A multi-stage path planning method and system for AGV based on moss mouse algorithm

The AGV path planning is optimized through the moss mouse algorithm combined with Rich_Moore and TCD* algorithm, which solves the problems of randomness and low node utilization in traditional RRT algorithms, and achieves better path planning and obstacle avoidance effects, improving the transportation efficiency of AGV and Energy consumption utilization rate.

CN118394092BActive Publication Date: 2024-11-08NANJING UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410806694.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-21
Publication Date
2024-11-08
Estimated Expiration
2044-06-21

AI Technical Summary

Technical Problem

Traditional RRT algorithms are too random in AGV path planning and have low node utilization, which leads to the resulting paths being not optimal, making it difficult to effectively avoid obstacles and optimize resource utilization.

Method used

Using a multi-stage path planning method based on the moss mouse algorithm, the initial effective path is generated through the improved random tree algorithm, combined with the Rich_Moore neighbor tracking method to expand the feasible area, use the TCD* algorithm to find the optimal path, and avoid obstacles in real time through the dynamic four-sided window algorithm , and finally adopt flexible high-order Bezier curve optimization path.

Benefits of technology

It improves the smoothness and efficiency of the path, reduces the redundant path nodes, improves the transportation efficiency and energy consumption utilization of AGV, and ensures the safety and stability of the transportation process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118394092B_ABST
    Figure CN118394092B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of logistics AGV path planning, and discloses a multi-stage path planning method and system for AGV based on the moss mouse algorithm, the method comprising: obtaining automatic guided vehicle environmental data, using an improved random tree algorithm to generate an initial valid path; expanding into a feasible area based on the initial valid path, adaptively rasterizing the feasible area, and expanding it in combination with the Rich_Moore neighbor tracking method; finding the optimal path through the TCD* algorithm, and removing redundancy from the path nodes; segmenting the obtained optimal path according to the nodes, using a dynamic four-sided window algorithm for real-time obstacle avoidance in the segments, and further optimizing the path; executing the obtained optimal path. The present invention can improve the path smoothness, reduce large-scale changes, ensure a smooth and worry-free transportation process, improve the algorithm convergence speed, balance the global and local search capabilities, reduce redundant path nodes, improve AGV transportation efficiency, and reduce energy consumption.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of logistics AGV path planning, and in particular to an AGV multi-stage path planning method and system based on a moss mouse algorithm. Background Art

[0002] An Automated Guided Vehicle (AGV) is a transport vehicle equipped with an electromagnetic or optical automatic guidance device that can travel along a specified guidance path and has safety protection and various transfer functions. AGV can replace manual transfer, loading and unloading, and handling of goods, effectively reducing the intensity of manual labor, improving work efficiency, and enhancing the safety of working in some dangerous and complex environments. It is currently widely used in logistics, machinery, electronics, chemicals and other industries. AGV path planning and obstacle avoidance algorithms are key technologies for realizing AGV autonomous navigation, which determine whether AGV can complete tasks efficiently and safely in complex environments. In recent years, it has become one of the important research hotspots in the AGV field.

[0003] AGV path planning refers to searching for an optimal or approximately optimal running path from the starting state to the target state for the AGV in a specific application scenario according to specific performance indicators. The traditional RRT algorithm is too random in planning, the node utilization rate is low, and the resulting path is not optimal. The path planning algorithm needs to consider various constraints, such as static obstacles, dynamic obstacles, and regional restrictions. Therefore, the obstacle avoidance algorithm is the core component of the path planning algorithm, which is used to judge, predict and avoid collisions between AGV and obstacles, thereby ensuring the safety and reliability of the AGV during operation. The goal of path planning can be to maximize the efficiency of the AGV, reduce travel time, reduce energy consumption, optimize resource utilization, etc. In order to achieve the above functional requirements, the present invention proposes an AGV multi-stage path planning method based on random + heuristic algorithm, which takes into account multiple requirements such as efficient calculation, threat avoidance, and action optimization, and has great benefits. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides an AGV multi-stage path planning method based on the moss rat algorithm to solve the problems of excessive randomness in the planning of the traditional RRT algorithm, low node utilization and non-optimal path.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an AGV multi-stage path planning method based on the moss mouse algorithm, comprising:

[0008] Obtain the environmental data of the automated guided vehicle, including the plane length and width, obstacle locations, and start and end locations, and use the improved random tree algorithm to generate an initial valid path;

[0009] Expanding the feasible area based on the initial valid path, adaptively rasterizing the feasible area, and expanding it in combination with the Rich-Moore neighbor tracking method;

[0010] Find the optimal path in the feasible area through the TCD* algorithm and remove redundant nodes on the path;

[0011] The optimal path is segmented by nodes, and the dynamic four-sided window algorithm is used to avoid obstacles in real time to further optimize the path.

[0012] The obtained optimal path is executed so that the automatic guided vehicle reaches the destination along the optimal path.

[0013] As a preferred solution of the AGV multi-stage path planning method based on the moss mouse algorithm described in the present invention, generating an initial valid path includes:

[0014] The random dynamic target node sampling rate is set under the constraints of the map environment parameters, and the target node sampling rate is adjusted according to the working environment map road conditions, which is expressed as:

[0015]

[0016] Among them, pro goal Indicates the target node sampling rate, pro random The random number generated by the random function, num dir Indicates the number of obstacles that the line connecting the AGV's current position and the target node passes through, Ppt bac Indicates the percentage of obstacle area in the entire map, d nos Indicates the distance the AGV has traveled for this task, d togoal Indicates the straight-line distance between the current position of the AGV and the target node;

[0017] Optimize the obstacle ratio in the environment and set the step size, expressed as:

[0018] Step curr =(2-(obs / S) ln obs )Step ini

[0019] Among them, Step iniIndicates the initial step length, Step curr represents the current step length, S represents the area of ​​the circle with the diameter of the line connecting the tree node closest to the random point and the target point, and obs represents the number of obstacles in the circle;

[0020] Select the next sampling point and determine whether the target point is covered by the random tree. If it is covered, a valid random tree path is generated. Otherwise, repeat the steps and reset the random dynamic target node sampling rate. The sampling method is expressed as:

[0021]

[0022] Among them, P goal represents the target point, P nearest Represents the distance P on the existing random tree rand The nearest node, P rand represents a random sampling point within the map range, P new Represents a new node to which the random tree is extended.

[0023] As a preferred solution of the AGV multi-stage path planning method based on the moss mouse algorithm described in the present invention, wherein: the expanded feasible area includes:

[0024] Rasterize the map according to the appropriate size and cut the chart into several small squares of equal size. The grid size requirement is expressed as:

[0025]

[0026] Among them, Size grid Indicates the grid side length, size indicates the optional side length, Z is a positive integer set, lg indicates the length of the rectangular map, wd indicates the width of the rectangular map, step curr Indicates the real-time step size of the random tree;

[0027] Through Rich_Moore's cellular neighbor automaton, the effective path obtained by the random tree is expanded outward to obtain a feasible area. The cellular automaton model is expressed as:

[0028] A=(Dm moore ,State,Set,f)

[0029] Among them, A represents the cellular automaton model, Dm moore represents the parameter dimension of the cell evolution rule, Sate represents the finite discrete state of the cell, Set represents the set of all potential neighboring cells in the grid neighborhood where the current path node is located, and f is the evolution rule function value.

[0030] As a preferred solution of the AGV multi-stage path planning method based on the moss mouse algorithm described in the present invention, the evolution rule f scores each neighboring cell grid. If the neighboring cell is an obstacle grid, it is immediately eliminated, and the 12 non-obstruction neighboring cells with the highest scores are selected as the current path node expansion area. When calculating the potential neighboring cell evolution rule function, the evolution rule function of the secondary neighboring cell is called, and the evolution rule function is expressed as:

[0031]

[0032] in, represents potential neighbor v i The evolution rule function value of represents the barrier-free property of each potential neighboring cell, v j =(x j ,v j ) represents each potential neighbor v i Each potential neighbor cell is the second-level potential neighbor cell, Set_of_v i Represents the grid v where the potential neighbor cell currently under consideration is located i =(x i ,y i ) All potential neighbors v in the neighborhood j =(x j ,y j ), represents the second-level potential neighbor v j In v j The number of levels in the neighboring cell structure, P goal =(x goal ,y goal ) is another form of expression for the target point. Indicates v i Expansion Direction Coefficient.

[0033] As a preferred solution of the AGV multi-stage path planning method based on the moss mouse algorithm described in the present invention, the D* algorithm improved by the TCD* algorithm coordinates the threat cost and the transportation cost in the feasible area, and secondary plans the optimal path for the current transportation task, including:

[0034] Create two lists, OPEN list and CLOSE list, and put the target node P goal Add to OPEN list, OPEN list means to store the locations reachable from the current point, CLOSElist means to store the points reached during the search, and create WARN list at the same time;

[0035] Perform path addressing and check whether the OPEN list is empty;

[0036] If there is no node in the OPEN list, the algorithm terminates and path planning fails;

[0037] Otherwise, check whether the node with the smallest cost estimate in the OPEN list is the starting node. If the smallest node is the starting node, the path planning is successful and the execution continues;

[0038] If the smallest node is not the starting node, add the node to the CLOSElist list and check whether there are obstacles in the first-level neighborhood of the node;

[0039] If there are obstacle nodes, add the non-expandable nodes in the second-level neighborhood to the WARNlist, add the nearby expandable nodes that are not in the OPEN list and WARN list to the OPEN list, and repeat the path addressing operation;

[0040] Perform a backtracking check. If the environmental change affects the nodes along the path, perform local re-planning, add the previous node of the changed node to the OPENlist list, and return to the path addressing operation. Otherwise, continue to execute and the path is determined and generated.

[0041] As a preferred solution of the AGV multi-stage path planning method based on the moss mouse algorithm described in the present invention, the path addressing optimizes the heuristic function of the rasterization heuristic algorithm, adds a threat cost function to the cost function, and optimizes and adjusts the parameters of the heuristic term and the non-threat cost term of the cost function, which is expressed as:

[0042] f(n)=h * (n)+g(n)

[0043] h * (n) = h(n) + (1 + Ppt bac )s(n)

[0044] h(n)=|x n -x goal |+|y n -y goal |

[0045] Where f(n) represents the grid side length, h * (n) represents the threat cost function, h(n) is the non-threat cost term of the cost function, represents the cost value between the node and the target point, s(n) represents the threat cost function, Ppt bac represents the percentage of obstacle area in the entire map, and g(n) represents the function heuristic term;

[0046] The threat cost function s(n) is expressed as:

[0047]

[0048] Among them, s(n) represents the threat cost function of the current path node n, α is a scaling factor, (x k ,y k ) represents the center coordinate of obstacle k, (x n ,y n ) represents the coordinates of the current path node n, a k represents the effective coefficient of the threat cost. If there is no obstacle in the two-level prediction neighborhood of the current path node, then set a k is 0, otherwise it is set to 1, s k is the threat coefficient brought by obstacle k to the node n passing through the current path, and M is the obstacle set in the two-level predicted neighborhood of the current path node;

[0049] By using the Euclidean distance as a heuristic function, a stability weighted root sw of not less than 1 is introduced. The heuristic term g(n) and the stability weighted root sw are expressed as:

[0050]

[0051] Among them, g(n) represents the heuristic term of the current path node n, sw represents a stability weighted root, and the value is the number of obstacles num in the neighborhood of all nodes before the current path node n carr The standard deviation σ plus 1, It represents the average number of obstacles in the neighborhood of all nodes before the current path node n. represents the number of obstacles in the neighborhood of each node before the current path node n, (x start ,y start ) represents the starting coordinates of the path, x n ,y n ) represents the coordinates of the current path node n.

[0052] As a preferred solution of the AGV multi-stage path planning method based on the moss mouse algorithm described in the present invention, further optimizing the path includes:

[0053] The path segmentation selects the grouping interval according to the path tortuosity and the number of path nodes. The segmentation interval determination formula is expressed as:

[0054]

[0055]

[0056] Among them, od represents the group spacing of the path segment, D represents the approximate number of path node spacing, pro sharp represents the probability of sharp bends at nodes in the path, Indicates the turning angle at each node, State node Indicates the zigzag state of each node. If the turning angle is greater than π / 3, it is a sharp bend node, otherwise it is a non-sharp bend node. Sharp indicates that the node is in a zigzag state, and normal indicates that the node is in a flat state.

[0057] By using a flexible high-order Bezier curve equation with the grouping interval od as the order, the smoothness of each segmented path is optimized to generate an optimal path.

[0058] In a second aspect, the present invention provides an AGV multi-stage path planning system based on the moss mouse algorithm, comprising:

[0059] The data collection and planning module is used to obtain the environmental data of the AGV, including the plane length and width, obstacle locations, and start and end locations, simplify the route performance, and use the improved random tree algorithm to generate the initial valid path;

[0060] A first processing module is used to expand the feasible area based on the initial valid path, adaptively rasterize the feasible area, and expand it in combination with the Rich-Moore neighbor tracking method;

[0061] The second processing module is used to find the optimal path in the feasible area through the TCD* algorithm and remove redundant nodes on the path;

[0062] The optimization module is used to segment the optimal path according to the nodes. The segmentation uses the dynamic four-sided window algorithm to avoid obstacles in real time and further optimize the path.

[0063] The execution module is used to execute the obtained optimal path so that the automatic guided vehicle reaches the destination along the optimal path.

[0064] In a third aspect, the present invention provides an electronic device, comprising:

[0065] Memory and processor;

[0066] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the AGV multi-stage path planning method based on the moss mouse algorithm are implemented.

[0067] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the AGV multi-stage path planning method based on the moss mouse algorithm.

[0068] Compared with the prior art, the present invention has the following beneficial effects: the present invention introduces an improved D* algorithm in combination with other dynamic obstacle avoidance algorithms, performs target constraints on the randomness of RRT, efficiently obtains a better path, adds adaptive field of view, step size, hazard factor and target point sampling rate to the RRT algorithm, and uses Rich_Moore's cellular automaton method to expand the feasible area around the feasible path planned by the improved RRT algorithm, rasterize the feasible area, find the shortest route in the feasible area through the improved D* algorithm, and finally divide the obtained global path into segments and use a flexible high-order Bezier curve equation method to smooth the path, improve the path smoothness, reduce large-scale changes, ensure a smooth and worry-free transportation process, improve the algorithm convergence speed, balance the global and local search capabilities, reduce redundant path nodes, improve AGV transportation efficiency, and reduce energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0070] Figure 1 It is a schematic diagram of the overall process of the AGV multi-stage path planning method based on the moss mouse algorithm according to an embodiment of the present invention;

[0071] Figure 2 A flow chart of an AGV two-stage path planning method of an AGV multi-stage path planning method based on a moss mouse algorithm according to an embodiment of the present invention;

[0072] Figure 3 A flow chart of a random tree addressing method of a multi-stage path planning method for AGV based on a moss mouse algorithm according to an embodiment of the present invention;

[0073] Figure 4 A flowchart of rasterization and expansion of feasible areas of a multi-stage path planning method for AGV based on the moss mouse algorithm according to an embodiment of the present invention;

[0074] Figure 5 A comparison diagram of the improved cellular neighbor structure of the AGV multi-stage path planning method based on the moss mouse algorithm according to an embodiment of the present invention;

[0075] Figure 6 A Rich_Moore improved cellular neighbor structure diagram of the AGV multi-stage path planning method based on the moss mouse algorithm according to an embodiment of the present invention;

[0076] Figure 7A TCD* algorithm flow chart of an AGV multi-stage path planning method based on a moss mouse algorithm according to an embodiment of the present invention;

[0077] Figure 8 A flow chart of path smoothness optimization using a flexible high-order Bezier curve equation method of a multi-stage path planning method for AGVs based on a moss mouse algorithm according to an embodiment of the present invention;

[0078] Fig. 9 This is a diagram showing the effect of a traditional RRT algorithm path planning according to an embodiment of the present invention;

[0079] Fig.10 This is a path planning effect diagram of a weight constraint-guided random path exploration tree method of an AGV multi-stage path planning method based on a moss mouse algorithm according to an embodiment of the present invention;

[0080] Fig.11 A schematic diagram of a traditional RRT algorithm planning according to an embodiment of the present invention;

[0081] Fig.12 This is a schematic diagram of the process and planning results of an AGV multi-stage path planning method based on the moss mouse algorithm described in an embodiment of the present invention. DETAILED DESCRIPTION

[0082] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0083] Example 1

[0084] Reference Figure 1-Figure 8 , which is an embodiment of the present invention, provides an AGV multi-stage path planning method based on the moss mouse algorithm. Based on the application of intelligent sampling in path planning, combined with the dynamic rasterization evaluation function method and the curve equation optimization method, the AGV can have both expansion speed and real-time obstacle avoidance capabilities during the addressing process. The process is as follows Figure 1-Figure 2 As shown, including:

[0085] S100, obtains the environmental data of the automated guided vehicle, including the length and width of the plane, the location of obstacles, and the starting and ending locations, simplifies the transportation performance, and uses the improved random tree algorithm to generate an initial valid path.

[0086] It should be noted that in order to ensure that the automatic guided vehicle (AGV) is subject to guidance constraints while rapidly expanding the path and avoid wasting computing resources, a dynamic target node sampling rate combining randomness and weight is used to guide the random tree expansion sampling direction to a certain extent, such as Figure 3 As shown, the specific steps are as follows:

[0087] The random dynamic target node sampling rate is set under the constraints of the map environment parameters. According to the specific working environment map road environment parameters, the target node sampling rate is adjusted to guide the direction of the random exploration path to improve the addressing efficiency and optimality, which is expressed as:

[0088]

[0089] Among them, pro goal Indicates the target node sampling rate, pro random The random number generated by the random function, num dir Indicates the number of obstacles that the line connecting the AGV's current position and the target node passes through, Ppt bac Indicates the percentage of obstacle area in the entire map, d nos Indicates the distance the AGV has traveled in this mission, d togoal Indicates the straight-line distance between the current position of the AGV and the target node;

[0090] Optimize the obstacle ratio in the environment and set the step length. That is, during the search process, the step length varies with the ratio of obstacles to the circular area. The larger the ratio, the more obstacles there are in the circular area and the smaller the step length. On the contrary, the larger the step length, the faster the target point can be searched. The specific formula is:

[0091] Step curr =(2-(obs / S) ln obs )Step ini

[0092] Among them, Step ini Indicates the initial step length, Step curr represents the current step length, S represents the area of ​​the circle with the diameter of the line connecting the tree node closest to the random point and the target point, and obs represents the number of obstacles in the circle;

[0093] Select the next sampling point and determine whether the target point is covered by the random tree. If it is covered, a valid random tree path is generated. Otherwise, repeat the steps and reset the random dynamic target node sampling rate. The sampling method is expressed as:

[0094]

[0095] Among them, P goal represents the target point, P nearestRepresents the distance P on the existing random tree rand The nearest node, P rand represents a random sampling point within the map range, P new Represents a new node to which the random tree is extended.

[0096] S200, expanding into a feasible region based on the initial valid path, adaptively rasterizing the feasible region, and expanding it in combination with the Rich-Moore neighbor tracking method.

[0097] It should be noted that, in order to further explore the optimal path, the present invention generates a feasible area by expanding the effective path on the rasterized map based on the grid method and the Rich-Moore type cellular neighbor automaton, so as to optimize the path within a certain limit later.

[0098] Further, such as Figure 4 As shown, the extended feasible region includes,

[0099] Rasterize the map according to the appropriate size, that is, cut the map into several small squares of equal size. The grid size requirement is expressed as:

[0100]

[0101] Among them, Size grid Indicates the length of the grid side, size indicates the optional side length, Z is a positive integer set, lg indicates the length of the rectangular map, if the shape is irregular, it is completed as a rectangle, if it is not a positive integer, then take the largest positive integer not greater than the length, wd indicates the width of the rectangular map, if the shape is irregular, it is completed as a rectangle, if it is not a positive integer, then take the largest positive integer not greater than the width, step curr Indicates the real-time step size of the random tree;

[0102] From the characteristics of the random tree, we know that this path is not the shortest. There must be a path shorter than the initial path in the surrounding area. Through Rich_Moore's cellular neighbor automaton, the effective path obtained by the random tree is expanded outward to obtain a feasible area. The cellular automaton model is expressed as:

[0103] A=(Dm moore ,State,Set,f)

[0104] Among them, A represents the cellular automaton model, Dm moore represents the parameter dimension of the cellular evolution rule, State represents the finite discrete state of the cell, and the cell can be in binary form {0,1}, where 0 represents that there is no obstacle in the first layer of the cellular neighborhood structure of the grid node, and 1 represents that there is an obstacle in the first layer of the cellular neighborhood structure of the grid node. Set represents all potential neighboring cells v in the neighborhood of the grid v0 = (x0, y0) where the current path node is located. i=(x i ,y i ), recorded as:

[0105]

[0106] Each node has a total of 20 neighbors, such as Figure 5 , Figure 6 As shown, there are 16 directions that can be reached within 2 steps starting from the cell. In order to limit the diffusion range and reduce the amount of quadratic programming calculations, the step size is limited to 1 step in the four directions of upper left, upper right, lower left, and lower right. f is the value of the evolution rule function.

[0107] It should be noted that the evolution rule f comprehensively considers factors such as the barrier-free properties of each potential neighbor, the weighted sum of the evolution rule function values ​​of the secondary neighbor, and the penalty of the distance to the target point, and scores each neighbor grid. If the neighbor is an obstacle grid, it will be eliminated immediately; then the 12 non-obstruction neighbor grids with the highest scores are selected as the expansion area of ​​the current path node. This method involves recursive thinking in the calculation, that is, the evolution rule function of the secondary neighbor needs to be called in the process of calculating the evolution rule function of the potential neighbor. The evolution rule function is expressed as:

[0108]

[0109] in, represents potential neighbor v i The evolution rule function value of represents the barrier-free property of each potential neighboring cell, v j =(x j ,y j ) represents each potential neighbor v i Each potential neighbor cell is the second-level potential neighbor cell, Set_of_v i Represents the grid v where the potential neighbor cell currently under consideration is located i =(x i ,y i ) All potential neighbors v in the neighborhood j =(x j ,y j ), represents the second-level potential neighbor v j In v j The number of layers in the neighboring cell structure is 1 for the closest layer and 2 for the outer layer. goal =(x goal ,y goal ) is another expression of the target point. Indicates v i Expansion Direction Coefficient.

[0110] S300, using the TCD* algorithm to find the optimal path in the feasible area and remove redundant nodes from the path.

[0111] In order to quickly generate the optimal path in the feasible area while ensuring safety, this embodiment, after expanding the feasible area from the effective path, uses an improved D* algorithm based on the TCD* algorithm to coordinate the threat cost and the transportation cost in the feasible area, and obtains the optimal path for the current transportation task through quadratic programming, such as Figure 7 As shown, including,

[0112] Create two lists, OPEN list and CLOSE list, and put the target node P goal Add to the OPEN list. The OPEN list stores the locations that can be reached from the current point. It records all possible nodes. The CLOSE list stores the points that have been reached during the search. It avoids repeated searches of nodes that have been searched during the search process. At the same time, a WARN list is created to save the non-expandable nodes in the feasible area during the addressing process.

[0113] Perform path addressing and check whether the OPEN list is empty;

[0114] If there is no node in the OPEN list, the algorithm terminates and path planning fails;

[0115] Otherwise, check whether the node with the smallest cost estimate in the OPEN list is the starting node. If the smallest node is the starting node, the path planning is successful and the execution continues;

[0116] If the smallest node is not the starting node, add the node to the CLOSElist list and check whether there are obstacles in the first-level neighborhood of the node;

[0117] If there are obstacle nodes, add the non-expandable nodes in the second-level neighborhood to the WARN list, add the nearby expandable nodes that are not in the OPEN list and WARN list to the OPEN list, and repeat the path addressing operation;

[0118] Perform a backtracking check. If the environmental change affects the nodes along the path, perform local re-planning, add the previous node of the changed node to the OPEN list, and return to the path addressing operation. Otherwise, continue to execute and the path is determined and generated.

[0119] Furthermore, the path addressing step optimizes the heuristic functions of common rasterization heuristic algorithms such as A* and D*, adds a threat cost function to the cost function, and optimizes the parameters of the heuristic term and the non-threat cost term of the cost function to further avoid obstacles and improve the addressing efficiency. The specific optimization parts are as follows:

[0120] f(n)=h * (n)+g(n)

[0121] h * (n) = h(n) + (1 + Ppt bac )s(n)

[0122] h(n)=|x n -x goal |+|y n -y goal |

[0123] Where f(n) represents the grid side length, h * (n) represents the threat cost function, h(n) is the non-threat cost term of the cost function, which represents the cost value between the node and the target point, that is, the Manhattan distance, s(n) represents the threat cost function, Ppt bac represents the percentage of obstacle area in the entire map, and g(n) represents the function heuristic term;

[0124] The threat cost function s(n) is expressed as:

[0125]

[0126] Among them, s(n) represents the threat cost function of the current path node n, α is a scaling factor, the square root operation is removed from the Euclidean distance term for computational efficiency, and the scaling factor is used to appropriately reduce the algebraic value in the denominator, (x k ,y k ) represents the center coordinate of obstacle k, (x n ,y n ) represents the coordinates of the current path node n, a k represents the effective coefficient of the threat cost. If there is no obstacle in the two-level prediction neighborhood of the current path node, then set a k is 0, otherwise it is set to 1, s k is the threat coefficient brought by obstacle k to the node n passing through the current path, and M is the obstacle set in the two-level predicted neighborhood of the current path node.

[0127] It should be noted that, considering that the path node expansion in the TCD* algorithm is a multi-level structure, the Manhattan distance and diagonal distance are greatly affected by a single dimension. If the difference in one dimension is large, the heuristic function value becomes larger, ignoring the influence of other dimensions.

[0128] By using the Euclidean distance as a heuristic function and introducing a stability weighted root sw of not less than 1, the cost from the current node to the target point is increased in the proportion of the total cost, which helps to increase the search depth and prevent the algorithm from falling into the local optimum. The heuristic term g(n) and the stability weighted root sw are expressed as:

[0129]

[0130] Among them, g(n) represents the heuristic term of the current path node n, sw represents a stability weighted root, and the value is the number of obstacles num in the neighborhood of all nodes before the current path node n carr The standard deviation σ of is added by 1 to adjust the influence of the heuristic term. It represents the average number of obstacles in the neighborhood of all nodes before the current path node n. represents the number of obstacles in the neighborhood of each node before the current path node n, x start ,y start ) represents the starting point coordinates of the path, (x n ,y n ) represents the coordinates of the current path node n.

[0131] S400, segmenting the obtained optimal path according to nodes, using a dynamic four-sided window algorithm to avoid obstacles in real time, and further optimizing the path.

[0132] It should be noted that in order to address the undesirable phenomena such as AGV jamming and cargo falling caused by overly tortuous paths in daily sorting and transportation work, this embodiment divides the established optimal path into segments, and uses flexible-order Bezier curves to optimize overly tortuous and sharp sections, improve the path smoothness, reduce sharp turns and large turns, and avoid transportation accidents.

[0133] Further, such as Figure 8 As shown in the flow chart, further optimization paths include:

[0134] Path segmentation selects the appropriate grouping spacing, that is, the order of the later Bezier curve, according to the path tortuosity and the number of path nodes. The segmentation interval determination formula is expressed as:

[0135]

[0136] Among them, od represents the grouping distance of the path segment, that is, how many node distances are involved in each segment, D represents the approximate number of path node distances, pro sharp represents the probability of sharp bends at nodes in the path, Indicates the turning angle at each node, State nodeIndicates the bending state of each node. If the turning angle is greater than π / 3, it is a sharp bend node, otherwise it is a non-sharp bend node;

[0137] Through the flexible high-order Bezier curve equation with the group spacing od as the order, the smoothness of each segmented path is optimized to generate the optimal path.

[0138] Furthermore, the entire path is set with num node Control points (i.e. path nodes) After segmentation, several segment intervals {P m·od ,P m·od+1 ,…,P (m+1)od}, the number of intervals is That is, the last interval length is allowed to be less than the given spacing, and the flexible high-order Bezier curve equation is expressed as:

[0139]

[0140] Among them, B(t) is the flexible high-order Bezier curve equation, num node is the total number of nodes in the path, k is the node traversal number, from 0 to num node -1, m is the segment interval number, od is the grouping spacing of the path segments in the above steps, and is also the order of each path segment curve, P k are the coordinates of the node numbered k, and t is a parameter with a value range of [0,1], which is related to the position of the current node in the segment interval.

[0141] S500, executing the obtained optimal path, so that the automatic guided vehicle AVG reaches the destination along the optimal path.

[0142] It should be noted that the present invention aims to solve the problems of excessive randomness in planning of traditional RRT algorithm, low node utilization and non-optimal path. The improved D* algorithm is combined with other dynamic obstacle avoidance algorithms to impose target constraints on the randomness of RRT and efficiently obtain a better path. First, an adaptive field of view, step size, hazard factor and target point sampling rate are added to the RRT algorithm; then, a cellular automaton method named Rich_Moore is proposed to expand the feasible area around the feasible path planned by the improved RRT algorithm, rasterize the feasible area, and find the shortest route in the feasible area through the improved D* algorithm; finally, the obtained global path is segmented and the flexible high-order Bezier curve equation method is used to smooth the path, improve the path smoothness, reduce large-scale changes, and ensure a smooth and worry-free transportation process.

[0143] The present invention is applied to AGV path planning in logistics warehousing, which can improve the algorithm convergence speed, balance the global and local search capabilities, reduce path redundant nodes, improve AGV transportation efficiency, and significantly reduce energy consumption.

[0144] The above is a schematic scheme of a multi-stage path planning method for AGV based on the moss rat algorithm in this embodiment. It should be noted that the technical scheme of the multi-stage path planning system for AGV based on the moss rat algorithm and the technical scheme of the multi-stage path planning method for AGV based on the moss rat algorithm mentioned above belong to the same concept. The details of the technical scheme of the multi-stage path planning system for AGV based on the moss rat algorithm in this embodiment that are not described in detail can all be referred to the description of the technical scheme of the multi-stage path planning method for AGV based on the moss rat algorithm mentioned above.

[0145] The AGV multi-stage path planning system based on the moss mouse algorithm in this embodiment includes:

[0146] The data collection and planning module is used to obtain the environmental data of the AGV, including the plane length and width, obstacle locations, and start and end locations, simplify the route performance, and use the improved random tree algorithm to generate the initial valid path;

[0147] The first processing module is used to expand the feasible area based on the initial valid path, adaptively rasterize the feasible area, and expand it in combination with the Rich-Moore neighbor tracking method;

[0148] The second processing module is used to find the optimal path in the feasible area through the TCD* algorithm and remove redundant nodes on the path;

[0149] The optimization module is used to segment the optimal path according to the nodes. The segmentation uses the dynamic four-sided window algorithm to avoid obstacles in real time and further optimize the path.

[0150] The execution module is used to execute the obtained optimal path so that the automatic guided vehicle reaches the destination along the optimal path.

[0151] This embodiment further provides an electronic device, which is applicable to the case of multi-stage path planning of AGV based on the moss mouse algorithm, including:

[0152] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the AGV multi-stage path planning method based on the moss mouse algorithm as proposed in the above embodiment.

[0153] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the AGV multi-stage path planning method based on the moss mouse algorithm as proposed in the above embodiment is implemented.

[0154] The storage medium proposed in this embodiment and the AGV multi-stage path planning method based on the moss rat algorithm proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0155] Through the above description of the implementation method, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccessMemory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform the methods of various embodiments of the present invention.

[0156] Example 2

[0157] Refer to Table 1-Table 2 and Figure 9-12 , which is an embodiment of the present invention, provides an AGV multi-stage path planning method based on the moss mouse algorithm. In order to verify its beneficial effects, the comparison results of two schemes are provided.

[0158] like Figure 9-10 , simulates a 21*21 indoor storage space (the actual fully usable area is 20*20, with one row and one column reserved as incomplete rows and columns to simulate the edge of the storage in the real scene). The AGV needs to start from the storage point (0, 0) to reach the specified P goal =(x goal ,y goal ), and in this embodiment, it is set to (3.5, 10). A simulation comparison experiment is conducted on a weight constraint-guided random path exploration tree method, and the comparison results are shown in Table 1.

[0159] Table 1: Comparison of path planning effects between the method of the present invention and the traditional method.

[0160]

[0161] From the comparison in Table 1, it can be seen that although the effective path obtained by the weight constraint-guided random path exploration tree method proposed in the present invention is similar to that of the traditional method, since the expansion direction is restricted by various parameters and weight coefficients, a lot of unnecessary waste branches are reduced, the cost of exploration is reduced, and the efficiency of obtaining effective paths is also higher. The random tree exploration direction is effectively converged, the calculation cost is saved, and the exploration efficiency is improved.

[0162] like Figure 11-Figure 12 , simulating a 30*30 complex indoor storage space, the AGV needs to start from the storage point (5, 2) to reach the specified P goal =(x goal ,y goal ), and this embodiment is set to (21, 17). The TCD* algorithm and flexible multi-order Bezier curve can be used in the feasible area of ​​RRT path extension to achieve the optimal route with both low cost and high smoothness. The planning time, path length, and number of inflection points of various methods obtained through simulation at each stage are shown in Table 2.

[0163] Table 2: Comparative data of each stage of the optimal route.

[0164]

[0165]

[0166] It can be seen from Table 2 that the fusion algorithm described in the present invention can shorten the path length by nearly 50% without affecting the efficiency, effectively reduce the number of path turning points, and improve the path smoothness.

[0167] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A multi-stage path planning method for AGV based on the moss mouse algorithm, It is characterized in that include: Obtain the environmental data of the AGV, including the plane length and width, obstacle locations, and start and end locations, and use the improved random tree algorithm to generate an initial valid path; Expanding the feasible area based on the initial valid path, adaptively rasterizing the feasible area, and expanding it in combination with the Rich-Moore neighbor tracking method; Find the optimal path in the feasible area through the TCD* algorithm and remove redundant nodes on the path; The optimal path is segmented by nodes, and the dynamic four-sided window algorithm is used to avoid obstacles in real time to further optimize the path. Execute the obtained optimal path so that the automatic guided vehicle reaches the destination along the optimal path; Generating an initial valid path includes: The random dynamic target node sampling rate is set under the constraints of the map environment parameters, and the target node sampling rate is adjusted according to the working environment map road condition environment parameters, which is expressed as: Among them, pro goal Indicates the target node sampling rate, pro random The random number generated by the random function, num dir Indicates the number of obstacles that the line connecting the AGV's current position and the target node passes through, Ppt bac Indicates the percentage of obstacle area in the entire map, d nos Indicates the distance the AGV has traveled for this task, d togoal Indicates the straight-line distance between the current position of the AGV and the target node; Optimize the obstacle ratio in the environment and set the step size, expressed as: Step curr =(2-(obs / S) lnobs )Step ini Among them, Step ini Indicates the initial step length, Step curr represents the current step length, S represents the area of ​​the circle with the diameter of the line connecting the tree node closest to the random point and the target point, and obs represents the number of obstacles in the circle; Select the next sampling point and determine whether the target point is covered by the random tree. If it is covered, a valid random tree path is generated. Otherwise, repeat the steps and reset the random dynamic target node sampling rate. The sampling method is expressed as: Among them, P goal represents the target point, P nearest Represents the distance P on the existing random tree rand The nearest node, P rand represents a random sampling point within the map range, P new Represents new nodes for random tree expansion; The feasible areas for expansion include: Rasterize the map according to the appropriate size and cut the chart into several small squares of equal size. The grid size is expressed as: Among them, Size grid Indicates the grid side length, size indicates the optional side length, Z is a positive integer set, lg indicates the length of the rectangular map, wd indicates the width of the rectangular map, step curr Indicates the real-time step size of the random tree; Through Rich_Moore's cellular neighbor automaton, the effective path obtained by the random tree is expanded outward to obtain a feasible area. The cellular automaton model is expressed as: A=(Dm moore ,State,Set,f) Among them, A represents the cellular automaton model, Dm moore represents the parameter dimension of the cell evolution rule, State represents the finite discrete state of the cell, Set represents the set of all potential neighboring cells in the grid neighborhood where the current path node is located, and f is the evolution rule function value; The evolution rule f scores each neighboring cell grid. If the neighboring cell is an obstacle grid, it is immediately eliminated, and the 12 non-obstruction neighboring cells with the highest scores are selected as the expansion area of ​​the current path node. When calculating the potential neighboring cell evolution rule function, the evolution rule function of the secondary neighboring cell is called. The evolution rule function is expressed as: in, represents potential neighbor v i The evolution rule function value of represents the barrier-free property of each potential neighboring cell, v j =(x j ,y j ) represents each potential neighbor v i Each potential neighbor cell is the second-level potential neighbor cell, Set_of_v i Represents the grid v where the potential neighbor cell currently under consideration is located i =(x i ,y i ) All potential neighbors v in the neighborhood j =(x j ,y j ), represents the second-level potential neighbor v j In v j The number of levels in the neighboring cell structure, P goal =(x goal ,y goal ) is another expression of the target point. Indicates v i Expansion directional coefficient; The D* algorithm improved by the TCD* algorithm coordinates the threat cost and transportation cost in the feasible area, and secondary planning is performed to find the best path for the current transportation task, including: Create two lists, OPEN list and CLOSE list, and put the target node P goal Add to OPEN list, OPENlist means to store the locations reachable from the current point, CLOSE list means to store the points reached during the search, and create WARNlist list at the same time; Perform path addressing and check whether the OPEN list is empty; If there is no node in the OPEN list, the algorithm terminates and path planning fails; Otherwise, check whether the node with the smallest cost estimate in the OPENlist is the starting node. If the smallest node is the starting node, the path planning is successful and the execution continues; If the smallest node is not the starting node, add the node to the CLOSElist list and check whether there are obstacles in the first-level neighborhood of the node; If there are obstacle nodes, add the non-expandable nodes in the second-level neighborhood to the WARN list, add the nearby expandable nodes that are not in the OPENlist and WARN list to the OPEN list, and repeat the path addressing operation; Perform backtracking check. If the environment change affects the nodes along the path, perform local re-planning, add the previous node of the changed node to the OPENlist list, and return to the path addressing operation. Otherwise, continue to execute and the path is determined and generated. The path addressing optimizes the heuristic function of the rasterization heuristic algorithm, adds a threat cost function to the cost function, and optimizes and adjusts the parameters of the heuristic term and the non-threat cost term of the cost function, which is expressed as: f(n)=h * (n)+g(n) h * (n)=h(n)+(1+Ppt bac )s(n) h(n)=|x n -x goal |+|y n -y goal | Where f(n) represents the grid side length, h * (n) represents the threat cost function, h(n) is the non-threat cost term of the cost function, represents the cost value between the node and the target point, s(n) represents the threat cost function, Ppt bac represents the percentage of obstacle area in the entire map, and g(n) represents the function heuristic term; The threat cost function s(n) is expressed as: Among them, s(n) represents the threat cost function of the current path node n, α is a scaling factor, (x k ,y k ) represents the center coordinate of obstacle k, (x n ,y n ) represents the coordinates of the current path node n, a k represents the effective coefficient of the threat cost. If there is no obstacle in the two-level prediction neighborhood of the current path node, then set a k is 0, otherwise it is set to 1, s k is the threat coefficient brought by obstacle k to the node n passing through the current path, and M is the set of obstacles in the two-level predicted neighborhood of the current path node; By using the Euclidean distance as a heuristic function, a stability weighted root order Sw not less than 1 is introduced. The heuristic term g(n) and the stability weighted root order Sw are expressed as: Among them, g(n) represents the heuristic term of the current path node n, sw represents a stability weighted root, and the value is the number of obstacles num in the neighborhood of all nodes before the current path node n carr The standard deviation σ plus 1, It represents the average number of obstacles in the neighborhood of all nodes before the current path node n. represents the number of obstacles in the neighborhood of each node before the current path node n, (x start ,y start ) represents the starting point coordinates of the path, (x n ,y n ) represents the coordinates of the current path node n; Further optimization paths include: The path segmentation selects the grouping interval according to the path tortuosity and the number of path nodes. The segmentation interval determination formula is expressed as: Among them, od represents the group spacing of the path segment, D represents the approximate number of path node spacing, pro sharp represents the probability of sharp bends at nodes in the path, Indicates the turning angle at each node, State node Indicates the bending state of each node. If the turning angle is greater than π / 3, it is a sharp bend node, otherwise it is a non-sharp bend node; By using a flexible high-order Bezier curve equation with the grouping interval od as the order, the smoothness of each segmented path is optimized to generate an optimal path.

2. A system using the AGV multi-stage path planning method based on the moss mouse algorithm as claimed in claim 1, It is characterized in that include, The data collection and planning module is used to obtain the environmental data of the AGV, including the plane length and width, obstacle locations, and start and end locations, simplify the route performance, and use the improved random tree algorithm to generate the initial valid path; A first processing module is used to expand the feasible area based on the initial valid path, adaptively rasterize the feasible area, and expand it in combination with the Rich-Moore neighbor tracking method; The second processing module is used to find the optimal path in the feasible area through the TCD* algorithm and remove redundant nodes on the path; The optimization module is used to segment the optimal path according to the nodes. The segmentation uses the dynamic four-sided window algorithm to avoid obstacles in real time and further optimize the path. The execution module is used to execute the obtained optimal path so that the automatic guided vehicle reaches the destination along the optimal path.

3. An electronic device, It is characterized in that include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the AGV multi-stage path planning method based on the moss mouse algorithm described in claim 1 are implemented.

4. A computer-readable storage medium, It is characterized in that It stores computer executable instructions, which, when executed by a processor, implement the steps of the AGV multi-stage path planning method based on the moss mouse algorithm described in claim 1.