fminsearch. The specific operation is to use the max function to calculate the fitness values of each path in the population, so as to obtain the index value worst_id of the worst path. Then, the coordinate values and fitness value of the worst path are replaced with the new path P and the fitness value newfit. This can continuously eliminate poor path schemes and improve the quality of the entire population.

[0037] Where: The new path P and the fitness value newfit are the paths obtained by the Nelder-Mead simplex method through Fminsearch.

[0038] Finally, the fitness values of the optimal paths for each evaluation iteration of the algorithm are recorded in Convergence. By recording these values, the convergence situation of the algorithm during the iteration process can be intuitively observed, which helps to analyze the performance of the algorithm and the effect of optimizing path planning. By continuously optimizing the path through local enhanced search, the moving robot can find a better travel path in a complex environment.

[0039] The mobile robot path planning method based on the DMSALA method consists of modules such as parameter initial setting, initialization, main loop parameter setting, global search, local search, new path evaluation, and hybrid Nelder-Mead local enhanced search. The specific implementation steps are as follows: a) In the initialization stage of the mobile robot path planning algorithm, set the initial global fitness optimal value Score to infinity, denoted as Score = inf, indicating that the optimal path solution has not been obtained yet; assign the initial global optimal position X best as the zero vector X best = zeros(1, dim), representing the initial state of the path coordinates; initialize the fitness value of each individual as a zero vector of 1*N, denoted as fitness = zeros(1, N), that is, the fitness of each path plan is in an unevaluated state; at the same time, initialize the convergence curve storage parameter Convergence as empty, denoted as Convergence = [], to record the convergence characteristics during the algorithm iteration process.

[0040] b) In the initial fitness evaluation stage, first initialize the evaluation counter Fes = 0 and the global optimal path cost Score. After decoding each path in the population, fobj takes the path length as the optimization objective and calculates its comprehensive performance through the fitness function fobj. b) Construct a dynamic adaptive adjustment mechanism C to adapt to the dynamic change requirements of path planning for mobile robots in complex environments; set a small-range perturbation factor p to provide local adjustment parameters for path optimization. c) Determine the execution direction of the search strategy by calculating the value of E. When E > 1, trigger the global search module to conduct a global exploration of the mobile robot's working space. If the random number rand < 0.3, start a hybrid search strategy based on the global optimal solution and random individuals, integrating the historical optimal path information and the random exploration strategy to search for a better path; when rand ≥ 0.3, execute the adaptive step-size random walk strategy to conduct random sampling exploration in the global space. d) If the calculation result satisfies E ≤ 1, then start the local search module to focus on the path optimization in the currently explored area. When rand < 0.5, execute the spiral search strategy within the limit, performing a spiral local search centered on the current position; when rand ≥ 0.5, further judge that if rand < p, adopt the adaptive probability Levy flight strategy to apply random perturbations to the current optimal path X best to enhance the global exploration ability of the algorithm and avoid falling into local optimal solutions; if rand ≥ p, then execute the small-range perturbation strategy to finely adjust the path. e) During the new path evaluation phase, the new paths generated by both global and local searches are evaluated using an objective function to determine their fitness. This value quantifies how well the path solution meets the mobile robot's mission objectives. A greedy strategy is employed to compare the fitness of the new path with the current path, selecting the optimal path solution for iterative path optimization. f) When the ratio of the number of evaluations Fes to the maximum number of evaluations Maxfes satisfies Fes / Maxfes>0.7, and the result of Fes modulo 5 mod(Fes,5)=0, that is, when the evaluation process exceeds 70% of the total number of evaluations and every 5 evaluations, perform the gradient-free Nelder-Mead local enhancement search. best The Fminsearch function is called with the search option options('Display', 'off') as input parameters to perform a local search and output the fitness value newfit of the new path. The new path P and its fitness value newfit are used to replace the worst individual in the population, maintaining population diversity and ensuring that the algorithm continues to explore better solutions. g) After the algorithm iteration is terminated, the global optimal path and the optimal fitness value are recorded, and the global optimal path planning solution is output for the mobile robot.

[0041] Performance Analysis: To verify the effectiveness of this invention, we conducted experiments on a computer running the Windows 11 Pro operating system, an Intel(R) Core(TM) i9-14900KF 3.20 GHz processor, and 128 GB of memory. The experiments demonstrated significant performance improvements. On the CEC2017-CEC2022 datasets, we selected 20 test functions and found 15 optimal solutions. Compared to other algorithms, the ALA algorithm found 10 optimal solutions, the WMA algorithm found 5 optimal solutions, the PSO, BKA, BWO, and GWO algorithms each found 1 optimal solution, and neither the SFOA nor the GOOSE algorithms found an optimal solution, as shown in Table 1.

[0042] Table 1 Comparison results of some functions of the proposed method and other 8 algorithms in the CEC2017~2022 test set To verify the effectiveness of this method in mobile robot path planning, we constructed three simulated map environments of varying sizes: 10x10, 20x20, and 30x30 grid maps, with minimizing the robot's path length as the core optimization objective. Within each map, we simulated the complex obstacle layouts faced by the mobile robot. Through repeated runs, we recorded the optimal path length value, average, and standard deviation from 30 independent experiments.

[0043] In a 10*10 small-scale map, the present invention quickly identifies obstacle gaps and efficiently plans the shortest path through a collaborative strategy of global search and local optimization. Figure 2 As shown in the figure, as the map size expands to 20*20 and 30*30, the algorithm continues to accurately locate the optimal solution in complex environments with its dynamic adaptive adjustment mechanism and intelligent search mode switching capabilities. Figure 3 and Figure 4 As shown in Table 2, under three simulation environments of different scales, the present invention not only successfully obtains the global optimal value of path length, but also the average optimal value is the shortest path, and the standard deviation value is stable at a low level. The present invention successfully finds the optimal value and the average optimal value on three maps of different scales, proving that the method has excellent stability and robustness.

[0044] Table 2 Data comparison between the proposed method and other eight algorithms on three maps of different scales The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. A robot path planning method based on the artificial lemming algorithm, characterized by: The specific steps of the method include: Step 1: Map construction: Construct a grid map for mobile robot path planning, discretize the environment map into a grid map, including the obstacle part, the path planning starting point, the path planning end point, and the passable map part; each grid coordinate (x, y) corresponds to a node; Step 2: Path planning based on the artificial lemming algorithm: including parameter setting, initialization, initial path fitness evaluation, global search, local search, new path evaluation, and local enhanced search. The detailed steps are as follows: Step 2.1: Parameter setting: Set the global fitness optimal value Score and initialize it to infinity Score = inf; the initial global optimal position X best is a 1×dim zero vector X best =zeros(1,dim), where dim is the dimension of the artificial lemming algorithm, i.e., the number of nodes in the corresponding path planning. The initial fitness value of each individual is a 1×N zero vector fitness=zeros(1,N), where N represents the number of random paths initially generated, which will serve as the population size searched by the algorithm. The initial convergence curve storage parameter Convergence is empty, Convergence=[], which is used to store the current optimal fitness. The parameters UB and LB are the upper and lower bounds of the variable, respectively, used to limit the path nodes to not exceed the boundaries of the constructed map, ensuring that the planned path is feasible in the actual environment. Step 2.2: Initialization: After inputting the parameters dim, N, UB, and LB, the algorithm generates an initial path solution space vector X of size N×dim: the lower bound LB of the variable plus the difference between UB and LB is multiplied by the random number rand; each random initial path X i,j The node coordinates are randomly distributed in the valid area of ​​the constructed map; the robot's traversable path, that is, the initial path solution space vector X is expressed as: X i,j =LB j +rand*(UB j -LB j ) Among them: UB j and LB j are the upper and lower bounds of the j-th variable, i = 1, 2, ..., N, j = 1, 2, ..., dim; the value range of the random number rand is (0, 1); Step 2.3: Initial fitness evaluation: First, initialize the evaluation counter Fes = 0 and the global optimal path cost Score; after decoding each path in the population, calculate its comprehensive performance using the fitness function fobj. fobj uses path length as the optimization target and comprehensively considers obstacle collision penalties and smoothness constraints; in mobile robot path planning, a shorter path length means a shorter robot travel distance, less energy and time consumed, and avoiding collisions with obstacles is a basic requirement for path feasibility, while path smoothness affects the stability and efficiency of the robot's motion; a greedy strategy is used to update the global optimal path X best The fitness value Score and the evaluation times Fes increase with iteration. When Fes reaches the preset threshold, the optimization is terminated. The preset threshold is the maximum evaluation times Maxfes. Finally, the coordinate sequence of the global optimal path of the robot is output, that is, the global optimal path X best ; Step 2.4: Establish a dynamic adaptive adjustment mechanism: Use the dynamic adaptive search radius strategy to dynamically adjust the search range C, expressed as: C=r min +(r max -r min )*exp(-λ*Fes / Maxfes) Where: Maximum search radius r max is the Euclidean norm, denoted as r max =norm(UB-LB), minimum search radius r min =0.01*r max ; Radius attenuation coefficient λ = 5, Fes is the number of evaluations, Maxfes is the maximum number of evaluations; The search range C decays exponentially with the number of evaluations Fes. Initially, the maximum search radius r is the length of the map diagonal. max Perform a large-scale scan of the environment, and later use r min Precision optimization path; In order to prevent the search range from falling into the local optimum, a small-scale perturbation factor p is set, which is expressed as: p=p min +(p max -p min )*(d / r max ) Where: p max is the maximum jump probability, p min is the minimum jump probability; to reflect the degree of dispersion of the population in the solution space and to determine whether the artificial lemming algorithm converges prematurely, calculate the standard deviation of each column of X, diversity = std(X). The larger the standard deviation, the greater the individual differences in that dimension, that is, the better the population diversity; calculate the average value of the standard deviation of each dimension, d = mean(diversity), which represents the average diversity of the entire population in all dimensions; Step 2.5: Set the exploration-exploitation switching threshold: The path planning method based on the artificial lemming algorithm intelligently switches between the exploration and exploitation modes through the exploration-exploitation switching threshold E, where E is expressed as: E=2*log(1 / rand)*θ Where: the evaluation progress parameter θ = 2*arctan(1-Fes / Maxfes) decreases monotonically with iteration, achieving a natural transition from global scanning to local optimization; log(1 / rand) follows a standard exponential distribution; Step 2.6: Global Search: When the exploration-exploitation switching threshold E is greater than 1, that is, E>1, global search begins. This stage adopts a dual-mode hybrid strategy, specifically including: Step 2.6.1: When the random number rand is less than 0.3, that is, rand < 0.3, start a hybrid search strategy based on all optimal and random individuals to advance the current path towards the historical optimal direction. At the same time, introduce Brownian motion noise to avoid regular oscillation, which is expressed as: X i (Fes+1)=X best (Fes)+F*B*(r*(X best (Fes)-X i (Fes))+(1-r)*(X i (Fes)-X a (Fes))) Where: X i (Fes+1) is the path of the i-th search individual at the Fes+1 evaluation, that is, the possible travel path of the mobile robot at this stage; X best (Fes) is the current optimal path; X i (Fes) is the path of the i-th search individual at the Fes-th evaluation; X a (Fes) is a randomly selected path among individuals in the population, a∈1,2,…,N; r=2*rand-1 is a random number uniformly distributed in the range [-1,1]; B=exp(-x 2 / 2) / (2*π)1 / 2 is Brownian motion noise; F is the random switching flag of the search direction; Step 2.6.2: When the random number rand is greater than or equal to 0.3, that is, rand ≥ 0.3, start the random walk strategy with adaptive step size. This strategy allows the mobile robot to explore new paths in the global space by random walk, which is expressed as: X i (Fes+1)=X i (Fes)+F*r2*(X best (Fes)-X a (Fes)) Among them: step size r2=rand*(1+sin(0.5*Fes))*C is the random walk factor, which is dynamically adjusted with the number of iterations; Step 2.7: Local Search: When the exploration-exploitation switching threshold E is less than 1, that is, E<1, local search begins, specifically including: Step 2.7.1: When the random number rand is less than 0.5, that is, rand < 0.5, around the current optimal path X best Start a spiral search, expressed as: X i (Fes+1)=X best (Fes)+F*X i (Fes)*S*r3 Where: spiral search coefficient S = rad*(sin(2*π*r3)+cos(2*π*r3)), r3 is a random number in (0,1); spiral search radius rad = sqrt(sum(X best (Fes)-X i (Fes))^2), sqrt is the square root function; Step 2.7.2: When the random number rand is greater than or equal to 0.5, i.e., rand≥0.5, and if the random number rand is less than the small-range perturbation factor p, i.e., rand<p, start the adaptive probabilistic Lévy flight and randomly perturb the current optimal path X best , so that the mobile robot has the possibility to jump out of the current local optimal path and explore a wider space, which is expressed as: X i (Fes+1)=X best (Fes)+F*G*Levy(dim)*(X best (Fes)-X i (Fes)) Where: jump intensity coefficient G = 2*(sign(rand-0.5))*(1-Fes / Maxfes), sign is the sign function, the value of G changes dynamically between [-2*(Fes / Maxfes), 2*(Fes / Maxfes)], controlling the amplitude of the disturbance; in Lévy flight, beta = 1.5 is the characteristic exponent of the Lévy distribution, controlling the heavy-tail characteristic of the step length; gamma is the gamma function, Lt = 1+beta, sigma is the scale parameter of the Lévy step length, controlling the overall amplitude of the step length, sigma=(gamma*Lt*sin(π*beta / 2) / (gamma*(Lt / 2)*beta*2^((beta1) / 2)))^(1 / beta), Levy=randn(1,dim)*sigma / abs(randn(1,dim))^(1 / beta), rand(1,dim) is a Gaussian distributed random number of dim dimension; Step 2.7.3: When the random number rand is greater than or equal to 0.5, that is, rand ≥ 0.5, if the random number rand is greater than or equal to the small-scale perturbation factor p, that is, rand ≥ p, start the small-scale perturbation strategy. In the mobile robot path planning, this strategy uses Gaussian distributed random numbers to calculate the current optimal path X. best Perform a small-scale perturbation, expressed as: X i (Fes+1)=X best (Fes)+p1*C*randn(1,dim) Among them: p1 = 0.05 is the basic scaling factor, which controls the amplitude of the disturbance; randn (1, dim) is the Gaussian distribution random number of dim dimension. Since Gaussian disturbance has strong local development ability, it has a great influence on X. best Small-scale perturbations can effectively prevent the path from falling into local narrow areas; Step 2.8: New path evaluation: The new path obtained in the global search phase and the local search phase is used to calculate the fitness value of the new path through the objective function fobj, which is expressed as: newfitness=fobj(X i (Fes+1)) Among them: newfitness is the fitness value of the new path; The relationship between the fitness value of the new path and the fitness value of the current path is compared by the greedy strategy. If the new path X i When the fitness value newfitness of (Fes+1) is better than the historical value fitness(i) of the individual, the path data is updated immediately to ensure the continuous evolution of the population; if the new path fitness value is also better than the global optimal Score, the path planning coordinates X are updated. best , the fitness value fitness(i) of the path is updated to the optimal path fitness value Score; Step 2.9: Local enhanced search: Introduce the local search enhancement module and use the simplex method Nelder-Mead to find the current global optimal path X best Conduct refined searches and enhance the algorithm's development capabilities to improve convergence accuracy and the quality of solution paths, while maintaining population diversity by replacing the worst paths in the population; When the ratio of the current number of evaluations Fes to the maximum number of evaluations Maxfes is greater than 0.7, that is, Fes / Maxfes>0.7, it means that the algorithm has reached the late stage of iteration. Mod(Fes,5)=0 means that the remainder of the current number of iterations Fes divided by 5 is 0. When Fes / Maxfes>0.7 and mod(Fes,5)=0 means that the maximum number of evaluations is exceeded, a gradient-free Nelder-Mead local search is performed every 5 evaluations. The specific steps of this local enhanced search are: First, when performing local search, the function Fminsearch is used. This function takes the objective function fobj, the current global optimal path X best and the search options options as inputs. After local search by the Fminsearch function, the fitness value newfit of the new path is output. This value reflects the quality of the new path. If the fitness value newfit of the new path obtained by local enhanced search is better than the fitness value Score of the current global optimal path, that is, newfit < Score, then update the global optimal path X best and its fitness value Score. This means that a better moving robot travel path has been found and it is used as the new global optimal path; Then, use the new path P and fitness value newfit obtained by Fminsearch to replace the worst path in the population: use the function max to calculate the fitness value of each path in the population, and thus obtain the index value worst_id of the worst path; then, use the new path P and fitness value newfit to replace the coordinate value and fitness value of the worst path, eliminating the poor path solution and improving the quality of the entire population; Finally, the fitness value of the optimal path for each evaluation iteration of the algorithm is recorded in Convergence. By recording these values, you can intuitively observe the convergence of the algorithm during the iteration process, help analyze the algorithm's performance and the effect of optimizing path planning. By continuously optimizing the path through local enhancement search, the mobile robot can find a more optimal travel path in complex environments. Step 3: Based on the fitness value of the recorded optimal path, output the optimal path sequence to obtain the final planned path.

2. The robot path planning method based on the artificial lemming algorithm as claimed in claim 1, characterized in that: In step 1, the environment map is discretized into three grid maps of different sizes: 10*10, 20*20, and 30*30.

3. The robot path planning method based on the artificial lemming algorithm as claimed in claim 1, characterized in that: In step 2.4, in the initial search phase of the algorithm, when the number of evaluations Fes approaches 0, i.e., Fes→0, and the number of evaluations exponent approaches 1, i.e., exp(Fes)→1, the search range C is approximately equal to r max , that is, C≈r max When the robot is moving, it can search for possible paths in a larger area.

4. The robot path planning method based on the artificial lemming algorithm as claimed in claim 1, characterized in that: In step 2.4, in the later stage of the search of the algorithm, the number of evaluations Fes tends to Maxfes, that is, Fes→Maxfes, the ratio of the number of evaluations Fes to Maxfes tends to 1, that is, (Fes / Maxfes)→1, the exponent of the radius attenuation coefficient λ tends to 0, that is, exp(-λ)→0, and the search range C tends to r min , that is, C≈r min When , the algorithm search tends to local fine search, which enables the robot to make fine adjustments near the discovered better path and converge to the optimal solution.

5. The robot path planning method based on the artificial lemming algorithm as claimed in claim 1, characterized in that: In step 2.4, the maximum jump probability p max =0.5, minimum jump probability p min = 0.1, to ensure that jitter is eliminated during the convergence phase.

6. The robot path planning method based on the artificial lemming algorithm according to claim 1, wherein: In step 2.4, the degree of dispersion of the population in the solution space is such that when the population distribution is dispersed, that is, when the path distribution is dispersed, the average value d of the standard deviation of each dimension is approximately equal to r max , the small-scale perturbation factor p tends to the maximum jump probability p max , that is, d≈r max ,p→p max , maintaining 50% perturbation intensity to achieve high jump probability.

7. The robot path planning method based on the artificial lemming algorithm according to claim 1, wherein: In step 2.4, the degree of dispersion of the population in the solution space is such that when the population distribution is concentrated, that is, when the path distribution is concentrated, the average value d of the standard deviation of each dimension tends to 0, and the small-scale perturbation factor p tends to the maximum jump probability p. min , that is, d→0, p→p min , reduce the jump probability, and retain 10% perturbation intensity to prevent the algorithm from falling into the local optimum.

8. The robot path planning method based on the artificial lemming algorithm according to claim 1, wherein: In step 2.6.1, F is the random switching flag for the search direction, which further increases the randomness and diversity of the search, helping the mobile robot to more comprehensively explore navigable paths globally. It is expressed as:

9. The robot path planning method based on the artificial lemming algorithm according to claim 1, wherein: In step 2.6.2, during the initial search phase of the algorithm, the step length is maintained at 1.5 times the baseline step length, enabling the mobile robot to explore within a larger range and increasing the possibility of discovering new paths.

10. The robot path planning method based on the artificial lemming algorithm according to claim 1, wherein: In step 2.6.2, in the later stage of the search of the algorithm, the step size is reduced to 0.6 times the base step size, which helps to improve the stability of the path and avoid excessive fluctuations.

Images