Automatic guided vehicle path planning method based on selection probability correction ant colony algorithm

Abstract

The invention discloses an automatic guided vehicle path planning method based on a selection probability correction ant colony algorithm, relates to the field of intelligent optimization algorithms, and is suitable for logistics path optimization, robot navigation, intelligent traffic scheduling, network routing optimization and other application scenes needing to solve a combinatorial optimization problem. According to the method, an ant colony algorithm automatic guided vehicle path planning model is constructed, a straight movement preference probability correction mechanism is established, a sudden change point probability correction strategy is formulated, a pheromone redistribution rule is designed, the short-distance direction change frequency is reduced, the occurrence frequency of sudden change points is reduced, the attraction of a later-stage high-quality path is enhanced, and the path smoothness is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent optimization algorithms, and particularly to a path optimization method based on the Ant Colony Optimization (ACO). This method is applicable to optimal path planning in complex dynamic environments, including but not limited to logistics path optimization, robot navigation, intelligent transportation scheduling, network routing optimization, and other application scenarios that require solving combinatorial optimization problems. Background Art

[0002] Automated guided vehicles are increasingly widely used in fields such as intelligent manufacturing, warehousing logistics, and the service industry. One of its core technologies is the path planning problem. The goal of path planning is to design an optimal path for the automated guided vehicle from the starting point to the target point, ensuring that obstacles are avoided and tasks are efficiently completed under known environmental information. Traditional path planning methods are mostly based on deterministic algorithms (such as A* and Dijkstra algorithms). Although these methods perform well in static environments, they often lack flexibility and robustness in complex dynamic scenarios.

[0003] As an intelligent optimization algorithm, the Ant Colony Optimization (ACO) has been widely applied to path planning problems due to its swarm intelligence characteristics and global search ability. Dorigo et al. first proposed the ant colony algorithm, and its core idea is based on bionics. By simulating the behavior of ants releasing and perceiving pheromones during the foraging process, path optimization is achieved. Since then, the ant colony algorithm has been widely applied to fields such as vehicle path optimization and robot navigation. However, the traditional ant colony algorithm has significant deficiencies when solving path planning problems in dynamic complex environments. For example, under the dual attraction of the heuristic value and pheromone intensity, the ant colony tends to be conservative in the later stage of exploration and is easily trapped in local optima. In addition, the algorithm has a weak adaptability to complex environments during the path planning process and is difficult to effectively handle sharp turns, dead ends, and other problems.

[0004] In recent years, researchers have improved the performance of the ant colony algorithm through various strategies. For example, some studies focus on optimizing the pheromone distribution strategy, such as introducing a dynamic pheromone evaporation mechanism and a pheromone concentration optimization method based on multi-scale updates; other studies focus on designing more effective heuristic functions to enhance the global search ability of ants. In addition, some work is dedicated to enhancing the algorithm's adaptability to dynamic environments, such as local area optimization by combining map knowledge and methods based on dynamic path replanning. Although the above studies have improved the performance of the ant colony algorithm to a certain extent, most methods still have the following limitations:

[0005] 1. Lack of effective handling of sharp turns and complex terrains, which easily leads to uneven paths;

[0006] 2. Insufficient improvements to the pheromone distribution strategy caused the algorithm to over-rely on the initial path in the later stages of exploration, affecting global optimization performance;

[0007] 3. In dynamic scenarios, algorithm flexibility and efficiency still need to be improved.

[0008] Therefore, there is an urgent need for an automated guided vehicle path planning method based on a selection probability-corrected ant colony algorithm to solve the above problems. Summary of the Invention

[0009] A method for automated guided vehicle path planning based on a selection probability-corrected ant colony algorithm is characterized by constructing an ant colony algorithm-based automated guided vehicle path planning model, establishing a straight preference probability correction mechanism, formulating a mutation point probability correction strategy, and designing pheromone redistribution rules to reduce the frequency of short-distance direction changes, reduce the number of mutation points, enhance the attractiveness of later high-quality paths, and improve path smoothness. The method specifically includes the following steps:

[0010] (1) Constructing an ant colony algorithm automatic guided vehicle path planning model:

[0011] Construct the optimization objective function of the ant colony algorithm automatic guided vehicle path planning, which is to minimize the path L k :

[0012]

[0013] Among them, L k is the length of the path taken by the kth ant from the starting point to the end point, q is the total number of ants, which is 25, n is the number of nodes, c ik is a node c that the kth ant passes through i , c (i+1)k The kth ant passes through node c ik The next node, d(c ik ,c (i+1)k ) is node c ik and c (i+1)k The Manhattan distance between Ф is the starting point, z Ф k is the starting decision variable, z Ф k There are only two values 0 and 1, z Ф k is 1, which means the kth ant starts from the starting point, z Ф k is 0, which means the kth ant did not start from the starting point, z ψ is the end point, z ψ k is the endpoint decision variable, z ψ kThere are only two values, 0 and 1, for z ψ k being 1 means that the k-th ant ends at the destination, z ψ k being 0 means that the k-th ant does not end at the destination, (c ik ,c (i+1)k ) is the node c ik and c (i+1)k forms the edge, E is the set of passable edges, c jk is a node passed by the k-th ant c j ;

[0014] (2) Establish a straight-line preference probability correction mechanism:

[0015] Calculate the selection probability of each candidate point:

[0016]

[0017] where O is the set of all h candidate points, c x is a candidate point in O, x ∈ [1, h], c i is the current node, c j is the node for which the probability is currently being calculated, j ∈ [1, h], P ij k is the probability that ant k at node c i selects node c j , α is the pheromone weight value, with a value of 1, τ ij is the pheromone concentration between node c i and c j , τ ij α is the α-th power of τ ij , β is the heuristic information value weight, with a value of 5, η ij is the heuristic information value between node c i and c j , η ij β is the β-th power of η ij , τ ix is the pheromone concentration between node c i and c x , τ ix α is the α-th power of τ ix , η ix is the heuristic information value between node c i and c x , η ix β is the β-th power of η ix ;

[0018] Adjust the selection probability of straight-ahead candidate points:

[0019]

[0020] Among them, P is the set of selection probabilities of all candidate points, P max is the maximum probability in the set P, P B is the selection probability of the straight-ahead candidate point, P B ' is the adjusted selection probability of the straight-ahead candidate point;

[0021] Generate a random number r1 distributed on [0, 1], and select the first P u ≥ r1 from P u (P u is the selection probability of the candidate point c u ), and set the node c u corresponding to P u as the next node;

[0022] (3) Develop a mutation point probability correction strategy:

[0023] ① Judge whether W(c i ) * W(c u ) is less than 0 (W(c i ) is the direction vector of the node c i , W(c u ) is the direction vector of the node c u ). If so, execute the mutation point probability correction strategy; otherwise, execute ⑥;

[0024] ② Enter the current node c i into the mutation point set M and the mutation point set M k of the k-th ant, and record the number of times the node c i appears in M as m i ;

[0025] ③ Judge whether m i is less than or equal to 5. If so, execute ④, otherwise execute ⑥;

[0026] ④ Enter the node c i into the skip point set T and the skip point set T k of the k-th ant, delete the node c i from M k and O, delete P i from P (P i is the selection probability of the candidate point c i ), and record the number of times the node c i appears in T as t i , and when t i is equal to five, for m iAdjust with formula (4):

[0027]

[0028] where f is the change in the excellent rate;

[0029]

[0030] where g is the number of times of y k <Y k ;

[0031] ⑤ Generate another random number r2 distributed on [0, 1], and select the first P from P that satisfies P u ≥ r2, and set the node c u corresponding to P u as the next node; u

[0032] ⑥ Update the next node to the current node, repeat the above process until ant k reaches the end point, calculate and record the length L k of the path that the k-th ant walks from the starting point to the end point, and add pheromone to the current path through the pheromone reallocation rule;

[0033] ⑦ Determine whether M k ∩ T k is an empty set. If so, the mutation point probability correction strategy ends. Otherwise, after finding the end point, perform the following steps. Specifically: sequentially take out the element c k from M k , extract the points adjacent to c k from S k and calculate the length of this section of the path and record it as Y k . If it is calculated again, take the average value. Sequentially take out the element c k from T k , extract the points adjacent to c k from S k and calculate the length of this section of the path and record it as y k . If it is calculated again, take the average value. If y k <Y k , then g = g + 1;

[0034] (4) Design the pheromone reallocation rule:

[0035] After ant k finds the end point, add pheromone to the path passed this time:

[0036]

[0037] where Δτ ij k is the k-th ant at node c​i and c j The pheromone concentration added between them, ρ is the pheromone evaporation rate, with a value of 0.3;

[0038]

[0039] where S k is the set of all nodes passed by ant k from the starting point to the ending point, Q is the pheromone gain weight, with a value of 1;

[0040] When the number of iterations reaches 5 times, change Δτ according to formula (8) ij k :

[0041]

[0042] where V is the current number of iterations, and L b is the current shortest path length;

[0043] Select the smaller value between L b and L k as the new L b value. The next ant starts from the starting point and repeats the above process until the last ant reaches the ending point. Output the optimal path in the form of an ordered coordinate sequence and send it to the automated guided vehicle through the control module. The automated guided vehicle drives to the ending point sequentially according to this path.

[0044] In formula (3) of the innovative part (2) of this article, the entire mutation point probability correction in the third part is completely innovative. The pheromone reallocation is formula (8), and the three mechanisms are linked.

[0045] Differences from other algorithms: Other algorithms mainly focus on pheromone iteration and the optimization of pheromone generation. However, this optimization directly targets the optimization of node selection probability. In nature, it is difficult for ants to have the motivation to continue looking for a better solution after finding a suboptimal solution, unless the original path is no longer passable. The same is true for the ant colony algorithm. This is the significance of the mutation point probability correction mechanism. The straight-line probability correction mechanism also provides a tendency for the ant colony from the probability, which is the most fundamental factor determining path selection. The mutation point probability correction can prevent the tendency from being too high and reduce the possibility.

[0046] Other algorithms mainly optimize the pheromone evaporation rate and the pheromone generation method. Just like there is a difficult problem, others' optimization is to use a better solution method, while the mutation point probability correction mechanism can make the problem simpler. The principle of the mutation point probability correction strategy is to first discover the mutations that occur frequently, which are the points of sharp turns, and then implement them as a dynamic wall to prohibit passage. After prohibiting passage several times, check the prohibiting effect to decide whether to continue storing the dynamic wall. The prohibiting effect has excellent, good, and poor levels, which are evaluated using formula 5. When the effect is excellent, that is, m iRemain unchanged, the dynamic wall continues to exist m i When =5, the dynamic wall is established, and the effect is good. The dynamic wall is closed but there is still the possibility of opening, that is, m i =0, the effect is poor m i =-10, the dynamic wall is closed more thoroughly and it is difficult to open again. The placement of the dynamic wall can be observed through the red area in the figure. Those red areas are equivalent to the dynamic wall, that is, the mask. If the path is found again after stabilization, it will be simple. Also, as described in the manual, various adverse situations can be defined as mutation points, so that the position of the wall can be adjusted dynamically, making the map of each AGV different, but it can help them exclude wrong options. Generally, pheromones are operated, but in this invention, the selection probability is directly targeted because the ant colony algorithm has a large inertia and strong measures such as adjusting the probability or prohibiting passage points are required to make it find a better solution.

[0047] The most useful part of the mutation point probability correction mechanism is how to judge the quality of the modification, because only by walking through the entire path and standing from a global perspective can the quality of the changed points be found. Add the point set M k , T k The process of sending the mutation points and skipped points existing in the current path needs to be verified and compared after the end. After discovering the signal, all mutation points are traced back. The two points on its left and right are the neighborhood range for path length calculation and comparison, so that the dynamic wall can be placed or removed according to the actual situation. It imitates the communication between ants

[0048] There is no obvious difference in the straight-line probability correction compared with the paper, and there is also pheromone reallocation.

[0049] The linkage of the three mechanisms roughly optimizes the straight-line probability and causes certain problems. The mutation point probability correction solves the problems arising from the straight-line probability optimization and conducts fine optimization. The pheromone reallocation mechanism ensures that the results of the fine optimization can be preserved and not submerged by the interfering pheromones during the initial exploration. Brief Description of the Drawings

[0050] Figure 1 It is the average path planning length graph provided by the example of the present invention;

[0051] Figure 2 It is the optimal path planning length graph provided by the example of the present invention;

[0052] Figure 3 It is the effect graph of the mutation point probability correction strategy provided by the example of the present invention;

[0053] Figure 4 It is the path simulation graph of the algorithm of the present invention provided by the example of the present invention. Detailed Embodiment

[0054] An automatic guided vehicle path planning method based on a selection probability corrected ant colony algorithm, characterized in that an ant colony algorithm automatic guided vehicle path planning model is constructed, a straight-line preference probability correction mechanism is established, a mutation point probability correction strategy is formulated, and a pheromone reallocation rule is designed to reduce the short-distance direction change frequency, reduce the number of mutation points, enhance the attraction of high-quality paths in the later stage, and improve the path smoothness. The specific steps are as follows:

[0055] (1) Construct an ant colony algorithm automatic guided vehicle path planning model:

[0056] Construct an optimization objective function for the ant colony algorithm automatic guided vehicle path planning. The objective function is to minimize the path L k :

[0057]

[0058] where L k is the path length that the k-th ant walks from the starting point to the ending point, q is the total number of ants, with a value of 25, n is the number of nodes, c ik is a node that the k-th ant walks through c i , c (i+1)k is the next node that the k-th ant walks through node c ik , d(c ik , c (i+1)k ) is the Manhattan distance between node c ik and c (i+1)k , z Ф is the starting point, z Ф k is the starting point decision variable, z Ф k has only two values 0 and 1, z Ф k is 1, representing that the k-th ant starts from the starting point, z Ф k is 0, representing that the k-th ant does not start from the starting point, z ψ is the ending point, z ψ k is the ending point decision variable, z ψ k has only two values 0 and 1, z ψ k is 1, representing that the k-th ant ends at the ending point, z ψ k is 0, representing that the k-th ant does not end at the ending point, (c ik , c (i+1)k ) is the edge composed of nodes c ik and c (i+1)k , E is the set of passable edges, c jkis a node c passed by the k-th ant j ;

[0059] Specifically, the path planning model of the ant colony algorithm for automated guided vehicles consists of an outer iteration and an inner loop. The entire algorithm performs 20 rounds of outer iterations, and each round of outer iteration contains 25 inner loops to simulate the process of the ant colony constructing a path; each inner loop corresponds to the process of a single ant constructing a path from the starting point to the ending point; when specifically used, the number of outer iterations and the number of inner loops can be set according to the terrain complexity and the amount of computing resources as required. What is shown in this article is a complete inner loop;

[0060] (2) Establish a straight-line preference probability correction mechanism:

[0061] Calculate the selection probability of each candidate point:

[0062]

[0063] where O is the set of all h candidate points, c x is a candidate point in O, x ∈ [1, h], c i is the current node, c j is the node for which the probability is currently being calculated, j ∈ [1, h], P ij k is the probability that ant k located at node c<� i selects node c j α is the pheromone weight value, with a value of 1, τ ij is the pheromone concentration between nodes c i and c j τ ij α is the α-th power of τ ij β is the heuristic information value weight, with a value of 5, η ij is the heuristic information value between nodes c i and c j η ij β is the β-th power of η ij τ ix is the pheromone concentration between nodes c i and c x τ ix α is the α-th power of τ ix η ix is the heuristic information value between nodes c i and c x η ix β is the β-th power of η ix ;

[0064] Adjust the selection probability of the straight-ahead candidate points:

[0065]

[0066] Among them, P is the set of selection probabilities of all candidate points, and P max is the maximum probability in the set P, and P B is the selection probability of the straight-ahead candidate point, and P B ' is the adjusted selection probability of the straight-ahead candidate point;

[0067] Generate a random number r1 distributed on [0, 1], and select the first P that satisfies P u ≥ r1 from P u (P u is the selection probability of the candidate point c u ), and set the node c u [[ID=2⑥]]corresponding to P u as the next node;

[0068] Specifically, the time point of introducing the straight-ahead preference probability correction mechanism has a significant impact on the performance of the algorithm. If the straight-ahead preference probability correction mechanism intervenes too early, the path will converge to the sub-optimal solution too quickly, reducing the global search ability of the algorithm and making it difficult to jump out of the local optimal solution; if the mechanism is introduced too late, the algorithm has basically converged and the pheromone distribution tends to be stable. At this time, introducing the mechanism is difficult to achieve the expected optimization effect and affects the overall optimization effect; therefore, a reasonable selection of the timing of introducing the straight-ahead preference probability correction mechanism is crucial for improving the stability and global search ability of the algorithm; in the example of the present invention, it is set to be introduced after 5 outer iterations, and the effect can be seen Figure 1 , Figure 2 ; it should be noted that among the three mechanisms, the straight-ahead preference probability correction mechanism must take effect first, and the mutation point probability correction strategy and the pheromone reallocation rule are supplementary optimizations on this mechanism;

[0069] (3) Formulate the mutation point probability correction strategy:

[0070] ① Judge whether W(c i ) * W(c u ) is less than 0 (W(c i ) is the direction vector of the node c i , and W(c u ) is the direction vector of the node c u ), if so, execute the mutation point probability correction strategy; otherwise, execute ⑥;

[0071] ② Enter the current node c i into the mutation point set M and the mutation point set M k of the k-th ant, and record the node c iWhen the number of occurrences of M is m i ;

[0072] ③ Judge m i Whether it is less than or equal to 5. If so, execute ④; otherwise, execute ⑥;

[0073] ④ Enter the node c i into the skip point set T and the skip point set T of the k-th ant k , delete the node c i from M k and O, delete P i from P (P i is the selection probability of the candidate point c i ), record that the number of occurrences of the node c i in T is t i , when t i is equal to five, adjust m i using formula (4):

[0074]

[0075] where f is the change in the excellent rate;

[0076]

[0077] where g is the number of times of y k <Y k ;

[0078] ⑤ Generate another random number r2 distributed in [0,1], and select the first P in P that satisfies P u ≥r2, and set the node c u corresponding to P u as the next node; u ;

[0079] ⑥ Update the next node to the current node, repeat the above process until the ant k reaches the end point, calculate and record the length L of the path that the k-th ant walks from the starting point to the end point k , and increase the pheromone for the current path through the pheromone redistribution rule;

[0080] ⑦ Judge M k ∩T k Whether it is an empty set. If so, the mutation point probability correction strategy ends; otherwise, after finding the end point, execute the following steps. Specifically: sequentially take out the element c k from M, extract the points adjacent to c k from S k , calculate the length of this section of the path and record it as Y k , if calculated again, take the average value, and from T k kSuccessively take out the element c from it k , from S k Extract c k The adjacent points and calculate the length of this section of the path, denoted as y k , if calculated again, take the average value. If y k <Y k , then g = g + 1;

[0081] Specifically, after implementing the straight-line preference probability correction mechanism, a mutation point probability correction strategy is introduced; in the example of the present invention, after 7 outer iterations, the mutation point probability correction strategy is introduced, and the effect can be seen Figure 1 , Figure 2 ; to more accurately evaluate the impact of a certain node on the overall path quality, it is necessary to consider the local structure within the neighborhood of the node; specifically, in the example of the present invention, 2 nodes on each side of the mutation point are selected as the evaluation range, and the width of the neighborhood selection directly affects the sensitivity and robustness of the determination: a smaller neighborhood is more sensitive to local changes and is suitable for capturing fine-grained mutations; while a larger neighborhood can smooth local fluctuations and enhance the stability of the judgment; therefore, the neighborhood width should be set according to actual factors such as path characteristics, environmental complexity, and target smoothness; in addition to defining the nodes with turning angles exceeding the threshold in the path as mutation points in the example of the present invention, other automatic guided vehicles themselves can also be regarded as mutation points and incorporated into the unified mutation point probability correction strategy for processing, so that the automatic guided vehicle can dynamically avoid obstacles and re-plan the path to adapt to the dynamic map environment; it can be passed Figure 3 Observe whether the effect of the mutation point probability correction strategy reaches the expectation Figure 3 In it, the red squares represent the nodes in the set M that are determined to be blocked by the mutation point probability correction strategy (that is, the nodes c corresponding to m i > 5 in M i ), the green squares represent the nodes in the set M that are determined to be passable by the mutation point probability correction strategy (that is, the nodes c corresponding to m i ≤5 in M i ), the black squares represent the obstacles in the environment, and the white squares are the points not in M; if the red squares can fill dead ends and other areas, reducing the difficulty of the algorithm to obtain the optimal solution, it is considered that the introduction time is appropriate; it can be adjusted according to the actual usage scenario and the Figure 3 generated after the initial operation i to adjust the introduction time of the mutation point probability correction strategy, the update strategy of m i (that is, formula (4)), the update strategy trigger threshold (that is, the threshold of t

[0082] (4) Design the pheromone reallocation rule:

[0083] After ant k finds the end point, add pheromone to the path passed this time:

[0084]

[0085] where Δτ ij k is the pheromone concentration added by the k-th ant between nodes c i and c j , ρ is the pheromone evaporation rate, with a value of 0.3;

[0086]

[0087] where S k is the set of all nodes passed by ant k from the starting point to the ending point, Q is the pheromone gain weight, with a value of 1;

[0088] When the number of iterations reaches 5, change Δτ according to formula (8) ij k :

[0089]

[0090] where V is the current number of iterations, and L b is the current shortest path length;

[0091] Specifically, V is the number of outer iterations, the time to start the pheromone reallocation rule, after the straight-line preference probability correction mechanism is implemented. In the example of the present invention, the start time is set to be after 5 outer iterations, and the effect can be seen Figure 1 , Figure 2 ; In practical applications, it can be selected to start between the implementation of the straight-line preference probability correction mechanism and the implementation of the mutation point probability correction strategy. By suppressing the pheromone residue of non-optimal paths and enhancing the pheromone retention of high-quality paths, this rule effectively reduces the interference in the search process and improves the algorithm convergence speed and the quality of the solution;

[0092] Select the smaller value between L b and L k as the new L b value. The next ant starts from the starting point and repeats the above process until the last ant reaches the ending point. Output the optimal path in the form of an ordered coordinate sequence and send it to the automated guided vehicle through the control module. The automated guided vehicle travels to the ending point in sequence according to this path. The simulation result of the example of the present invention is as Figure 4 .

Claims

1. An automatic guided vehicle path planning method based on a selection probability corrected ant colony algorithm, characterized in that Specifically, it includes the following steps: (1) Construct an ant colony algorithm automatic guided vehicle path planning model: Construct an optimization objective function for the path planning of an automated guided vehicle using the ant colony algorithm, where the objective function is to minimize the path L k : Among them, L k is the path length that the k-th ant walks from the starting point to the ending point, q is the total number of ants, with a value of 25, n is the number of nodes, c ik is a node that the k-th ant has walked through, c i ; c (i+1)k is the next node after the k-th ant has walked through node c ik , d(c ik , c (i+1)k ) is the Manhattan distance between node c ik and c (i+1)k , z Ф is the starting point, z Ф k is the starting point decision variable, z Ф k has only two values, 0 and 1, z Ф k being 1 represents that the k-th ant starts from the starting point, z Ф k being 0 represents that the k-th ant does not start from the starting point, z ψ is the ending point, z ψ k is the ending point decision variable, z ψ k has only two values, 0 and 1, z ψ k being 1 represents that the k-th ant ends at the ending point, z ψ k being 0 represents that the k-th ant does not end at the ending point, (c ik , c (i+1)k ) is the edge formed by node c ik and c (i+1)k , E is the set of passable edges, c jk is a node that the k-th ant has walked through, c j ; (2) Establish a straight-line preference probability correction mechanism: Calculate the selection probability of each candidate point: Among them, O is the set of all h candidate points, and c x is a candidate point in O, x ∈ [1, h], and c i is the current node, and c j is the node for which the probability is currently being calculated, j ∈ [1, h], and P ij k is the probability that ant k located at node c i selects node c j α is the pheromone weight value, with a value of 1, and τ ij is the pheromone concentration between node c i and c j τ ij α is the α-th power of τ ij β is the heuristic information value weight, with a value of 5, and η ij is the heuristic information value between node c i and c j η ij β is the β-th power of η ij τ ix is the pheromone concentration between node c i and c x τ ix α is the α-th power of τ ix η ix is the heuristic information value between node c i and c x η ix β is the β-th power of η ix ; Adjust the selection probability of the straight-line candidate point: Among them, P is the set of the selection probabilities of all candidate points, and P max is the maximum probability in the set P, and P B is the selection probability of the straight-ahead candidate point, and P B ' is the adjusted selection probability of the straight-ahead candidate point; Generate a random number r1 distributed on [0, 1], and select the first P from P that satisfies P u ≥r1 of P u (P u is the selection probability of the candidate point c u , and set the node c u corresponding to P u as the next node; (3) Develop a mutation point probability correction strategy: ① Determine whether W(c i ) * W(c u ) is less than 0 (W(c i ) is the direction vector of node c i , and W(c u ) is the direction vector of node c u ). If so, execute the mutation point probability correction strategy; otherwise, execute ⑥; ② Input the current node c i into the mutation point set M and the mutation point set M of the k-th ant k , and record the node c i with the number of occurrences in M being m i ; ③ Determine m i Whether it is less than or equal to 5. If so, execute ④; otherwise, execute ⑥; ④ Enter node c i into the skip point set T and the skip point set T of the k-th ant k , and delete node c i from M k and O, and delete P i from P (P i is the selection probability of candidate point c i ), record that the number of times node c i appears in T is t i , when t i is equal to five, adjust m i using formula (4): Among them, f is the change in the excellent rate; where g is the degree of y k <Y k ; ⑤ Generate another random number r2 distributed on [0, 1], and select the first P from P that satisfies P u ≥ r2 u , and set the node c u corresponding to P u as the next node; ⑥ Update the next node to the current node and repeat the above process until ant k reaches the end point. Calculate and record the length L of the path that the k-th ant has walked from the starting point to the end point. k , and increase the pheromone for the current path through the pheromone reallocation rule; ⑦ Judge M k ∩T k Whether it is an empty set. If so, the mutation point probability correction strategy ends. Otherwise, after finding the end point, execute the following steps. Specifically: sequentially take out the element c k from M k , extract the points adjacent to c k from S k and calculate the length of this section of the path, denoted as Y k . If it is calculated again, take the average value. Sequentially take out the element c k from T k , extract the points adjacent to c k from S k and calculate the length of this section of the path, denoted as y k . If it is calculated again, take the average value. If y k <Y k , then g = g + 1; (4) Design a pheromone reallocation rule: After ant k finds the end point, it increases the pheromone on the path passed this time: where, Δτ ij k is the pheromone concentration added by the k-th ant between nodes c i and c j , ρ is the pheromone evaporation rate, with a value of 0.3; Among them, S k is the set of all nodes that ant k has passed from the starting point to the ending point, Q is the pheromone gain weight, and its value is 1; When the number of iterations reaches 5, change Δτ according to formula (8). ij k : where V is the current iteration number, and L b is the current shortest path length; Select the smaller value of L b and L k as the new value of L b The next ant starts from the starting point and repeats the above process until the last ant reaches the end point. The optimal path is output in the form of an ordered coordinate sequence and sent to the automatic guided vehicle through the control module. The automatic guided vehicle travels to the end point in sequence according to this path.

Images