Complex workshop multi-AGV collision avoidance path planning method based on space-time grid depth exploration

Abstract

The invention provides a complex workshop multi-AGV collision avoidance path planning method based on space-time grid depth exploration, and belongs to the technical field of intelligent manufacturing and automatic control, and the method comprises the steps: defining a parameter set for each trolley, and initializing multi-AGV scheduling parameters; constructing a workshop rasterized map model; the coordinates are mapped into a two-dimensional array, and a path-finding maze is built; receiving a transportation task instruction, and analyzing information of a starting node and a target node; judging whether the current instruction can be executed immediately or not, and if the current instruction cannot be executed immediately, waiting until available AGVs exist; the path planning efficiency is optimized by combining dynamic obstacle marking, space-time collision detection and direction weight probability search, path planning is executed, and the AGV state is updated; and displaying the car path as a space-time trajectory, performing time analysis on a space intersection point, analyzing whether collision occurs or not, and triggering real-time path re-planning. According to the method, the problems of multi-agent collaborative scheduling and collision avoidance during complex workshop path planning are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent manufacturing and automatic control, and particularly relates to a multi-AGV collision avoidance path planning method for complex workshops based on spatio-temporal grid depth exploration, which is particularly suitable for the intelligent workshop environment with multi-level distributed management and control. The present invention improves the multi-agent cooperation algorithm to realize multi-dimensional optimization such as material transportation, equipment scheduling, and path planning in the workshop, and significantly improves production efficiency and resource utilization rate. Background Art

[0002] With the rapid development of Industry 4.0 and intelligent manufacturing, the intelligent workshop, as the core unit of modern manufacturing, has become an important carrier for realizing efficient production. The intelligent workshop realizes the high automation and intelligence of the production process by integrating technologies such as the Internet of Things, big data, and artificial intelligence. However, due to the complexity and dynamics of the workshop environment, traditional centralized management methods are unable to cope with problems such as multi-AGV (Automated Guided Vehicle) cooperative scheduling and path planning.

[0003] In the prior art, the A algorithm, as a classic path planning algorithm, although it performs excellently in a static road network, has the following problems in the actual workshop environment:

[0004] (1) Single steering angle: The traditional A algorithm only supports path planning in 8 directions and cannot meet the requirements of multi-angle paths in the actual workshop.

[0005] (2) Low multi-AGV scheduling efficiency: The traditional A algorithm is only applicable to single-AGV path planning and is difficult to efficiently handle the problem of multi-AGV cooperative scheduling.

[0006] (3) Path conflict and collision risks: When multiple AGVs are running, path conflicts and collisions are difficult to avoid, resulting in a decrease in the operating efficiency of the workshop.

[0007] (4) Insufficient algorithm efficiency: In a complex workshop environment, the traditional A algorithm has a large amount of calculation and is difficult to meet the requirements of real-time scheduling.

[0008] Therefore, there is an urgent need for an improved multi-agent cooperation algorithm to realize the efficient scheduling and path planning of multiple AGVs in the intelligent workshop. Summary of the Invention

[0009] To solve the above technical problems, the present invention proposes a multi-AGV collision avoidance path planning method for complex workshops based on spatio-temporal grid depth exploration, which specifically includes the following technical solutions:

[0010] The specific technical solution is as follows: A complex workshop multi-AGV collision avoidance path planning method based on spatio-temporal grid depth exploration, and the method includes the following steps:

[0011] Step 1: Define a parameter set for each vehicle, and initialize the multi-AGV scheduling parameters; construct a workshop grid map model according to the actual terrain of the workshop; map the coordinates into a two-dimensional array, and build a pathfinding maze;

[0012] Step 2: Receive the transportation task instruction, and parse the information of the starting node and the target node; determine whether the current instruction can be executed immediately. If it cannot be executed immediately, wait for a period of time until there is an available AGV; if the instruction can be executed immediately, go to Step 3;

[0013] Step 3: Generate a path planning result based on the extended steering angle, multi-AGV collaborative scheduling, collision avoidance, and weighted path search;

[0014] Step 4: Execute the path planning result, update the AGV status and dynamic obstacle information in real time, and complete the task scheduling;

[0015] Step 5: Display the AGV path as a spatio-temporal trajectory, perform time analysis on the spatial intersection points, analyze whether there is a collision, and trigger real-time path replanning.

[0016] The present invention has the following beneficial effects:

[0017] Extended steering angle: The traditional algorithm can only select eight directions, namely up, down, left, right, and the diagonals. It is too mechanical and single when planning paths. In the present invention, through the expansion of neighborhood nodes, angles such as ±27°, ±63°, ±117°, ±153° are added to the original eight directions, making the path planning of the vehicle more flexible and diverse, and the angle steering more diversified.

[0018] Dynamic obstacle and collision avoidance algorithm: By setting dynamic obstacles, the obstacles are taken into consideration when calling the algorithm, and the running status of the AGV is monitored in real time to generate a suitable path to avoid path conflicts and collisions during the operation of multiple AGVs.

[0019] Multi-AGV scheduling algorithm: Each AGV is given independent attributes (such as size, speed, starting position, etc.), and the algorithm of the present invention is called for path planning according to its specific attributes. The AGV closest to the starting point and available is preferentially scheduled to reduce the waiting time and improve the scheduling efficiency.

[0020] Weighted path search: According to the relative position relationship between the starting point and the target point, weights are assigned to the search probabilities of neighborhood nodes to increase the search probability towards the optimal path and optimize the algorithm efficiency.

[0021] Improve path planning flexibility: The algorithm of the present invention expands the steering angle, making path planning more diverse and adaptable to complex workshop environments. Enhance the multi-AGV scheduling efficiency: Through dynamic obstacle handling and multi-AGV collaborative scheduling algorithms, significantly improve the workshop operation efficiency and reduce waiting time. Reduce the collision risk: Adopt a collision avoidance algorithm to effectively avoid path conflicts and collisions during the operation of multiple AGVs. Optimize the algorithm efficiency: The weighted path search algorithm significantly reduces the computational amount, improves the algorithm operation efficiency, and meets the real-time scheduling requirements of the workshop. Description of the Drawings

[0022] Figure 1 It is a flowchart of the complex workshop multi-AGV collision avoidance path planning method based on spatio-temporal grid depth exploration of the present invention;

[0023] Figure 2 It is a flowchart of the application scenario of the present invention;

[0024] Figure 3 It is the path of the traditional algorithm;

[0025] Figure 4 It is the path of the present invention;

[0026] Figure 5 It is a comparison chart of operation efficiency;

[0027] Figure 6 It is another comparison chart of operation efficiency;

[0028] Figure 7 It is a flowchart for the vehicle to avoid collision;

[0029] Figure 8 It is a visualization result chart of the operation path. Detailed Embodiments

[0030] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above objectives, the present invention adopts the following technical solutions.

[0031] The present invention provides a complex workshop multi-AGV collision avoidance path planning method based on spatio-temporal grid depth exploration, as Figure 1 , Figure 2 shown, and the method includes the following steps:

[0032] Step 1: Define a parameter set for each trolley, and initialize multi-AGV scheduling parameters, including trolley size, speed, available status, and dynamic obstacle identification; according to the actual terrain of the workshop, construct a grid map model of the workshop, and set the obstacle attribute with 0 / 1 variables; map the coordinates into a two-dimensional array and build a pathfinding maze. The specific mapping formula is as follows:

[0033] ;

[0034] Among them, is the distance from the map boundary coordinates to the origin coordinates, is the expansion factor. The size of the expansion factor determines the refinement degree of the workshop grid and is closely related to the algorithm efficiency.

[0035] Step 2: Receive the material transportation task instruction from the scheduling center, and parse the start node and target node information; determine whether the current instruction can be executed immediately. If it cannot be executed immediately, for example, when there is no available AGV at present, wait for a period of time until there is an available AGV; if the instruction can be executed immediately, go to S3.

[0036] Step 3: Generate a path based on the algorithm of the present invention. The improvements include expanding the steering angle, multi-AGV collaborative scheduling, collision avoidance mechanism, and weighted path search.

[0037] Step 4: Execute the path planning result, and update the AGV status and dynamic obstacle information in real time to complete task scheduling.

[0038] Step 5: The path visualization module displays the AGV path as a spatio-temporal trajectory, performs time analysis on spatial intersections, analyzes whether there is a collision, and triggers real-time path replanning.

[0039] The specific content of Step 3 includes:

[0040] Step 3-1: Add the second circle of 16 neighborhood nodes on the basis of the 8-neighborhood of the traditional A algorithm to achieve the expansion of the steering angles of ±27°, ±63°, ±117°, and ±153°;

[0041] Step 3-2: Assign an independent attribute set to each AGV, including real-time position, speed, path, and dynamic obstacle identification, and preferentially select an available AGV using the Euclidean distance;

[0042] Step 3-3: Achieve collision avoidance through dynamic obstacle setting and departure time adjustment;

[0043] Step 3-4: Apply probability weights to the neighborhood node search direction based on the target node orientation to optimize the calculation efficiency;

[0044] The specific content of Step 3-1 includes:

[0045] First, define the second - ring neighborhood nodes of the current node as , where n ∈ {0, 1, 2}, to achieve the expansion of 8 - neighborhood nodes; then calculate the Manhattan distance between each neighborhood node and the target node, and establish a 16 - direction angle expansion model; divide the neighborhood nodes into left and right node sets, make an initial orientation judgment, and select the actual neighborhood node set; based on the determined neighborhood nodes, perform path selection through the cost function .

[0046] The multi - AGV collaborative scheduling in step 3 - 2 is implemented as follows:

[0047] Create an independent attribute set for each AGV, including vehicle name, size specification, real - time speed, path coordinate sequence, and dynamic obstacle marker;

[0048] Adopt the Euclidean distance formula:

[0049] ;

[0050] Sort the small vehicles by distance, comprehensively consider factors such as vehicle size and whether they are idle, select the nearest small vehicle for the scheduling task, generate a vehicle - specific path planning space, and update the vehicle status.

[0051] The collision avoidance mechanism in step 3 - 3 is implemented as follows: Conduct spatio - temporal analysis on the paths of the scheduled AGVs, and mark the areas where vehicle_path and vehicle_time are located in the map;

[0052] When planning the paths of subsequent AGVs, consider whether the current planned path overlaps with the previous path. The judgment criteria are as follows:

[0053] ;

[0054] ;

[0055] Among them, is the vehicle size, The function is used to judge spatial overlap. When there is a spatial intersection at any part of the two vehicles, it is considered that a collision has occurred; for time overlap, take an interval. If two vehicles pass through the same location and are very close in time, it is considered that a collision has occurred. The calculation method of the time threshold is as follows:

[0056] ;

[0057] Here, a small AGV is taken for calculation, that is, the length is , That is, when the speed is , the calculated threshold is approximately .

[0058] In addition, when either of the Boolean variables Time and Space is true, the departure time of the AGV is adjusted or the path is re-planned until there is no longer a collision.

[0059] The weighted path search in step 3-4 is implemented as follows: Based on the initial starting and target nodes, an orientation determination function is established:

[0060] ;

[0061] Set the selection probability of neighborhood nodes:

[0062] ;

[0063] Here, the selection probability can be modified according to the actual map. When the target node is on the right side of the current node, the probability of selecting the right-side node set as the neighborhood nodes is greater; when the target node is on the left side of the current node, the probability of selecting the right-side node set as the neighborhood nodes is smaller.

[0064] Dynamically adjust the expansion order of the OPEN table nodes based on the probability distribution, and preferentially expand the neighborhood nodes in the target direction.

[0065] Step S3-1: Expand the steering angle

[0066] The 8-neighborhood nodes of the first circle of the current node are specifically: the neighborhood nodes in the up, down, left, right, upper left, lower left, upper right, and lower right eight directions. On this basis, 16 neighborhood nodes of the second circle of the current node are added, that is, points such as (x + 2, y), (x + 2, y + 1), (x + 2, y - 1), so that the angles traveled by the trolley are expanded, including the eight positions of ±27°, ±63°, ±117°, and ±153°. Figure 3 , Figure 4 is a comparison chart of the traditional algorithm and this algorithm. It can be seen that the improved algorithm has better linearity and the running angle is also more in line with the actual shortest path (that is, the line segment between two points is the shortest).

[0067] Step S3-2: Implement the multi-AGV scheduling algorithm

[0068] The main idea of this algorithm to implement multi-AGV scheduling is to add multiple status flags to each trolley, and obtain the trolley running information and complete the scheduling by reading and judging the status flags.

[0069] Each trolley class mainly includes the following attributes:

[0070] Name: The name of the vehicle, such as "Small Vehicle 1", "Large Vehicle 3", etc.;

[0071] Size: The size of the vehicle. For large vehicles, it is 2 For the 1.5 specification, for small vehicles, it is 1 For the 1 specification (unit: m);

[0072] Speed: For large vehicles, it is 0.45 m / s, and for small vehicles, it is 0.65 m / s;

[0073] start_pos, goal_pos: The names of the starting and target nodes of the vehicle in the input instruction;

[0074] Path: Used to store the path planned by the vehicle using the algorithm of the present invention;

[0075] Time: Used to store the time when the vehicle reaches each point;

[0076] Ordernumber: The instruction number for calling the vehicle;

[0077] vehicle_obstacle: Used to set dynamic obstacles for the vehicles called later;

[0078] Other flag bits, such as: useful (to judge whether the vehicle is available), nouse (to judge whether the vehicle is currently running), current_pos (to update the current specific position of the vehicle).

[0079] By judging the current state of the vehicle, the program will preferentially select the vehicle that is closest to the starting node in terms of Euclidean distance and available for scheduling. By calling the algorithm of the present invention, the target node and the current node are input to generate a path and store the path of the vehicle. Visualization is performed when the end point is reached.

[0080] Step S3-3: Add an anti-collision algorithm

[0081] Since the instructions are input one by one, there is a sequence relationship in the order of calling the vehicles. For the first vehicle to be scheduled, it can directly use the algorithm of the present invention and store the path. For the vehicles scheduled later, the flowchart for monitoring and avoiding collisions is as Figure 7 .

[0082] If the path planned by the car overlaps with the previous car path in time and space, the car will take measures to delay departure to stagger the time and avoid collision; in addition, since the car is in constant motion, in order to avoid setting all points on the car's running path as obstacles, a method of dynamically selecting obstacles is adopted here, that is, setting certain segments on the car's path as obstacles. When the current car has not reached the end, the dynamic obstacles are retained and affect the calculation of the path of the later scheduled car using the algorithm of the present invention. When the car reaches the end, the dynamic obstacles are cleared and the state is restored. The final path of the car will be stored in the path array of the car in the form of coordinates, and the overall path of the car can be obtained by backtracking from the end. Figure 5 This is a comparison chart of operating efficiency; Figure 6 Another operating efficiency comparison chart is shown below.

[0083] After completing the above steps, call this algorithm and perform visualization. In order to intuitively display the paths of different cars and whether they collide, this algorithm displays all paths on a graph. At the same time, in order to intuitively reflect the running time of the car (that is, the running time of the car in this step) to determine whether there is a collision, the spatial intersection points are marked and their respective times are displayed. If the time difference between the spatial intersections of the cars is within the threshold, it means that the two have collided.

[0084] The final path visualization result is as follows Figure 8 .

[0085] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A complex workshop multi-AGV collision avoidance path planning method based on spatio-temporal grid depth exploration, characterized in that, The method includes the following steps: Step 1: Define a parameter set for each trolley, and initialize the multi-AGV scheduling parameters; according to the actual terrain of the workshop, construct a grid map model of the workshop; map the coordinates into a two-dimensional array, and build a pathfinding maze; Step 2: Receive the transportation task instruction, and parse the information of the starting node and the target node; determine whether the current instruction can be executed immediately. If it cannot be executed immediately, wait for a period of time until there is an available AGV; If the instruction can be executed immediately, go to Step 3; Step 3: Generate a path planning result based on the extended steering angle, multi-AGV collaborative scheduling, collision avoidance, and weighted path search; Step 4: Execute the path planning result, and update the AGV status and dynamic obstacle information in real time to complete the task scheduling; Step 5: Display the AGV path as a spatio-temporal trajectory, perform time analysis on the spatial intersection points, analyze whether there is a collision, and trigger real-time path replanning.

2. A method for collision avoidance path planning of multiple AGVs in a complex workshop based on spatio-temporal grid depth exploration according to claim 1, characterized in that, The specific content of Step 3 includes: Step 3-1: Define the OPEN table as a set of points to be inspected. Take the current node as the parent node. Define the neighborhood nodes in the first circle as the eight nodes above, below, left, right, upper left, upper right, lower left, and lower right of the current node, and add them to the OPEN table as objects to be inspected. According to the designed cost function, select the node with the minimum cost from the neighborhood nodes as the next arrival position. Extend the coordinates based on the neighborhood nodes in the first circle, and add the neighborhood nodes in the second circle to achieve the extension of the steering angles of ±27°, ±63°, ±117°, and ±153°; Step 3-2: Assign an independent attribute set to each AGV, and use the Euclidean distance to preferentially select an available AGV for multi-AGV collaborative scheduling; Step 3-3: Achieve collision avoidance through dynamic obstacle setting and departure time adjustment; Step 3-4: Apply a probability weight to the search direction of the neighborhood nodes based on the orientation of the target node, and optimize the calculation efficiency for weighted path search.

3. A method for collision avoidance path planning of multiple AGVs in a complex workshop based on spatio-temporal grid depth exploration according to claim 2, characterized in that, The specific content of Step 3-1 includes: First, define the current node All the neighborhood nodes of are ; then calculate the Manhattan distance between each neighborhood node and the target node, and establish a 16-direction angle expansion model; divide the neighborhood nodes into left and right node sets, make an initial orientation judgment, and select the actual neighborhood node set; based on the determined neighborhood nodes, select a path through the cost function ; among them, represents the estimated distance from the initial position through node n to the target position; is the actual distance from the initial position to the current position; is the estimated distance of the best path from the current position to the target position.

4. A complex workshop multi-AGV collision avoidance path planning method based on spatio-temporal grid depth exploration according to claim 2, characterized in that, The implementation of the multi-AGV collaborative scheduling in Step 3-2 is as follows: Create an independent attribute set for each AGV; Use the Euclidean distance formula: ； Among them, , , respectively represent the horizontal and vertical coordinate values of the current node and the target node, is the distance value between two points; Sort the AGVs by distance, select the nearest trolley for the scheduling task, generate a vehicle-specific path planning space, and update the AGV status.

5. A complex workshop multi-AGV collision avoidance path planning method based on spatio-temporal grid depth exploration according to claim 2, characterized in that, The implementation of the collision avoidance in Step 3-3 is as follows: Perform spatio-temporal analysis on the scheduled AGV path and mark the area where and are located in the map; When planning the path for the AGV, consider whether the current planned path overlaps with the previous path. The judgment criteria are as follows: ； ； Among them, represents the time when the current trolley reaches a certain position; represents the time when the running trolley reaches this position; respectively represent the path coordinates of the current trolley and the running trolley, The function is used to judge spatial overlap, represents a boolean variable to judge whether there is time overlap; is also a boolean variable to judge whether there is spatial overlap; when any part of the two AGVs has a spatial intersection, it is considered that a collision occurs; for time overlap, an interval is taken. When the two AGVs pass by the same location very closely, it is considered that a collision occurs; In addition, when either the boolean variables Time or Space is true, adjust the AGV departure time or replan the path until there is no collision.

6. A method for collision avoidance path planning of multiple AGVs in a complex workshop based on spatio-temporal grid depth exploration according to claim 2, characterized in that, The implementation of the weighted path search in Step 3-4 is as follows: Based on the initial starting and target nodes, establish an orientation determination function: ； Among them, represents the abscissa of the target node, represents the abscissa of the current node, represents the determined orientation; Set the probability of selecting neighborhood nodes: ； representing a previously selected set of right nodes, representing the probability of using the set of right nodes as a neighborhood node set; When the target node is on the right side of the current node, the probability of selecting the set of right-side nodes as neighborhood nodes is greater; when the target node is on the left side of the current node, the probability of selecting the set of right-side nodes as neighborhood nodes is smaller; Dynamically adjust the expansion order of the openList nodes based on the probability distribution, and preferentially expand the neighborhood nodes in the target direction.

7. A complex workshop multi-AGV collision avoidance path planning method based on spatio-temporal grid depth exploration according to claim 1, characterized in that, The multi-AGV scheduling parameters include the trolley size, speed, available status, and dynamic obstacle identification.

8. A method for collision avoidance path planning of multiple AGVs in a complex workshop based on spatio-temporal grid depth exploration according to claim 1, characterized in that, When establishing the workshop grid map model, a 0 / 1 variable is used to set whether it is an obstacle attribute.

9. A method for collision avoidance path planning of multiple AGVs in a complex workshop based on spatio-temporal grid depth exploration according to claim 1, characterized in that, The coordinates are mapped to a two-dimensional array, and the specific mapping formula is as follows: ； Among them, is the distance between the map boundary coordinates and the origin coordinates, is the expansion factor, which is used to set the grid density, is the original coordinate, is the coordinate after mapping.

10. A method for collision avoidance path planning of multiple AGVs in a complex workshop based on spatio-temporal grid depth exploration, characterized in that, The independent attribute set includes the vehicle name, size specification, real-time speed, path coordinate sequence, and dynamic obstacle marker.

Images