A mobile robot path planning method based on improved African wild dog algorithm

Abstract

The application discloses a mobile robot path planning method based on an improved African wild dog algorithm, and comprises the following steps: S1, generating an initial population; S2, calculating population fitness; S3, wild dogs performing siege behavior; S4, wild dogs performing pursuit behavior; S5, wild dogs performing scavenging behavior; S6, detecting whether a new position of a wild dog individual is better than an original position, and updating position information; S7, calculating individuals needing to perform survival behavior, and calculating a new position of the individuals; and S8, outputting optimal wild dog individual optimal path information when a maximum cycle number is not reached. The improved African wild dog algorithm adopting the novel strategy has obvious effects in solving mobile robot path planning, and the overall performance is better than basic African wild dog algorithm and other improved ant colony optimization algorithms.

Description

Technical Field

[0001] This invention relates to the field of robot path planning technology, specifically a mobile robot path planning method based on an improved African wild dog algorithm. Background Technology

[0002] Path planning, simply put, is a technique for finding a safe, collision-free path for a mobile robot that meets specific requirements. Given different constraints in various application environments, finding the optimal path that satisfies the planning objectives is a crucial task in path planning research. Although numerous scholars and experts both domestically and internationally have made significant breakthroughs in robot path planning, proposing and improving many algorithms to address problems in its applications, some issues remain unresolved. Therefore, overcoming and solving these problems in path planning algorithms is one of the most important challenges in the field of robotics research.

[0003] Currently, many experts and scholars at home and abroad are dedicated to the research of path planning algorithms. Commonly used optimization algorithms include artificial potential field method, immune algorithm, ant colony optimization algorithm, neural network, particle swarm optimization algorithm and African wild dog algorithm.

[0004] The Dingo Optimization Algorithm (DOA) is a novel optimization algorithm proposed in 2021 by Hernán Peraza-Vázquez et al., which simulates the hunting process of African wild dogs. This algorithm simulates four hunting behaviors of African wild dogs: encirclement, pursuit, scavenging, and survival. In each iteration, each individual performs one or more of these four behaviors, resulting in a new position. Compared to other optimization algorithms, the Dingo algorithm boasts advantages such as simple logic, wide adaptability, and random search. However, it also has some drawbacks that need improvement. For example, during algorithm initialization, the positions of the wild dogs are randomly generated, resulting in generally low fitness. Secondly, when the wild dog group engages in encirclement and pursuit behaviors, it only searches for new positions near the current global optimum, which may lead to excessively high similarity in the positional information of the wild dog population, causing them to get trapped in local optima. Finally, the survival strategy of the wild dogs only applies to inferior individuals, resulting in a somewhat one-sided effect.

[0005] To address the aforementioned problems, this invention proposes an Improved Dingo Optimization Algorithm (IDOA), which improves upon three shortcomings of the traditional IOA algorithm. First, it incorporates known starting point information to construct a new initialization method for the wild dog's position. Second, referencing the displacement concept of particle swarm optimization, the pursuit strategy of the wild dogs is redesigned, taking into account the suboptimal individuals in the population. Finally, by adjusting the individual selection strategy for survival behavior, the dispersion of population fitness does not affect the selection process. Simulation results verify the superiority of the improved IOA algorithm in mobile robot path planning. Summary of the Invention

[0006] The purpose of this invention is to provide a mobile robot path planning method based on an improved African wild dog algorithm. This method can overcome the shortcomings of the traditional African wild dog algorithm, such as low wild dog adaptability and low pathfinding efficiency, thereby improving the efficiency of the algorithm in solving path planning tasks.

[0007] In traditional African wild dog algorithms, the initial positions of wild dogs are mainly generated randomly, which often results in low fitness of the initial solutions. This invention proposes a population initialization method that incorporates endpoint information. For the position information of each wild dog, i.e., path information, the transition probabilities between nodes are calculated according to equation (1) during generation. Let d be the probability that node I will move to node J from the location information of the m-th stray dog. IJ Let d be the Euclidean distance between node I and node J. JE Let J be the Euclidean distance between node J and destination E, and allowed be the adjacent passable nodes.

[0008]

[0009] In addition, the traditional African wild dog algorithm mainly relies on equation (2) when performing pursuit, where x new,i For the new location of the i-th stray dog, x i Let x be the current position of the i-th stray dog. best The globally optimal solution is the known optimal position, i.e., the globally optimal path solution. β1 is a uniformly random number on [-2, 2], β2 is a uniformly random number on [-1, 1], and x... r This represents the location of a random stray dog.

[0010]

[0011] As can be seen from equation (2), the essence of the pursuit behavior is to make the wild dogs search near the globally optimal individual. However, if the globally optimal solution is a local optimum, it may cause the wild dog population to fall into a local optimum. Based on the particle displacement model in the particle swarm optimization algorithm, this invention designs a new pursuit strategy, as shown in equation (3), where α i Let ε be the global learning factor for the i-th wild dog. i Its random learning factor, α i With ε i All are random numbers in [0,1], and satisfy α i +ε i =1, α for each wild dog i With ε i The parameters are assigned at the beginning of the algorithm and remain unchanged throughout the process. R To determine a random position from the population using a roulette wheel method, the probability of selection for each individual wild dog is calculated according to equation (4), fitness i Let represent the fitness of the i-th wild dog individual.

[0012]

[0013]

[0014] Then, in the original algorithm, at the end of each iteration, survival behaviors are performed on wild dog individuals with low survival probabilities in the population. The survival probability of each wild dog is related to its fitness value, calculated by equation (5). Individuals with low survival probabilities, i.e., less than 0.3, need to go back to the vicinity of the current best individual to forage for food in order to survive, thereby increasing their own survival probability.

[0015]

[0016] Obviously, Equation (5) calculates the survival probability of an individual mainly due to the absolute value of the fitness difference between the best and worst individuals in the population. When the fitness values ​​of the population are relatively dispersed, Equation (5) will have difficulty selecting inferior individuals who need to survive. This invention proposes a new survival behavior determination formula, as shown in Equation (6). Where needsr i Let x be the discriminant for whether the i-th individual needs to survive. r Let be a random wild dog individual in the population. Equation (6) uses the individual's quality in the population as the criterion, which satisfies the design requirement that better individuals have a low probability of survival and worse individuals have a high probability of survival, and is no longer affected by the different degrees of population fitness dispersion.

[0017]

[0018] The technical solution adopted by this invention to solve its technical problem is:

[0019] A mobile robot path planning method based on an improved African wild dog algorithm includes the following steps:

[0020] S1. Initialize algorithm parameters, such as maximum number of iterations T, population size M, probability of siege or hunt behavior P, probability of siege behavior Q, and global learning factor α and random learning factor ε for each wild dog.

[0021] S2. Initialize the location information, i.e., the path information, of the m-th wild dog (1≤m≤M), and initialize the path node list pathlist to be empty:

[0022] S2.1 Set the current node of the path as the starting point;

[0023] S2.2 Calculate the transition probability of adjacent nodes according to equation (1);

[0024] S2.3 Add the current node of the path to pathlist, and use the roulette wheel method to transfer nodes according to the node transfer probability calculated by equation (1), and update the current node of the node;

[0025] S2.4. Is the current node the endpoint? If yes, add the current node to pathlist and execute S2.5; otherwise, execute S2.2.

[0026] S2.5. Have all wild dogs completed initialization? If yes, proceed to S3; otherwise, select a wild dog that has not completed initialization and proceed to steps S2.1-S2.5.

[0027] S3. Calculate the fitness of the wild dog population, which is the length of the path solution represented by the location information of individual wild dogs.

[0028] S4. Remove the first wild dog individual and proceed to S5;

[0029] S5. Take a random number and check if it is less than P. If it is, proceed to S6; otherwise, proceed to S9.

[0030] S6. Take a random number and check if it is less than Q. If it is, proceed to S7; otherwise, proceed to S8.

[0031] S7. The wild dogs perform a siege behavior according to formula (7), where x new,i For the new location of the i-th stray dog, x i Let x be the current position of the i-th stray dog. k Let na be the current position of the kth wild dog, na be the number of wild dogs attacking, take a random integer in [2, N / 2], β be a uniformly distributed random number in the interval [-2, 2], and execute S10;

[0032]

[0033] S8. The wild dog performs the pursuit action according to formulas (3) and (4), and executes S10.

[0034] S9. Wild dogs perform carrion-eating behavior according to formula (8), where b is a random number on [-1,1], and x r1 It is a random individual, and then S10 is executed;

[0035]

[0036] S10. Detect whether the new location of the wild dog individual is better than its original location. If so, update the location information; otherwise, keep the original location unchanged.

[0037] S11. Calculate the individuals who need to perform survival behaviors according to formula (6), and calculate the new positions of these individuals according to formula (9), where x r1 ,x r2 They are two random individuals.

[0038]

[0039] S12. Has the maximum number of loops been reached? If yes, then execute S13; otherwise, execute S3.

[0040] S13. Output the location information of the best wild dog individual, i.e., the optimal path information.

[0041] The beneficial effects of this invention are as follows: by constructing the initial wild dog's location solution using a path construction strategy that incorporates the endpoint, the fitness of the initial wild dog can be effectively improved, providing better preconditions for subsequent processing in the algorithm; by designing a novel pursuit strategy, a population of wild dogs with different preferences possesses stronger global and local search capabilities, thereby accelerating the convergence speed of the algorithm; and by adjusting the selection strategy for individuals exhibiting survival behavior, the degree of dispersion of population fitness does not affect the selection of individuals, thus improving the algorithm efficiency to a certain extent. Extensive simulation results demonstrate that the improved African wild dog algorithm using the novel strategy presented in this paper significantly improves the path planning performance for mobile robots, outperforming the basic African wild dog algorithm and other improved ant colony optimization algorithms. Attached Figure Description

[0042] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0043] Figure 1 Flowchart of the improved African wild dog algorithm of this invention;

[0044] Figure 2 The path obtained by the traditional African wild dog algorithm;

[0045] Figure 3 This invention improves the path obtained by the African Wild Dog algorithm;

[0046] Figure 4 Algorithm convergence curve;

[0047] Figure 5 This invention improves the path obtained by the African Wild Dog algorithm;

[0048] Figure 6 Convergence curve of the improved African Wild Dog algorithm of this invention; Detailed Implementation

[0049] The purpose of this invention is to provide a mobile robot path planning method based on an improved African wild dog algorithm. This method can overcome the shortcomings of the traditional African wild dog algorithm, such as low wild dog adaptability and low pathfinding efficiency, thereby improving the efficiency of the algorithm in solving path planning tasks.

[0050] The technical solution adopted by this invention to solve its technical problem is:

[0051] See Figure 1 This invention provides a mobile robot path planning method based on an improved African wild dog algorithm, comprising the following steps:

[0052] S1. Initialize algorithm parameters, such as maximum number of iterations T, population size M, probability of siege or hunt behavior P, probability of siege behavior Q, and global learning factor α and random learning factor ε for each wild dog.

[0053] S2. Initialize the location information, i.e., the path information, of the m-th wild dog (1≤m≤M). Initialize the path node list pathlist to be empty:

[0054] S2.1 Set the current node of the path as the starting point;

[0055] S2.2 Calculate the transition probability of adjacent nodes according to equation (1);

[0056] S2.3 Add the current node of the path to pathlist, and use the roulette wheel method to transfer nodes according to the node transfer probability calculated by equation (1); and update the current node of the node.

[0057] S2.4. Is the current node the endpoint? If yes, add the current node to the pathlist and execute S2.5; otherwise, execute S2.2.

[0058] S2.5. Have all wild dogs completed initialization? If yes, proceed to S3; otherwise, select a wild dog that has not completed initialization and proceed to steps S2.1-S2.5.

[0059] S3. Calculate the fitness of the wild dog population, which is the length of the path solution represented by the location information of individual wild dogs.

[0060] S4. Remove the first wild dog individual and proceed to S5;

[0061] S5. Take a random number and check if it is less than P. If it is, proceed to S6; otherwise, proceed to S9.

[0062] S6. Take a random number and check if it is less than Q. If it is, proceed to S7; otherwise, proceed to S8.

[0063] S7. The wild dogs perform a siege behavior according to formula (7), where x new,i Given the new location information of the i-th wild dog, na is the number of wild dogs attacking, take a random integer in [2, N / 2], β is a uniformly distributed random number in the interval [-2, 2], and execute S10;

[0064]

[0065] S8. The wild dog performs the pursuit action according to formulas (3) and (4), and then executes S10;

[0066] S9. Wild dogs perform carrion-eating behavior according to formula (8), where b is a random number on [-1,1], and x r1 These are individual random entities. Then execute S10;

[0067]

[0068] S10. Detect whether the new location of the wild dog individual is better than its original location. If so, update the location information; otherwise, keep the original location unchanged.

[0069] S11. Calculate the individuals who need to perform survival behaviors according to formula (6), and calculate the new positions of these individuals according to formula (9), where x r1 ,x r2 They are two random individuals.

[0070]

[0071] S12. Has the maximum number of iterations been reached? If yes, proceed to S13; otherwise, proceed to S3.

[0072] S13. Output the location information of the best wild dog individual, i.e., the optimal path information.

[0073] The beneficial effects of this invention are as follows: by using a path construction strategy that incorporates the endpoint to construct the location solution of the first-generation wild dogs, the fitness of the first-generation wild dogs can be effectively improved, providing better preconditions for subsequent processing of the algorithm; by designing a novel pursuit strategy, a population of wild dog individuals with different preferences has stronger global and local search capabilities, thereby accelerating the convergence speed of the algorithm; and by adjusting the selection strategy of individuals with survival behavior, the degree of dispersion of population fitness will not affect the selection of individuals, thus improving the efficiency of the algorithm to a certain extent.

[0074] The effects of this invention can be further illustrated by the following simulation experiments:

[0075] To verify the correctness and rationality of this method, the improved algorithm was simulated using Python in a 20×20 grid environment model, and compared with the basic African wild dog algorithm. The main parameters of the algorithm were set as follows: wild dog population size M = 50, maximum number of iterations max_iter = 100, probability of attack or pursuit behavior P = 0.5, and probability of attack behavior Q = 0.7. Simulation results are as follows. Figure 2 , Figure 3 , Figure 4 As shown in Table 1:

[0076] Table 1 Comparison of Simulation Results

[0077]

[0078] As shown in Table 1, the traditional DOA algorithm finds the optimal solution of 30.3847m in the 100th generation, while the improved African Wild Dog algorithm of this invention finds the optimal solution of 29.2132m in only the 11th generation. Furthermore, the initial solution of the improved algorithm is only 40.6273m, far superior to the initial solution of 44.8699m of the traditional algorithm. Therefore, the improved African Wild Dog algorithm of this invention has significant advantages over the traditional African Wild Dog algorithm in both solution finding performance and convergence speed.

[0079] To further verify the effectiveness of the improved African wild dog algorithm proposed in this invention, performance experiments were conducted to compare the improved African wild dog algorithm with Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), and other improved ant colony algorithms in the literature. Another improved ant colony algorithm is described in the paper "Application of Improved Ant Colony Algorithm in Path Planning" published by Yang Beichen et al. in the journal *Computer Applications Research* in 2022. Simulations were performed using the method of this invention under the grid environment conditions shown in Figure 8 of that article, and the experimental results were compared and analyzed. The results are as follows... Figure 5 , Figure 6And as shown in Table 2.

[0080] Table 2 Comparison of Simulation Results

[0081]

[0082] As shown in Table 2, the improved African Wild Dog algorithm of this invention achieves an optimal solution of 29.2132m, which is superior to other heuristic algorithms. Furthermore, the number of iterations is significantly less than other heuristic algorithms. Although the performance difference between the improved algorithm in the literature and the improved algorithm of this invention is not significant in terms of optimal solution quality and solution-finding speed, the improved algorithm of this invention exhibits better stability. The above experiments demonstrate the feasibility and practicality of the improved African Wild Dog algorithm proposed in this invention for path optimization.

[0083] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention, or modify them into equivalent embodiments, without departing from the scope of the present invention. Therefore, any simple modifications, equivalent substitutions, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention, without departing from the scope of the present invention, shall still fall within the protection scope of the present invention.

Claims

1. A mobile robot path planning method based on an improved African wild dog algorithm, characterized in that, Includes the following steps: S1. Algorithm parameter initialization, such as maximum number of iterations T, population size M, probability of siege or hunt behavior P, probability of siege behavior Q, and global learning factor for each wild dog. and random learning factor ; S2, Initialize the m-th wild dog (1≤) m The location information of ≤M, i.e., path information, initializes the path node list. Empty: S2.1 Set the current node of the path as the starting point; S2.2 Calculate the transition probability of adjacent nodes according to formula (1); (1) in Let be the probability that node I will move to node J from the location information of the m-th stray dog. Let be the Euclidean distance between node I and node J. Let J be the Euclidean distance between node J and endpoint E. Adjacent to walkable nodes; S2.3, Add the current node of the path to... In the process, the node transition probability is calculated according to Equation (1), and the node transition is performed using the roulette wheel method, and the current node is updated. S2.

4. Is the current node the endpoint? If so, add the current node to the endpoint. If the condition is met, proceed to S2.5; otherwise, proceed to S2.

2. S2.

5. Have all wild dogs completed initialization? If yes, proceed to S3; otherwise, select a wild dog that has not completed initialization and proceed to steps S2.1-S2.

5. S3. Calculate the fitness of the wild dog population, which is the length of the path solution represented by the location information of individual wild dogs. S4. Remove the first wild dog individual and proceed to S5; S5. Take a random number and check if it is less than the probability P of the siege or pursuit. If it is, proceed to S6; otherwise, proceed to S9. S6. Take a random number and check if it is less than the probability Q of the siege. If it is, proceed to S7; otherwise, proceed to S8. S7. Wild dogs carry out siege behavior according to formula (7), where For the first i New location information for a stray dog. na To represent the number of wild dogs attacking, take [2, M The random integer in / 2] Given a uniformly distributed random number in the interval [-2, 2], execute S10; The globally optimal solution is the known optimal location, i.e., the globally optimal path solution. For the first i The current location of the stray dog. For the first k The current location of the stray dog; S8. The wild dog performs the pursuit action according to formulas (3) and (4), and then executes S10; (3) (4) in, Let be the global learning factor for the i-th wild dog. A uniformly random number in the interval [-2, 2]. A uniformly random number in the interval [-1, 1]. M Size of a wild dog population Its random learning factor, and All are random numbers in the range [0,1], and satisfy the following conditions: Each stray dog and The parameters are assigned at the beginning of the algorithm and remain unchanged thereafter. To determine the location to be randomly selected from the population using a roulette wheel method, the probability of selection for each individual wild dog is calculated according to equation (4). Let be the fitness of the i-th wild dog individual; S9. The wild dog performs carrion-eating behavior according to formula (8), where b is a random number on [-1,1]. It is a random individual, and then S10 is executed; (8) S10. Detect whether the new location of the wild dog individual is better than its original location. If so, update the location information; otherwise, keep the original location unchanged. S11. A new survival behavior determination formula (6) is adopted. (6) in Let i be the discriminant for whether the i-th individual needs to survive. A random wild dog individual within the population; And based on a new survival behavior determination formula (6), the individuals who need to perform survival behaviors are calculated, and the new positions of these individuals are calculated according to formula (9), where, They are two random individuals. (9) S12. Has the maximum number of loops been reached? If yes, then execute S13; otherwise, execute S3. S13. Output the location information of the best wild dog individual, i.e., the optimal path information.

Images