CN113703452A - AGV path planning method for large-scale storage environment - Google Patents

AGV path planning method for large-scale storage environment Download PDF

Info

Publication number
CN113703452A
CN113703452A CN202110974440.7A CN202110974440A CN113703452A CN 113703452 A CN113703452 A CN 113703452A CN 202110974440 A CN202110974440 A CN 202110974440A CN 113703452 A CN113703452 A CN 113703452A
Authority
CN
China
Prior art keywords
agv
path
node
time
algorithm
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110974440.7A
Other languages
Chinese (zh)
Inventor
陈娟
张政
郭青
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Chemical Technology
Original Assignee
Beijing University of Chemical Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Chemical Technology filed Critical Beijing University of Chemical Technology
Priority to CN202110974440.7A priority Critical patent/CN113703452A/en
Publication of CN113703452A publication Critical patent/CN113703452A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle

Landscapes

  • Engineering & Computer Science (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to an AGV path planning method for a large-scale storage environment, which improves the single search speed by limiting the expansion of a search tree towards the direction of a target point of an AGV; the traditional MCTS algorithm is improved, node evaluation is carried out only when the search tree is expanded to a target point, the node evaluation is carried out once when the node expansion is carried out once, the time waste of a 'back propagation' process is avoided, and the algorithm efficiency is improved. Then, a DP-MCTS algorithm is established by combining the idea of DP algorithm multi-stage optimization problem and the improved MCTS algorithm, and the path planning efficiency of the AGV under a large-scale automatic storage system is further improved; the method can shorten the AGV path planning time, obtain the paths with the shortest distance and time of multiple AGVs, reduce the energy consumption of the multiple AGV systems, and conveniently and quickly select the optimal AGV path, thereby being beneficial to improving the real-time performance of the whole storage system.

Description

AGV path planning method for large-scale storage environment
Technical Field
The invention belongs to the technical field of AGV path planning, and particularly relates to an AGV path planning method for a large storage environment.
Background
The large-scale automatic storage system only needs a small amount of manual work and realizes a system for automatically storing and taking materials through an Automatic Guided Vehicle (AGV). As a primary transport in an automated storage system, an AGV undertakes the work of transporting goods from a specified origin to a specified destination. The operating efficiency of the AGVs has a significant impact on the overall efficiency of the system, and the operating efficiency of the AGVs depends on the efficiency of the path planning. The reasonable planning of the AGV path not only can improve the efficiency of the warehousing system, but also can save a large amount of energy. The multiple AGV path planning gradually evolves into a multi-target combination optimization problem: in addition to paying attention to algorithm rapidity, shortest path distance, minimum time and minimum energy consumption need to be considered. Therefore, the problem of path planning of multiple AGVs in a large-scale automated storage system is receiving wide attention.
The existing Dijkstra algorithm is one of the most effective methods for AGV path planning, but cannot meet the requirement of continuously expanding the storage environment scale; in a traditional Dynamic Programming (DP) algorithm, a sample set is iterated by randomly selecting a starting point, and a Value Function closest to an optimal strategy is obtained through multiple training, so that a problem-solving strategy is found. Along with the enlargement of the scale of the path planning environment, the training times and the single training time are increased, and the calculation amount of the DP algorithm in a large-scale automatic storage system is exponentially increased along with the increase of the number of the work sites; the traditional Monte Carlo Tree Search algorithm (MCTS) lacks a heuristic mechanism and cannot balance 'exploration' and 'utilization', and a 'blind Search' mechanism of the traditional Monte Carlo Tree Search algorithm has larger calculation amount. In addition, in the traditional MCTS node expansion process, the traversed nodes are evaluated only when the leaf nodes of the search tree expand to the target point, which is not efficient in large warehousing systems.
Therefore, how to conveniently and quickly select the optimal AGV path has become a technical problem to be solved urgently.
Disclosure of Invention
In order to solve the problems of low efficiency and poor real-time performance in the prior art, the invention provides an AGV path planning method for a large storage environment, which has the characteristics of higher efficiency, better real-time performance and the like.
According to the specific implementation mode of the invention, the AGV path planning method for the large-scale storage environment comprises the following steps:
performing area division based on areas determined by the positions of the driving starting point and the target point of the AGV to determine a path searching range of the AGV;
limiting the AGV node expansion process based on the positions of the current node and the target node in the search range so as to limit the search direction;
screening the node to a target node once when node expansion is performed once every time so as to find all feasible paths from the starting point to the target point of the AGV;
and scoring the feasible paths, selecting the feasible path with the minimum score as the distance-time shortest path of the AGV, and determining the optimal path of the AGV according to the priority.
Further, the AGV path planning method for a large storage environment further includes:
and performing pair-wise conflict pre-detection on the distance-time shortest paths of the AGVs, and determining a path conflict key segment and the length thereof.
Further, the AGV path planning method for a large storage environment further includes:
acquiring the priority of the related AGV with the path conflict, and modifying the path of the related AGV and/or the time for reaching each node in the path by adopting a self-adaptive method;
and when the AGV encounters a temporary obstacle, solving the path conflict by adopting an adaptive method.
Further, the AGV path planning method for a large storage environment further includes:
and acquiring a transportation task, finding an idle AGV which is closest to the task starting point Manhattan based on a scheduling algorithm, and carrying.
Further, the area division based on the determined positions of the travel start point and the target point of the AGV to determine the path search range of the AGV includes:
and performing region division in multiple stages based on a dynamic programming algorithm.
Further, the limiting the AGV node expansion process based on the positions of the current node and the target node in the search range to limit the search direction thereof includes:
and defining the searching direction of the node based on the depth-first search and the breadth-first search and the Monte Carlo tree searching algorithm added in the defining process.
Further, screening the node to the target node once every time node expansion is performed to find all feasible paths from the starting point to the target point of the AGV includes:
and establishing an adjacency list based on the load condition of the AGV, and limiting the expansion range of the search tree according to the positions of the starting point and the target point of the AGV before expanding the leaf nodes, so that the number of the unorthodox nodes is reduced.
Further, the scoring the feasible paths, selecting the feasible path with the minimum score as the distance-time shortest path of the AGVs, and determining the optimal path of the AGVs according to the priority includes:
and obtaining the shortest distance path based on the feasible path score, simultaneously adding a punishment factor for the AGV turning, and selecting the path with the minimum punishment as the optimal path of the AGV.
The invention has the beneficial effects that: the method comprises the steps that the expansion of a search tree towards a target point direction of the AGV is limited, and the single search speed is improved; the traditional MCTS algorithm is improved, node evaluation is carried out only when the search tree is expanded to a target point, the node evaluation is carried out once when the node expansion is carried out once, the time waste of a 'back propagation' process is avoided, and the algorithm efficiency is improved. Then, a DP-MCTS algorithm is established by combining the idea of DP algorithm multi-stage optimization problem and the improved MCTS algorithm, and the path planning efficiency of the AGV under a large-scale automatic storage system is further improved; the method can shorten the AGV path planning time, obtain the paths with the shortest distance and time of multiple AGVs, reduce the energy consumption of the multiple AGV systems, and conveniently and quickly select the optimal AGV path, thereby being beneficial to improving the real-time performance of the whole storage system.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of a method for AGV path planning for a large storage environment provided in accordance with an exemplary embodiment;
FIG. 2 is a diagram of a simulation of a 25X 34 large automated storage system including 10 AGVs according to an exemplary embodiment;
FIG. 3 is a graph comparing calculated time for finding a shortest distance-time path for a DP-MCTS algorithm proposed by the present invention and a conventional DP algorithm running 1000 times respectively in a single AGV scenario according to an exemplary embodiment;
FIG. 4 is a graph comparing energy consumption for a DP-MCTS algorithm with a conventional DP algorithm for finding a shortest distance-time path 1000 times for a single AGV scenario according to an exemplary embodiment;
FIG. 5 is a graph comparing the trend of path computation time for the classical DP algorithm and the DP-MCTS algorithm with increasing environmental size (number of nodes) in accordance with an exemplary embodiment;
FIG. 6 is a graph comparing the trend of system energy consumption for the classical DP algorithm and the DP-MCTS algorithm as the environmental size (number of nodes) increases, according to an exemplary embodiment;
FIG. 7 is a graph comparing the trend of system throughput changes for the classical DP algorithm and the DP-MCTS algorithm as the environment size (number of nodes) increases, provided in accordance with an exemplary embodiment;
FIG. 8 is another flow chart of a method for AGV path planning for a large storage environment provided in accordance with an exemplary embodiment;
FIG. 9 is a graph comparing the calculated time for an initial shortest path found by the DP-MCTS algorithm and the conventional DP algorithm running 1000 times when 10 AGVs are involved in a simultaneous travel according to an exemplary embodiment;
FIG. 10 is a graph comparing power consumption for DP-MCTS algorithm and conventional DP algorithm running 1000 times to find the initial shortest path when 10 AGVs are driven simultaneously according to an exemplary embodiment;
FIG. 11 is a graph comparing power consumption for DP-MCTS algorithm and conventional DP algorithm running 1000 times to find the initial shortest path when 10 AGVs are involved in traveling simultaneously according to an exemplary embodiment;
FIG. 12 is a comparison table of evaluation indexes of a 25X 34 large automated storage system according to an exemplary embodiment of the AGV path planning method for a large storage environment and a conventional method.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the examples given herein without any inventive step, are within the scope of the present invention.
Firstly, the evaluation indexes of the path planning algorithm of the large-scale automatic storage system can be summarized as follows: (1) system path Computation Time (CTR); (2) travel path length (RDA) for multiple AGVs; (3) multiple AGV paths complete utilization Time (Routes Time for AGV, RTA); (4) multiple AGV tasks (multiple AGVs all move from the starting point to the target point) finish use Time (Task Time for AGV, TTA)(s); (5) the number of completed AGVs moving from the start point to the target point (Finished Routes per Unit Time, FRUT) in a Unit Time; (6) the number of turns (Turn Numbers for AGVs, TNA) during the movement of a plurality of AGVs; (7) total Energy Cost for AGVs (TECA) of multiple AGVs.
Wherein CTR, RDA and TNA can be directly obtained through program simulation, and other evaluation indexes are calculated on the basis of the CTR, the RDA and the TNA: setting n AGV in the environment, wherein the moving time of the AGV in the grid graph between every two adjacent nodes is tmThe time taken to turn at a node is tδSuppose that the energy consumption of the system in performing the path calculation is E1W/h, the energy consumption of the AGV in the straight line is E2W/h, energy consumption at the time of parking and turning is E3W/h, taking 1min as unit time, and the calculation method of RTA, TTA, FRUT and TECA is as follows: RTA ═ RDA × tm+TNA×tδ，TTA＝RTA+CTR，FRUT＝n/60×TFR，TECA＝CTR×E1+RDA×E2+TNA×E3。
Referring to fig. 1, an embodiment of the present invention provides an AGV path planning method for a large storage environment, which specifically includes the following steps:
101. performing area division based on areas determined by the positions of the driving starting point and the target point of the AGV to determine a path searching range of the AGV;
a method for dividing the area of a large-scale automatic storage system is provided by adopting the idea of solving the problem in multiple stages by adopting a dynamic programming algorithm;
specifically, referring to fig. 2, let the dimension of the storage environment be m × n, and let the start coordinate of the AGV be (S)x,Sy) The abscissa of the target point is (G)x,Gy) The abscissa of the left boundary of the grid pattern is XlThe abscissa of the right border is XrThe ordinate of the upper boundary being YuThe ordinate of the lower boundary is Yd. The upper left corner coordinate of the path planning area of this AGV is
Figure BDA0003226803330000061
Coordinates of the upper right corner are
Figure BDA0003226803330000062
The coordinate of the lower left corner is
Figure BDA0003226803330000063
The coordinate of the lower right corner is
Figure BDA0003226803330000064
N 'total of route planning region'rowLine n'colAnd (4) columns. The method for dividing the path planning region comprises the following steps:
Figure BDA0003226803330000065
Figure BDA0003226803330000066
Figure BDA0003226803330000067
Figure BDA0003226803330000068
n'row＝max(Sx,Gx)-Xlformula (5)
n'col＝|Gy-SyEquation (6)
The formula (1) is a calculation method for coordinates of the upper left corner of the path planning area of the agent; the formula (2) is a path planning area upper right corner coordinate calculation method; formula (3) is a lower left corner coordinate calculation method; the formula (4) is a calculation mode of coordinates of a lower right corner; equations (5) and (6) are the number of lines (n ') in the route planning region'row) And column number (n'col) The method of (3).
102. Limiting the AGV node expansion process based on the positions of the current node and the target node in the search range so as to limit the search direction;
an improved monte carlo tree search algorithm incorporating a "pruning", i.e. a defined, process is used in conjunction with depth-first and breadth-first searches to provide a heuristic mechanism:
specifically, assume that the AGV start point coordinates are (S)x,Sy) The coordinates of the target point are (G)x,Gy) The coordinate of the current node is (V)x,Vy) The child node coordinate to be extended is (V'x,V'y) The method for selecting the child node comprises the following steps:
(1)IfGx≥Sx：
Figure BDA0003226803330000071
Figure BDA0003226803330000072
(2)IfGx<Sx：
Figure BDA0003226803330000073
Figure BDA0003226803330000074
equation (7) indicates that if the target point of the AGV is located at the lower right of the start point of the AGV, only the node located at the lower right of the current node is considered to be expanded; equation (8) indicates that if the target point of the AGV is located at the upper right of the start point of the AGV, only the node located at the upper right of the current node is considered to be expanded; equation (9) indicates that if the target point of the AGV is located at the lower left of the start point of the AGV, only the node located at the lower left of the current node is considered to be expanded; equation (10) indicates that if the target point of the AGV is located above and to the left of the start point of the AGV, only the node located above and to the left of the current node is considered to be expanded.
103. Screening the node to a target node once when node expansion is performed once every time so as to find all feasible paths from the starting point to the target point of the AGV;
namely, determining all feasible paths from the starting point to the target point of the AGV through single step updating, evaluating all the feasible paths, and determining the path with the shortest distance and time, wherein the method comprises the following steps:
and establishing an adjacency list according to whether the AGV executes a load task or an idle task, and limiting the expansion range of the search tree according to the positions of the starting point and the target point of the AGV before expanding the leaf nodes, so that the number of the nodes without joints is reduced, and the algorithm efficiency is improved.
Specifically, let U (v, t) ═ U (v) be the path of the AGV1,t1),u(v2,t2),…,u(vk,tk) }, element u (v)i,ti) Indicating the state, v, of the AGV reaching the ith node in the pathiNumber, t, representing the ith node in the pathiIndicating the time when the AGV reaches this node, i is 1,2,3 … k (k is the path length). Calculating an adjacency list according to whether each AGV executes a load task or an idle task: if it is executedIs a load task, and establishes an obstacle adjacency list adj according to the algorithm 1obs[](ii) a If no-load task is executed, then establish the obstacle-free adjacency list adj [ 2 ] according to Algorithm 2]。
Figure BDA0003226803330000081
In Algorithm 1, "G' [ i][j]：G’[i+1][j+1]→adjobs[]"represents a node G' [ i ]][j]And node G' [ i +1]][j+1]Adjoining, connecting the node G' i][j]Adding adjacency list adjobs[]In (1).
Figure BDA0003226803330000082
In Algorithm 2, "G' [ i ] [ j ]: g '[ i +1] [ j +1] → adj [ ] "indicates that the node G' [ i ] [ j ] is adjacent to the node G '[ i +1] [ j +1], and the node G' [ i ] [ j ] is added to the adjacency table adj [ ].
Finding all feasible path pseudo codes from the AGV starting point to the target point in the segmented area according to the adjacency list is as follows:
Figure BDA0003226803330000083
Figure BDA0003226803330000091
and traversing nodes adjacent to the starting point of the AGV in the area (a 3 rd row), then performing 'pruning' on all adjacent nodes, selecting the nodes which accord with the advancing direction of the AGV as the starting points, performing node expansion again, and adding the nodes into the path list path [ ] (a 4 th row and a 5 th row).
104. And scoring the feasible paths, selecting the feasible path with the minimum score as the distance-time shortest path of the AGV, and determining the optimal path of the AGV according to the priority.
Namely, selecting the path with the minimum Move value as the shortest distance-time path of the AGV, and determining the initial optimal path according to the priority, wherein the method comprises the following steps:
and (3) providing an evaluation standard based on the path 'Move' value, scoring all feasible paths from the starting point to the target point in the node process, and adding punishment to the paths to find out the path with the minimum 'Move' value, namely the path with the minimum turning from the starting point to the target point.
Specifically, the node evaluation process is advanced to the front of the simulation process, the shortest distance path is obtained by scoring the feasible paths, meanwhile, a penalty factor is added for AGV turning, and then the path with the smallest penalty is selected as the final path of the AGV. The improved MCTS algorithm scores the process when the leaf node is expanded once, and whether the expanded node is optimal or not can be obtained without a back propagation process after the search tree is expanded to a target point. At this time, only the path with the shortest distance from the starting point to the target point of the AGV and the least turning needs to be found from the search tree.
Let the AGV path be U (v, t) ═ U (v)1,t1),u(v2,t2),…,u(vk,tk) }, element u (v)i,ti) Indicating the state, v, of the AGV reaching the ith node in the pathiNumber, t, representing the ith node in the pathiIndicating the time when the AGV reaches this node, i ═ 1,2,3 … k (k is the path length), the pseudo code is as follows:
Figure BDA0003226803330000101
on the basis of finding all feasible paths according with the advancing direction of the AGV, selecting the path with the least turning according to the path Move value: traversal list path [ 2 ]]Each path (path 2)]i) Of ((V))'x,V'y) And (V)x,Vy) Wherein i is 1,2 … k, k is path [ 2 ]]iLength of (d). If node (V)x,Vy) If it is a turning point, the Move value of the path is +2, otherwise + 1. Then, the list path [ 2 ]]The Move of all the paths in the path list is aggregated and is in one-to-one correspondence with the paths. Finally, by selecting the list Move]The path corresponding to the minimum value is the shortest distance-time path (i.e., the path with the least turns) from the starting point to the target point of the AGV.
Referring to fig. 3 to 7, in the case of simultaneous operation of 10 AGVs, the improved DP-MCTS algorithm proposed by the present invention compares the calculated time of the shortest distance-time path in the automatic storage system with the conventional DP algorithm at different scales (different node numbers), the path energy consumption, and the throughput (number of transport tasks completed per unit time) of the multiple AGV system. The larger the automated warehousing system scale (the larger the number of nodes), the greater the advantage of the improved algorithm described in this embodiment. Therefore, the single search speed is improved by limiting the expansion of the search tree towards the target point direction of the AGV; the traditional MCTS algorithm is improved, node evaluation is carried out only when the search tree is expanded to a target point, the node evaluation is carried out once when the node expansion is carried out once, the time waste of a 'back propagation' process is avoided, and the algorithm efficiency is improved. Then, a DP-MCTS algorithm is established by combining the idea of DP algorithm multi-stage optimization problem and the improved MCTS algorithm, and the path planning efficiency of the AGV under a large-scale automatic storage system is further improved; the method can shorten the AGV path planning time, obtain the paths with the shortest distance and time of multiple AGVs, reduce the energy consumption of the multiple AGV systems, and conveniently and quickly select the optimal AGV path, thereby being beneficial to improving the real-time performance of the whole storage system.
Referring to fig. 8, the AGV path planning method for large storage environment according to the present invention includes the following steps:
201. performing area division based on areas determined by the positions of the driving starting point and the target point of the AGV to determine a path searching range of the AGV;
202. limiting the AGV node expansion process based on the positions of the current node and the target node in the search range so as to limit the search direction;
203. screening the node to a target node once when node expansion is performed once every time so as to find all feasible paths from the starting point to the target point of the AGV;
204. and scoring the feasible paths, selecting the feasible path with the minimum score as the distance-time shortest path of the AGV, and determining the optimal path of the AGV according to the priority.
205. And performing pair-wise conflict pre-detection on the distance-time shortest paths of the AGVs, and determining a path conflict key segment and the length thereof.
206. Acquiring the priority of the related AGV with the path conflict, and modifying the path of the related AGV and/or the time for reaching each node in the path by adopting a self-adaptive method;
207. and when the AGV encounters a temporary obstacle, solving the path conflict by adopting an adaptive method.
208. And acquiring a transportation task, finding an idle AGV which is closest to the task starting point Manhattan based on a scheduling algorithm, and carrying.
Specifically, in actual execution, when a transport task is received, an idle AGV is scheduled in real time to execute the task, and first, the AGV path planning method for the multi-use large-scale storage environment, which is provided by the above embodiment, plans the distance-time shortest path from the starting point to the target point for each AGV;
step 205, detecting whether path conflicts exist among the AGVs in the environment by adopting a path conflict detection method of multiple AGVs based on the path conflict key segment, and determining the length of the conflict key segment;
the method for detecting the path conflict of the multiple AGVs based on the path conflict key segment comprises the following steps: (1) collision pre-detection and (2) determination of collision critical section. Firstly, determining the priority (prio) of the AGV according to an equation (11); the initial shortest path for the AGV is then determined according to table 1.
Figure BDA0003226803330000121
TABLE 1 determination of initial Path
Figure BDA0003226803330000122
Figure BDA0003226803330000131
Suppose that there are p AGVs in the grid pattern at the same time, and the path is:
AGV1：
Figure BDA0003226803330000132
AGV2：
Figure BDA0003226803330000133
AGVp：
Figure BDA0003226803330000134
wherein k is1,k2…kpRespectively representing AGV1,AGV2…AGVpDistance-length of the shortest path in time.
Therefore, the node set through which p AGVs pass is S ═ S1,S2…SpWhere AGViThe nodes passing through are collected as
Figure BDA0003226803330000135
The time set of p AGVs passing through the node is T ═ T1,T2…TpWhere AGViThe time passing through the node is set as
Figure BDA0003226803330000136
(i ═ 1,2 … p). The path conflict detection is divided into two parts of conflict pre-detection and conflict key section determination, and pseudo codes of the path conflict pre-detection are as follows:
Figure BDA0003226803330000137
whether the same node exists between the initial distance-time shortest paths of p AGVs can be judged through the algorithm 5, and if two paths contain the same node, the time of the related AGVs reaching the nodes is further compared through a path conflict determining key segment (algorithm 6) to determine whether the two paths conflict. The pseudo code that determines the path conflict critical section is as follows:
Figure BDA0003226803330000138
Figure BDA0003226803330000141
the principle of determining path collision is: if two AGVs arrive at a duplicate node at the same time, then there is indeed a path conflict between the two AGVs. Since path collisions may occur on road segments of two adjacent nodes in addition to the nodes, but the list
Figure BDA0003226803330000142
In only store the AGViTime to reach each node in the path without storing the AGViTime on the road segment. Lines 1-3 of algorithm 6 are based on AGViCalculating the time of arriving at the road section according to the time of arriving at the node in the path; lines 4-8 indicate that
Figure BDA0003226803330000143
This comparison, where n represents the number of AGVs, results in a list of times each AGV reaches a node and a road segment. Then adding the node or section corresponding to the repeat element to the list section]And obtains the length of the conflict key section.
Step 206, if there is a path conflict, determining whether the related AGVs have an alternative optimal path and adopting a self-adaptive path conflict resolution strategy: the AGV selects another alternative path or waits before colliding with the key segment;
in order to ensure that the total system time for all AGVs to complete a task is minimized, the specification modifies the path or schedule of the AGVs with a small Manhattan distance (equation (12)); (1) if the target point of the agent with the large Manhattan distance is on the shortest path of the AGV with the small Manhattan distance, stopping the agent with the small Manhattan distance before entering the conflict key section and waiting for Len multiplied by run _ time + work _ time(s), wherein Len represents the length of the conflict key section; run _ time represents the travel time of each AGV between adjacent nodes; word _ time represents the time required for the AGV to stay on a specified node; (2) if the target point of the agent with the large Manhattan distance is not on the shortest path of the AGV with the small Manhattan distance, the AGV with the small Manhattan distance stops waiting for Len × run _ time(s) before entering the conflict key segment. And then modifying the final path schedule of the AGV with the small Manhattan distance to obtain the optimal collision-free path of all the AGVs.
On the plane, point Pi(xi,yi) And point Pj(xj,yj) The Manhattan distance of (A) is shown in equation (12).
d(i,j)＝|xi-xj|+|yi-yjEquation (12)
Wherein d (i, j) represents the Manhattan distance from i to j; x is the number ofiDenotes the abscissa, y, of iiIs the ordinate of i; x is the number ofjDenotes the abscissa, y, of jjIs the ordinate of j.
The method of determining the final path of the AGV is shown in Table 2:
table 2 determines the final path of an AGV
Figure BDA0003226803330000151
Repeating the step 2-3 until no conflict exists, and issuing an optimal path planning result and a time schedule to each AGV by the system;
step 207, when the AGV encounters a temporary obstacle during traveling according to the delivered route, the AGV stops before the temporary obstacle, then uses the current node as a starting point, and uses the improved DP-MCTS algorithm (as described in S11-S14) provided by the present invention to determine whether an alternative optimal route exists between the current node and the target point in real time, and then uses a self-adaptive method (directly uses the alternative route or waits for the obstacle to be removed) to solve the route conflict.
Step 208, when the multiple AGV system receives a new transportation task at t, the task starting point is set as (x)0,y0). At this time, it is automatedThere are p candidate AGVs (i.e. agents with idle transportation and power higher than 20%) in the storage system, and their locations are Loc { (x)1,y1),(x2,y2),…,(xp,yp)}. System real-time calculation list Loc [ 2 ]]Each element in (a) and a task starting point (x)0,y0) The Manhattan distance of (1) is calculated by the method shown in formula (12). The system stores the calculation results in a list
Figure BDA0003226803330000161
Wherein p is the number of the candidate AGVs, the optimal AGV is the AGV corresponding to the minimum Manhattan distance, namely argmin { D }M}。
Wherein figure 2 is a 25 x 34 large automated warehousing system employed by the present invention. Note: where the gray grid represents the determined obstacle storage node (representing shelf parking stations in an automated storage system) and the white grid represents the aisle of the AGV. Fig. 12 is a comparison table of evaluation indexes of the multiple AGV path planning algorithm provided by the present invention and a conventional algorithm in a 25 × 34 large-scale automatic warehousing system including 10 AGVs running simultaneously, and it can be seen from the comparison table that the method of the present embodiment can determine a collision-free distance-time shortest path for the multiple AGVs, and by using the method, energy consumption of the multiple AGVs system can be reduced, and the transport task amount (throughput in the automatic warehousing system) completed by the AGVs per unit time can be increased.
With reference to the comparison effects shown in fig. 9 to fig. 11, according to the AGV path planning method for a large storage environment provided by the embodiment of the present invention, by using the idea of optimizing the problem by stages by using the DP algorithm, the large automatic storage system is segmented according to the positions of the AGV start point and the target point, so that the environment dimension reduction is realized, and the number of training samples is reduced. The invention improves the traditional MCTS algorithm in the following three aspects according to the characteristics of a large-scale automatic storage system: (1) providing a node expansion process with pruning operation, endowing the algorithm with heuristic search capability according to the positions of a starting point and a target point of the AGV, and increasing the algorithm search speed; (2) the process of 'back propagation' is advanced to the process of 'simulation', so that the real-time evaluation of leaf nodes is realized, and the path optimization efficiency is improved; (3) and providing a path evaluation standard based on the 'Move' value, and reducing the turning times in the moving process of the AGV according to the minimum principle of the 'Move' value to obtain the path with the shortest time and distance of the AGV. Meanwhile, different path planning methods are provided according to whether the AGV is loaded or not, and the efficiency of the automatic storage system is further improved. And performing path conflict pre-detection on the basis of obtaining the time shortest path of each AGV so as to determine the priority of the AGV, conflict key segments and the Manhattan distance from the starting point of the AGV to the target point. And then determining a conflict resolution strategy according to the size of the Manhattan distance. By adopting the method, the distance-time shortest path can be planned for multiple AGVs, the energy consumption of the multiple AGV system is favorably reduced, and the real-time performance of the whole automatic storage system is favorably improved.
It is understood that the same or similar parts in the above embodiments may be mutually referred to, and the same or similar parts in other embodiments may be referred to for the content which is not described in detail in some embodiments.
It should be noted that the terms "first," "second," and the like in the description of the present invention are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. Further, in the description of the present invention, the meaning of "a plurality" means at least two unless otherwise specified.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps of the process, and alternate implementations are included within the scope of the preferred embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present invention.
It should be understood that portions of the present invention may be implemented in a combination of hardware and software. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, for a single AGV, a field programmable logic controller (FPGA) may be used, or a combination of a server and wireless communication may be used, that is, a path is planned on the server first, and then a planning result is transmitted to the AGV through wireless communication; and for the situation of multiple AGVs, because the online calculated amount is large, the multiple AGV paths can be planned only through the high-speed server, the driving states of the multiple AGVs and whether temporary obstacles appear on a route are monitored in real time, when the AGVs reach each node, the current positions are uploaded to the server through wireless communication, and the server performs path conflict detection/solution according to the current positions.
It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments.
In addition, functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium.
The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.

Claims (8)

1. An AGV path planning method for a large storage environment is characterized by comprising the following steps:
performing area division based on areas determined by the positions of the driving starting point and the target point of the AGV to determine a path searching range of the AGV;
limiting the AGV node expansion process based on the positions of the current node and the target node in the search range so as to limit the search direction;
screening the node to a target node once when node expansion is performed once every time so as to find all feasible paths from the starting point to the target point of the AGV;
and scoring the feasible paths, selecting the feasible path with the minimum score as the distance-time shortest path of the AGV, and determining the optimal path of the AGV according to the priority.
2. The AGV path planning method for large storage environment according to claim 1, further comprising:
and performing pair-wise conflict pre-detection on the distance-time shortest paths of the AGVs, and determining a path conflict key segment and the length thereof.
3. The AGV path planning method for large storage environment according to claim 2, further comprising:
acquiring the priority of the related AGV with the path conflict, and modifying the path of the related AGV and/or the time for reaching each node in the path by adopting a self-adaptive method;
and when the AGV encounters a temporary obstacle, solving the path conflict by adopting an adaptive method.
4. The AGV path planning method for large scale storage environment according to claim 3, further comprising:
and acquiring a transportation task, finding an idle AGV which is closest to the task starting point Manhattan based on a scheduling algorithm, and carrying.
5. The AGV path planning method for large storage environment according to claim 1, wherein said area division based on the positions of the travel start point and the target point of the AGV to determine the path search range of the AGV comprises:
and performing region division in multiple stages based on a dynamic programming algorithm.
6. The AGV path planning method for large storage environment according to claim 1, wherein said defining the AGV node expansion process based on the positions of the current node and the target node in the search range to define the search direction comprises:
and defining the searching direction of the node based on the depth-first search and the breadth-first search and the Monte Carlo tree searching algorithm added in the defining process.
7. The method of claim 1, wherein the step of performing a screening process on each node to a target node for each node expansion to find all feasible routes from the AGV start point to the target point comprises:
and establishing an adjacency list based on the load condition of the AGV, and limiting the expansion range of the search tree according to the positions of the starting point and the target point of the AGV before expanding the leaf nodes, so that the number of the unorthodox nodes is reduced.
8. The method of claim 1, wherein the scoring the feasible paths, selecting the feasible path with the smallest score as the shortest distance-time path of the AGVs, and determining the optimal path of the AGVs according to the priority comprises:
and obtaining the shortest distance path based on the feasible path score, simultaneously adding a punishment factor for the AGV turning, and selecting the path with the minimum punishment as the optimal path of the AGV.
CN202110974440.7A 2021-08-24 2021-08-24 AGV path planning method for large-scale storage environment Pending CN113703452A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110974440.7A CN113703452A (en) 2021-08-24 2021-08-24 AGV path planning method for large-scale storage environment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110974440.7A CN113703452A (en) 2021-08-24 2021-08-24 AGV path planning method for large-scale storage environment

Publications (1)

Publication Number Publication Date
CN113703452A true CN113703452A (en) 2021-11-26

Family

ID=78654338

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110974440.7A Pending CN113703452A (en) 2021-08-24 2021-08-24 AGV path planning method for large-scale storage environment

Country Status (1)

Country Link
CN (1) CN113703452A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114355926A (en) * 2021-12-29 2022-04-15 深圳市云鼠科技开发有限公司 Path planning method and device, robot and storage medium
CN114384908A (en) * 2021-12-16 2022-04-22 杭州申昊科技股份有限公司 Intelligent navigation path planning system and method for track robot
CN114626794A (en) * 2022-05-17 2022-06-14 山东西曼克技术有限公司 Internet of things-based warehousing vertical warehouse shuttle vehicle path planning method and system
CN117151596A (en) * 2023-11-01 2023-12-01 领先未来科技集团有限公司 Logistics management method, system and storage medium for storage AGVs (automatic guided vehicle) through Internet of things

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106527448A (en) * 2016-12-16 2017-03-22 浙江工业大学 Improved A* robot optimal path planning method suitable for warehouse environment
CN111708364A (en) * 2020-06-19 2020-09-25 南京理工大学 Improved AGV path planning method based on A-x algorithm
CN112161630A (en) * 2020-10-12 2021-01-01 北京化工大学 AGV (automatic guided vehicle) online collision-free path planning method suitable for large-scale storage system
CN112486178A (en) * 2020-12-03 2021-03-12 哈尔滨理工大学 Dynamic path planning method based on directed D (delta) algorithm

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106527448A (en) * 2016-12-16 2017-03-22 浙江工业大学 Improved A* robot optimal path planning method suitable for warehouse environment
CN111708364A (en) * 2020-06-19 2020-09-25 南京理工大学 Improved AGV path planning method based on A-x algorithm
CN112161630A (en) * 2020-10-12 2021-01-01 北京化工大学 AGV (automatic guided vehicle) online collision-free path planning method suitable for large-scale storage system
CN112486178A (en) * 2020-12-03 2021-03-12 哈尔滨理工大学 Dynamic path planning method based on directed D (delta) algorithm

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114384908A (en) * 2021-12-16 2022-04-22 杭州申昊科技股份有限公司 Intelligent navigation path planning system and method for track robot
CN114384908B (en) * 2021-12-16 2023-07-11 杭州申昊科技股份有限公司 Intelligent navigation path planning system and method for track robot
CN114355926A (en) * 2021-12-29 2022-04-15 深圳市云鼠科技开发有限公司 Path planning method and device, robot and storage medium
CN114626794A (en) * 2022-05-17 2022-06-14 山东西曼克技术有限公司 Internet of things-based warehousing vertical warehouse shuttle vehicle path planning method and system
CN117151596A (en) * 2023-11-01 2023-12-01 领先未来科技集团有限公司 Logistics management method, system and storage medium for storage AGVs (automatic guided vehicle) through Internet of things
CN117151596B (en) * 2023-11-01 2023-12-29 领先未来科技集团有限公司 Logistics management method, system and storage medium for storage AGVs (automatic guided vehicle) through Internet of things

Similar Documents

Publication Publication Date Title
CN113703452A (en) AGV path planning method for large-scale storage environment
CN109115226B (en) Route planning method for avoiding multi-robot conflict based on jumping point search
CN113074728B (en) Multi-AGV path planning method based on jumping point routing and collaborative obstacle avoidance
CN109949604B (en) Large parking lot scheduling navigation method and system
CN111532641B (en) Parallel path planning method for automatic guide vehicle in storage sorting
CN112833905B (en) Distributed multi-AGV collision-free path planning method based on improved A-x algorithm
CN109782757A (en) A kind of path dispatching method of more AGV systems based on subsection scheduling
Zhu et al. DSVP: Dual-stage viewpoint planner for rapid exploration by dynamic expansion
CN113075927A (en) Storage latent type multi-AGV path planning method based on reservation table
CN112525196B (en) AGV route planning and scheduling method and system based on multidimensional data
CN113625716B (en) Multi-agent dynamic path planning method
CN109115220B (en) Method for parking lot system path planning
CN115237135A (en) Mobile robot path planning method and system based on conflict
CN113899383B (en) Multi-vehicle deadlock prevention method, system, equipment and storage medium based on short path
CN111598332A (en) Workshop double-resource integrated scheduling method and system in intelligent manufacturing environment
CN113592158A (en) AGV and machine combined scheduling method in multi-AGV path planning and multi-AGV intelligent production line
CN114489062A (en) Workshop logistics-oriented multi-automatic guided vehicle distributed dynamic path planning method
CN111857149A (en) Autonomous path planning method combining A-algorithm and D-algorithm
CN115437382A (en) Multi-AGV path planning method and system for unmanned warehouse and equipment medium
Shi et al. Task allocation and path planning of many robots with motion uncertainty in a warehouse environment
CN112161630B (en) AGV (automatic guided vehicle) online collision-free path planning method suitable for large-scale storage system
CN115638804B (en) Deadlock-free unmanned vehicle online path planning method
CN116719312A (en) Multi-AGV unlocking method based on turn-back avoidance in single-way scene
CN114187781B (en) Distributed multi-vehicle cooperative behavior decision method and system
Haiming et al. Algorithm of path planning based on time window for multiple mobile robots in warehousing system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20211126

RJ01 Rejection of invention patent application after publication