CN107036618A - A kind of AGV paths planning methods based on shortest path depth optimization algorithm - Google Patents

Abstract

The present invention relates to a kind of AGV paths planning methods based on shortest path depth optimization algorithm, including：PC scheduling system receives task requests and sends solicited message to every AGV vehicle-mounted monitoring management system；The positional information of the positional information and task requests that perform task AGV is transferred to path optimizing system by AGV vehicle-mounted monitorings management system；Path optimizing system is by the feedback of the information of optimal path into PC scheduling system；After AGV completion tasks, task completed information is fed back into PC scheduling system.The present invention combs the detailed Optimization Steps of AGV path planning systems again, removes the flow of tradition AGV overcomplicated, on the basis for retaining critical workflow node, and the co-ordination relation between AGV system modules is planned again.

Description

An AGV path planning method based on the shortest path depth optimization algorithm

technical field

The invention relates to the technical field of an AGV path optimization method and a dispatch management system, in particular to an AGV path planning method based on a shortest path depth optimization algorithm.

Background technique

AGV is the abbreviation of Automated Guided Vehicle (Automated Guided Vehicle). It is the most common transportation equipment for delivering materials in unmanned workshops. The current AGV uses laser guidance for positioning and navigation in the industry, and the car is equipped with a PC. With programming, feedback, monitoring and other functions.

At present, the path planning algorithms commonly used in the AGV path planning system include Dijkstra, A* algorithm, etc. Among them, the Dijkstra algorithm is mainly used for the single-source shortest path, and the adjacency matrix is used to store the map model nodes, and the nodes are divided into two groups: one group stores the shortest The node set A of the path, and another set of node sets B that store undetermined paths, search for the nodes in the B set in turn according to the search method of the Dijkstra algorithm and add them to the A set. The distance from the starting node to each node in set A is not greater than the shortest distance to the nodes in set B. The existing Dijkstra algorithm search is a way of traversing all path nodes and node storage and search, which greatly reduces the running speed of the algorithm and makes the AGV path planning system extremely slow. The feedback of path information in the AGV path planning system mostly adopts alarm information feedback. When the AGV performs tasks abnormally, the AGV system will feed back the alarm information and take corresponding measures.

Contents of the invention

The purpose of the present invention is to provide a method for improving the utilization rate of nodes and road sections in the system, optimizing the workflow of path planning in the AGV system, and improving the efficiency and stability of the system. AGV path planning method based on the shortest path depth optimization algorithm.

In order to achieve the above object, the present invention adopts the following technical solutions: an AGV path planning method based on the shortest path depth optimization algorithm, the method includes the steps of the following order:

(1) Initialize the AGV dispatch management system. The PC dispatch system reads the path map model already established in the database, extracts the information of key nodes P and road sections L in the map, and stores the node information in an optimized storage method. At the same time, the PC The machine scheduling system reads all the AGV status information fed back by the laser navigation system, monitors and positions the AGV, and at the same time, the vehicle machine monitoring and management system receives the task requests for cargo delivery at different processing points in the system in real time, and the PC machine scheduling system receives the task requests and sends them The request information is sent to each AGV on-board monitoring and management system;

(2) After the AGV on-board monitoring and management system receives the task request information, it reads the current AGV running status and judges whether the current AGV is in the idle state: if it is not in the idle state, it will feedback that the current AGV is in the working state and does not receive the task request; if it is idle State, read the current position coordinate information of the AGV, and feed back the position coordinate information of the AGV to the PC dispatching system; the PC dispatching system calculates the weight of the delivery time of the AGV execution task path through the position coordinates of the requested task and the coordinate position of the AGV, according to The weight of the delivery time of the AGV execution task path sets the priority order of the AGV, selects the AGV with the highest priority to perform the task, and the AGV vehicle monitoring management system transmits the location information of the AGV and the location information of the task request to the path optimization system;

(3) After receiving the location information of the AGV, the route optimization system initializes and updates the information of the route map model in the database, records the current position of the AGV as the starting node P ₀ , and records the position of the processing station with the task request as the ending node Pn, The route optimization system _reads the node and road section information in the route map model, and adopts the optimal route algorithm based on the shortest transportation time, that is, the Dijkstra algorithm. The feasible path among them is evaluated according to the evaluation function and the optimal path is selected from it, and the path optimization system feeds back the information of the optimal path to the PC dispatching system;

(4) After the PC scheduling system receives the optimal path information fed back by the path optimization system, it converts the optimal path into motion control instructions that can be recognized by the vehicle-mounted industrial computer, and sequentially converts the task paths according to the priority order of the tasks. The motion control command is sent to the on-board control system, the on-board control system executes the action after receiving the motion command, and feeds back the information of the action execution to the on-board monitoring and management system, which reads the actual position information of the AGV in the laser navigation system , and compared with the difference between the action position information fed back by the vehicle control system, a linear interpolation strategy is adopted for the difference between the feedback action position and the actual position of the AGV, and fuzzy control is performed on the trajectory of the AGV. By adjusting the fuzzy controller Control parameters (k _α , k _β , k _γ ), where k _β is a proportional factor, k _α and k _γ are quantitative factors; after the AGV completes the task, it will feed back the task completion information to the PC scheduling system;

(5) After the PC scheduling system receives the task completion information, it sets a fixed waiting time value for the AGV. After the waiting time arrives, if there is a new task assignment, the AGV receives the task and completes it; if there is no new task assignment, the AGV Automatically start the automatic warehouse return program, assign the optimal warehouse return path, the AGV will automatically start and return to the warehouse, and release the resources of the nodes and road sections in the path.

In step (1), the corresponding key node P and road section L are established for the actual working environment of the logistics workshop, and the key node P and road section L in the map model are simplified on the basis of retaining the key delivery process. Establish a relatively sparse undirected map model G(P',L'), where P'={1,2,3...,n} is the set of nodes, that is, the node set, and L' is the arc set between nodes, namely Road section set, according to the information of P' and L', establish the distance matrix D=(d _ij ) _n*n of the undirected map, where,

Among them, L _ij represents the distance value between node P _i and node P _j ; +∞ means that node P _i is not adjacent to node P _j , and the distance is infinite; 0 means that node P _i coincides with node P _j .

In step (1), the optimized storage method refers to adopting the storage method of adjacency linked list to store to reduce the storage of useless matrix, and to establish a single-linked list of adjacency relationship for each node in the path map model, and list it The head pointer is stored in the form of a vector; the storage design of the key node P includes four fields: one is the serial number ID of the node, the other is the adjacent point field AdjacentP, which is used to store the nodes adjacent to the node P, and the third is the weight Domain WeightP is used to store the distance weight between node P and adjacent nodes, and the fourth is adjacent domain NextP, which is used to link the next node in the adjacency list of node P.

In step (3), all nodes in the map model are initialized, where the set of adjacent nodes is marked as set M, and the set of marked nodes in the optimal path that has been obtained is marked as set N, and set N contains only path nodes during initialization The starting node P ₀ in , the set of storage nodes waiting to be sorted in the set of adjacent nodes M is denoted as set R, for the nodes to be stored in the unordered sequence in the set R, that is, the nodes for which the optimal path has not been determined, it is assumed that there are n Unordered sequence nodes, store n unordered sequence nodes using heap sorting optimization to obtain a small top heap, create a one-dimensional array A[] to store n unordered sequence nodes, and perform minimum heap sorting on it, where Each node is stored in the storage order of a complete binary tree. The vertex of the tree stores the adjusted top node of the heap. The top node of the heap is used as the intermediate node, and the intermediate node is calculated from the starting node P The length weight of the optimal path reachable by each node in the complement set from ₀ to the set N of marked nodes through the intermediate node, and so on, and iteratively repeats for many times until all the nodes in the set M of adjacent nodes are added to the set In the set N of marked nodes, the shortest path set from the starting node P ₀ to any node P _k in the set M of adjacent nodes can be obtained, and P ₀ is selected as the starting node from the obtained shortest path set, _{Pn is} used as the path of the terminal node as the path for AGV to perform tasks.

In step (3), the formula of the evaluation function based on the distance evaluation function of the minimum delivery time is:

f(n)＝∑ω _ij +∑(f(θ _ij )+μ)

Among them, θ _ij is the turning angle of the AGV between node i and node j in the path, f(θ _ij ) is the turning time cost function of turning angle, μ is the cost coefficient of turning; the evaluation index of path walking is ∑ω _ij , ω _ij is the distance weight between node i and node j.

In step (3), described selecting optimal path according to Dijkstra algorithm comprises the following steps:

(6a) Initialization, where the optimal path node set N' only contains the starting node P ₀ , the unsorted node set R' contains all nodes in the adjacent node set M' except the set N'; and any The adjacent relationship and distance weight between two nodes are known, and the undirected graph and distance list of the map model in the system are initialized;

(6b) In the unsorted node set R', perform minimum heap sorting on the nodes to be sorted adjacent to the start node P ₀ , and select the node P _i with the smallest distance weight as the heap vertex, the start node P The distance weight between ₀ and node P _i is the distance weight of the shortest path, put node P _i into the set M', the distance weight of the optimal path from node P ₀ to node P _i is the delivery time , the shortest path is P ₀ ->P _i ;

(6c) Take the node P _i as the new intermediate node, select the nodes adjacent to the intermediate node P _i from the unsorted node set R', if the distance from the starting node P ₀ to the nodes in the set R' If the weight is smaller than the distance weight of the original path, modify the path that has been calculated through the intermediate node P _i in the unsorted node set R', where the modified path weight is the path that passes through the new intermediate node P _i Weight;

(6d) Repeat steps (6b) and (6c), iteratively process, traverse and search the unsorted nodes in the unsorted node set R', until all the nodes in the set R' are included to the final In the optimal path node set N' or until the optimal path is found between the starting node P ₀ and the ending node P _n .

In step (4), the vehicle-mounted industrial computer is the industrial computer carried on the AGV vehicle, and the vehicle-mounted control system includes a movement with the industrial computer as the upper computer, the ABB controller as the lower computer, and the closed-loop servo system as the execution module. The control system is used to control the steering, reversing, and entry and exit of the AGV.

In step (5), after the AGV completes the task of delivery and feeds back the task completion information, in order to avoid occupying the resources of nodes and road sections, after the waiting time arrives: if there is no task allocation, the AGV will start the automatic warehouse return program, and the PC will schedule The system sets the current position of the AGV as the starting node P' ₀ , and the idle position of the warehouse as P' _n , and the PC scheduling system plans the shortest path back to the warehouse for this AGV so that it can automatically return to the parking garage; after completion, The AGV releases the occupied node and road section resources; if there is a task assignment, the AGV accepts the task.

It can be seen from the above technical solution that the advantages of the present invention are: first, reorganize the detailed optimization steps of the AGV path planning system, remove the overly complicated process of the traditional AGV, and re-plan the AGV system on the basis of retaining key process nodes. Coordinated working relationship between modules; Second, the map model of the AGV system is established based on the actual environment of the workshop. On the basis of retaining the nodes of the key delivery process, the nodes and road sections in the map are simplified, and the nodes are stored in the database by calling the nodes. The complete data information that has been stored greatly improves the calling efficiency of node information; thirdly, optimizes the path search algorithm based on the shortest delivery time, that is, the Dijkstra algorithm, from the establishment of the node adjacency list, the storage method of the node structure, and the algorithm search way, etc., so as to greatly improve the stability and efficiency of the path planning system; fourth, the heap sorting method adopted for unordered nodes, obtain the minimum value of the small top heap, and use this vertex to correct the starting node through the middle The length of the shortest path reachable between a node and other adjacent nodes.

Description of drawings

Figure 1 is a system framework diagram of the AGV system;

Fig. 2 is a work flow chart of the present invention;

Figure 3 is a plan view of the actual working environment of the AGV;

FIG. 4 is a schematic structural diagram of an adjacency list node.

detailed description

As shown in Figure 2, an AGV path planning method based on the shortest path depth optimization algorithm, the method includes the following sequential steps:

(1) Initialize the AGV dispatch management system. The PC dispatch system reads the established path map model in the database, extracts the information of key nodes P and road sections L in the map, and stores the node information in an optimized storage method. At the same time, the PC The machine dispatching system reads all the AGV status information fed back by the laser navigation system, monitors and positions the AGV, and at the same time, the car machine monitoring and management system receives the task requests for cargo delivery at different processing points in the system in real time, and the PC machine dispatching system receives the task requests and sends them The request information is sent to each AGV on-board monitoring and management system;

(2) After the AGV on-board monitoring and management system receives the task request information, it reads the current AGV running status and judges whether the current AGV is in the idle state: if it is not in the idle state, it will feedback that the current AGV is in the working state and does not receive the task request; if it is idle State, read the current position coordinate information of the AGV, and feed back the position coordinate information of the AGV to the PC dispatching system; the PC dispatching system calculates the weight of the delivery time of the AGV execution task path through the position coordinates of the requested task and the coordinate position of the AGV, according to The weight of the delivery time of the AGV execution task path sets the priority order of the AGV, selects the AGV with the highest priority to perform the task, and the AGV vehicle monitoring management system transmits the location information of the AGV and the location information of the task request to the path optimization system;

(3) After receiving the location information of the AGV, the route optimization system initializes and updates the information of the route map model in the database, records the current position of the AGV as the starting node P ₀ , and records the position of the processing station with the task request as the ending node Pn, The route optimization system _reads the node and road section information in the route map model, and adopts the optimal route algorithm based on the shortest transportation time, that is, the Dijkstra algorithm. The feasible path among them is evaluated according to the evaluation function and the optimal path is selected from it, and the path optimization system feeds back the information of the optimal path to the PC dispatching system;

(4) After the PC scheduling system receives the optimal path information fed back by the path optimization system, it converts the optimal path into motion control instructions that can be recognized by the vehicle-mounted industrial computer, and sequentially converts the task paths according to the priority order of the tasks. The motion control command is sent to the on-board control system, the on-board control system executes the action after receiving the motion command, and feeds back the information of the action execution to the on-board monitoring and management system, which reads the actual position information of the AGV in the laser navigation system , and compared with the difference between the action position information fed back by the vehicle control system, a linear interpolation strategy is adopted for the difference between the feedback action position and the actual position of the AGV, and fuzzy control is performed on the trajectory of the AGV. By adjusting the fuzzy controller Control parameters (k _α , k _β , k _γ ), where k _β is a proportional factor, k _α and k _γ are quantitative factors; after the AGV completes the task, it will feed back the task completion information to the PC scheduling system;

(5) After the PC scheduling system receives the task completion information, it sets a fixed waiting time value for the AGV. After the waiting time arrives, if there is a new task assignment, the AGV receives the task and completes it; if there is no new task assignment, the AGV Automatically start the automatic warehouse return program, assign the optimal warehouse return path, the AGV will automatically start and return to the warehouse, and release the resources of the nodes and road sections in the path.

In step (1), the corresponding key node P and road section L are established for the actual working environment of the logistics workshop, and the key node P and road section L in the map model are simplified on the basis of retaining the key delivery process. Establish a relatively sparse undirected map model G(P',L'), where P'={1,2,3...,n} is the set of nodes, that is, the node set, and L' is the arc set between nodes, namely Road segment set, according to the information of P' and L', establish the distance matrix D=(d _ij ) _n*n of the undirected map, where,

Among them, L _ij represents the distance value between node P _i and node P _j ; +∞ means that node P _i is not adjacent to node P _j , and the distance is infinite; 0 means that node P _i coincides with node P _j .

As shown in Fig. 4, in step (1), described optimized storage mode refers to adopting the storage mode of adjacency linked list to store to reduce the storage of useless matrix, each node in the path map model is set up a single adjacency relationship Linked list, and its head pointer is stored in the form of vector; the storage design of the key node P includes four fields: one is the serial number ID of the node, and the other is the adjacent point field AdjacentP, which is used to store the data adjacent to the node P Node, the third is the weight field WeightP, which is used to store the distance weight between the node P and the adjacent node, and the fourth is the adjacency field NextP, which is used to link the next node in the adjacency list of the node P.

In step (3), all nodes in the map model are initialized, where the set of adjacent nodes is marked as set M, and the set of marked nodes in the optimal path that has been obtained is marked as set N, and set N contains only path nodes during initialization The starting node P ₀ in , the set of storage nodes waiting to be sorted in the set of adjacent nodes M is denoted as set R, for the nodes to be stored in the unordered sequence in the set R, that is, the nodes for which the optimal path has not been determined, it is assumed that there are n Unordered sequence nodes, store n unordered sequence nodes using heap sorting optimization to obtain a small top heap, create a one-dimensional array A[] to store n unordered sequence nodes, and perform minimum heap sorting on it, where Each node is stored in the storage order of a complete binary tree. The vertex of the tree stores the adjusted top node of the heap. The top node of the heap is used as the intermediate node, and the intermediate node is calculated from the starting node P The length weight of the optimal path reachable by each node in the complement set from ₀ to the set N of marked nodes through the intermediate node, and so on, and iteratively repeats for many times until all the nodes in the set M of adjacent nodes are added to the set In the set N of marked nodes, the shortest path set from the starting node P ₀ to any node P _k in the set M of adjacent nodes can be obtained, and P ₀ is selected as the starting node from the obtained shortest path set, _{Pn is} used as the path of the terminal node as the path for AGV to perform tasks.

In step (3), the formula of the evaluation function based on the distance evaluation function of the minimum delivery time is:

f(n)＝∑ω _ij +∑(f(θ _ij )+μ)

Among them, θ _ij is the turning angle of the AGV between node i and node j in the path, f(θ _ij ) is the turning time cost function of turning angle, μ is the cost coefficient of turning; the evaluation index of path walking is ∑ω _ij , ω _ij is the distance weight between node i and node j.

In step (3), described selecting optimal path according to Dijkstra algorithm comprises the following steps:

(6a) Initialization, where the optimal path node set N' only contains the starting node P ₀ , and the unsorted node set R' contains all nodes in the adjacent node set M' except the set N'; and any The adjacent relationship and distance weight between two nodes are known, and the undirected graph and distance list of the map model in the system are initialized;

(6b) In the unsorted node set R', perform minimum heap sorting on the nodes to be sorted adjacent to the start node P ₀ , and select the node P _i with the smallest distance weight as the heap vertex, the start node P The distance weight between ₀ and node P _i is the distance weight of the shortest path, put node P _i into the set M', the distance weight of the optimal path from node P ₀ to node P _i is the delivery time , the shortest path is P ₀ ->P _i ;

(6c) Take the node P _i as the new intermediate node, select the nodes adjacent to the intermediate node P _i from the unsorted node set R', if the distance from the starting node P ₀ to the nodes in the set R' If the weight is smaller than the distance weight of the original path, modify the path that has been calculated through the intermediate node P _i in the unsorted node set R', where the modified path weight is the path that passes through the new intermediate node P _i Weight;

(6d) Repeat steps (6b) and (6c), iteratively process, traverse and search the unsorted nodes in the unsorted node set R', until all the nodes in the set R' are included to the final In the optimal path node set N' or until the optimal path is found between the starting node P ₀ and the ending node P _n .

In step (4), the vehicle-mounted industrial computer is the industrial computer carried on the AGV vehicle, and the vehicle-mounted control system includes a movement with the industrial computer as the upper computer, the ABB controller as the lower computer, and the closed-loop servo system as the execution module. The control system is used to control the steering, reversing, and entry and exit of the AGV.

In step (5), after the AGV completes the task of delivery and feeds back the task completion information, in order to avoid occupying the resources of nodes and road sections, after the waiting time arrives: if there is no task allocation, the AGV will start the automatic warehouse return program, and the PC will schedule The system sets the current position of the AGV as the starting node P' ₀ , and the idle position of the warehouse as P' _n , and the PC scheduling system plans the shortest path back to the warehouse for this AGV so that it can automatically return to the parking garage; after completion, The AGV releases the occupied node and road section resources; if there is a task assignment, the AGV accepts the task.

As shown in Figure 1, the AGV system mainly includes PC scheduling management system, laser navigation system, vehicle machine servo control system, vehicle machine monitoring management system, etc. The vehicle machine monitoring management system reads the task request information of the processing station, accepts Task request, assign AGV to execute the task, the path planning system plans the path with the shortest time for the AGV by calling the node information in the database, the AGV executes the task and feeds back the task execution information to the PC scheduling management system in real time, and the PC scheduling management system passes Difference comparison, real-time adjustment to reduce the difference, so that the AGV can complete the task with high precision. When the task is completed, the AGV accepts the new task or executes the automatic warehouse return procedure.

As shown in Figure 3, Figure 3 is a node map model created for the actual working environment of the AGV. Key path nodes are established at the processing position, shelf position, detection position, cache position and charging maintenance position, and other unnecessary Nodes, establish a relatively sparse map model, thereby improving the operational stability of the system, shortening the search time of the algorithm and reducing the complexity of the algorithm.

In summary, the present invention reorganizes the detailed optimization steps of the AGV path planning system, removes the overly complicated process of the traditional AGV, and re-plans the coordination and working relationship between the various modules of the AGV system on the basis of retaining key process nodes; The map model of the AGV system is established based on the actual environment of the workshop. On the basis of retaining the nodes of the key transportation process, the nodes and road sections in the map are simplified, and the node information is greatly improved by calling the complete data information of the nodes already stored in the database. call efficiency.

Claims (8)

1. An AGV path planning method based on the shortest path depth optimization algorithm, characterized in that: the method comprises the steps of the following order: (1) Initialize the AGV dispatch management system. The PC dispatch system reads the established path map model in the database, extracts the information of key nodes P and road sections L in the map, and stores the node information in an optimized storage method. At the same time, the PC The machine dispatching system reads all the AGV status information fed back by the laser navigation system, monitors and positions the AGV, and at the same time, the car machine monitoring and management system receives the task requests for cargo delivery at different processing points in the system in real time, and the PC machine dispatching system receives the task requests and sends them The request information is sent to each AGV on-board monitoring and management system; (2) After the AGV on-board monitoring and management system receives the task request information, it reads the current AGV running status and judges whether the current AGV is in an idle state: if it is not in an idle state, it will feedback that the current AGV is in a working state and does not receive a task request; if it is idle State, read the current position coordinate information of the AGV, and feed back the position coordinate information of the AGV to the PC scheduling system; the PC scheduling system calculates the weight of the delivery time of the AGV execution task path through the position coordinates of the requested task and the AGV coordinate position, according to The weight of the delivery time of the AGV execution task path sets the priority order of the AGV, selects the AGV with the highest priority to perform the task, and the AGV vehicle monitoring management system transmits the location information of the AGV and the location information of the task request to the path optimization system; (3) After receiving the location information of the AGV, the route optimization system initializes and updates the information of the route map model in the database, records the current position of the AGV as the starting node P ₀ , and records the position of the processing station with the task request as the ending node Pn, The route optimization system _reads the node and road section information in the route map model, and adopts the optimal route algorithm based on the shortest transportation time, that is, the Dijkstra algorithm. The feasible path among them is evaluated according to the evaluation function and the optimal path is selected from it, and the path optimization system feeds back the information of the optimal path to the PC dispatching system; (4) After the PC scheduling system receives the optimal path information fed back by the path optimization system, it converts the optimal path into motion control instructions that can be recognized by the vehicle-mounted industrial computer, and sequentially converts the task paths according to the priority order of the tasks. The motion control command is sent to the on-board control system, the on-board control system executes the action after receiving the motion command, and feeds back the information of the action execution to the on-board monitoring and management system, which reads the actual position information of the AGV in the laser navigation system , and compared with the difference between the action position information fed back by the vehicle control system, a linear interpolation strategy is adopted for the difference between the feedback action position and the actual position of the AGV, and fuzzy control is performed on the trajectory of the AGV. By adjusting the fuzzy controller Control parameters (k _α , k _β , k _γ ), where k _β is a proportional factor, k _α and k _γ are quantitative factors; after the AGV completes the task, it will feed back the task completion information to the PC scheduling system; (5) After the PC scheduling system receives the task completion information, it sets a fixed waiting time value for the AGV. After the waiting time arrives, if there is a new task assignment, the AGV receives the task and completes it; if there is no new task assignment, the AGV Automatically start the automatic warehouse return program, assign the optimal warehouse return path, the AGV will automatically start and return to the warehouse, and release the resources of the nodes and road sections in the path. 2. the AGV path planning method based on the shortest path depth optimization algorithm according to claim 1, is characterized in that: in step (1), set up corresponding key node P and road section for the actual working environment of logistics workshop L, on the basis of retaining the key transportation process, simplify the key node P and road section L in the map model, and establish a relatively evacuated undirected map model G(P',L'), where P'={1,2 ,3...,n} is the set of nodes, i.e. the node set, L' is the arc set between the nodes, i.e. the road section set, and the distance matrix D=(d _ij ) _n of the undirected map is established according to the information of P' and L' _*n , where Among them, L _ij represents the distance value between node P _i and node P _j ; +∞ means that node P _i is not adjacent to node P _j , and the distance is infinite; 0 means that node P _i coincides with node P _j . 3. the AGV path planning method based on the shortest path depth optimization algorithm according to claim 1, is characterized in that: in step (1), described optimization storage mode refers to adopting the storage mode of adjacency linked list to store to reduce useless For the storage of the matrix, a single-linked list of adjacency relationship is established for each node in the path map model, and its head pointer is stored in the form of a vector; the storage design of the key node P includes four fields: one is the sequence of nodes The second is the adjacent point field AdjacentP, which is used to store the nodes adjacent to the node P, the third is the weight field WeightP, which is used to store the distance weight between the node P and the adjacent node, and the fourth is the adjacent field NextP, The next node in the adjacency list for linking node P. 4. the AGV path planning method based on the shortest path depth optimization algorithm according to claim 1, is characterized in that: in step (3), all nodes in the initialization map model, wherein adjacent node set is denoted as set M, has Find the set of marked nodes in the optimal path and record it as set N. At the time of initialization, set N only includes the starting node P ₀ in the path node, and the set of storage nodes waiting to be sorted in the set of adjacent nodes M is marked as set R. For the nodes to be stored in the unordered sequence in the set R, that is, the node for which the optimal path has not been determined, assuming that there are n nodes in the unordered sequence, the n nodes in the unordered sequence are optimally stored by heap sorting to obtain a small top heap, and a new A one-dimensional array A[] stores n unordered sequence nodes, and minimizes heap sorting. Each node is stored in the storage order of a complete binary tree, and the vertices of the tree are stored. After the top node of the heap, the top node of the heap is used as the intermediate node, and the optimal path of each node in the complement set from the starting node P ₀ to the set N of marked nodes is calculated through the intermediate node. Length weight, and so on, all nodes in the set M of adjacent nodes are added to the set N of marked nodes through repeated iterations, so that any one of the set M from the starting node P ₀ to the adjacent nodes can be obtained The shortest path set of node P _k , and select the path with P ₀ as the starting node and P _n as the terminating node from the obtained shortest path set as the path for AGV to perform tasks. 5. the AGV path planning method based on the shortest path depth optimization algorithm according to claim 1, is characterized in that: in step (3), described evaluation is based on the formula of the distance evaluation function of minimum delivery duration: f(n)＝∑ω _ij +∑(f(θ _ij )+μ) Among them, θ _ij is the turning angle of the AGV between node i and node j in the path, f(θ _ij ) is the turning time cost function of turning angle, μ is the cost coefficient of turning; the evaluation index of path walking is ∑ω _ij , ω _ij is the distance weight between node i and node j. 6. the AGV path planning method based on the shortest path depth optimization algorithm according to claim 1, is characterized in that: in step (3), described selecting optimum path according to Dijkstra algorithm comprises the following steps: (6a) Initialization, where the optimal path node set N' only contains the starting node P ₀ , and the unsorted node set R' contains all nodes in the adjacent node set M' except the set N'; and any The adjacent relationship and distance weight between two nodes are known, and the undirected graph and distance list of the map model in the system are initialized; (6b) In the unsorted node set R', perform minimum heap sorting on the nodes to be sorted adjacent to the start node P ₀ , and select the node P _i with the smallest distance weight as the heap vertex, the start node P The distance weight between ₀ and node P _i is the distance weight of the shortest path, put node P _i into the set M', the distance weight of the optimal path from node P ₀ to node P _i is the delivery time , the shortest path is P ₀ ->P _i ; (6c) Take the node P _i as the new intermediate node, select the nodes adjacent to the intermediate node P _i from the unsorted node set R', if the distance from the starting node P ₀ to the nodes in the set R' If the weight is smaller than the distance weight of the original path, modify the path that has been calculated through the intermediate node P _i in the unsorted node set R', where the modified path weight is the path that passes through the new intermediate node P _i Weight; (6d) Repeat steps (6b) and (6c), iteratively process, traverse and search the unsorted nodes in the unsorted node set R', until all the nodes in the set R' are included to the final In the optimal path node set N' or until the optimal path is found between the starting node P ₀ and the ending node P _n . 7. the AGV path planning method based on the shortest path depth optimization algorithm according to claim 1, is characterized in that: in step (4), the vehicle-mounted industrial computer is the industrial computer carried on the AGV car, and the vehicle-mounted The control system includes a motion control system with the industrial computer as the upper computer, the ABB controller as the lower computer, and the closed-loop servo system as the execution module, which are used to control the steering, reversing, and entry and exit of the AGV. 8. the AGV path planning method based on the shortest path depth optimization algorithm according to claim 1, is characterized in that: in step (5), after AGV completes the task of conveying, feedback task completion information, in order to avoid occupying node and road section resources, after the waiting time is up: if there is no task assigned, the AGV will start the automatic warehouse return procedure, and the PC scheduling system will set the current position of the AGV as the starting node P' ₀ , the idle position of the warehouse as P' _n , and the PC The computer scheduling system plans the shortest path back to the warehouse for this AGV, so that it can automatically return to the parking garage; after completion, the AGV releases the occupied node and road section resources; if there is a task assignment, the AGV accepts the task.