A static path planning method for warehouse AGV based on improved artificial potential field method

Abstract

The present invention relates to the field of computer science and optimization algorithms, and to a static path planning method for a storage AGV based on an improved artificial potential field method. The method models an unmanned storage environment through a grid method, and sets the map size and static obstacle coordinates. First, before confirming the starting and target nodes, the repulsion of static obstacles is calculated for all feasible areas. When multiple obstacles are adjacent, they are merged into a whole, and the direction of the repulsion is calculated by the centroid and the magnitude of the repulsion is calculated by the edge distance; if they are not adjacent, the repulsion of each obstacle is calculated separately and attenuated. After confirming the starting and target nodes, a constant gravitational force is generated, and the connectivity value of the feasible area is calculated by the BFS algorithm, and the connectivity value is used to adjust the gravitational force. Finally, combining the gravitational force, repulsion and connectivity value, select the node whose resultant angle is close to the target node and whose resultant force is larger and is pushed into the stack until the starting node is pushed into the stack, and finally output the node in the stack as the optimal path.

Description

Technical Field

[0001] The present invention belongs to the field of computer science and optimization algorithms, and in particular relates to a static path planning method for a storage AGV based on an improved artificial potential field method. Background Art

[0002] As an unmanned mobile platform suitable for ground operations in unmanned warehousing environments, AGV can autonomously plan driving paths based on environmental information. It is small and light, highly flexible, highly economical, and highly adaptable to complex environments. Its low cost, high efficiency, and intelligence have made it widely used in modern industry and logistics, and it is rapidly developing in the direction of intelligence. With the increasing demand for unmanned warehousing environments due to industrial automation and technological progress, it has become a top priority to further improve the intelligence and autonomy of AGV. AGV's path planning technology senses and analyzes external environmental information to plan a driving trajectory that meets safety standards. It is a key factor in ensuring that AGV can successfully complete its tasks in an unmanned warehousing environment, and directly affects its work efficiency and technological advancement.

[0003] The algorithms currently used in the field of unmanned equipment path planning all have obvious shortcomings. For example, the Dijkstra algorithm is known for its simple implementation and good robustness, but due to its high spatiotemporal complexity and low computational efficiency, it performs poorly when dealing with long-distance path planning in large-scale warehousing environments. The A* algorithm is known for its strong scalability and outstanding search capabilities, but due to its reliance on heuristic functions, its solution speed is slow, making it perform poorly in real-time obstacle avoidance scenarios. In contrast, the simulated annealing algorithm performs well in complex path solving and can escape local optimal solutions, but its convergence speed is slow and it is sensitive to parameter settings, making it difficult to cope with dynamically changing warehousing environments. Another commonly used algorithm is the rapidly expanding random tree (RRT), which has strong randomness in path search and can effectively handle path planning in high-dimensional space, but due to the randomness in the algorithm, the generated path may not be optimal and has limited efficiency in larger warehousing environments.

[0004] Among the path planning algorithms for AGVs in unmanned warehouse environments, the Artificial Potential Field (APF) method is widely used in AGV path planning because of its simple model design, convenient calculation, and certain robustness to control and sensor errors. This method assumes that the target point exerts gravity on the AGV and the obstacles exert repulsion on it, and uses the combined force of gravity and repulsion to form a virtual potential field to guide the AGV to move from the starting point to the target point. However, the traditional artificial potential field method has some limitations in the warehouse environment. First, the map construction accuracy is insufficient, which may cause the AGV to fail in the obstacle avoidance process; second, when the AGV approaches the target point, it is easy to wander around the target point, resulting in the failure to reach the target in the end; in addition, the AGV may fall into the "local potential energy minimum point" (local optimal solution), causing it to stagnate or be unable to move forward. Therefore, in warehouse operations, when using this method, the AGV sometimes cannot reach the designated target smoothly, affecting the operation efficiency and the effectiveness of path planning.

[0005] Through the above analysis, the specific technical problems faced by the traditional artificial potential field method (APF) and most improved related algorithms in the global path planning problem are:

[0006] (1) The target is unreachable: In the traditional artificial potential field method, as the AGV gradually approaches the target point, the gravitational function gradually decreases to zero, making it difficult for the AGV to accurately reach the target point. In addition, when there are static obstacles near the target point, the repulsive function will gradually increase, causing the direction of the resultant force on the AGV to be opposite to the target point, making it impossible for the AGV to reach the target position smoothly;

[0007] (2) Local optimal solution: The traditional artificial potential field method can easily cause the AGV to fall into a local optimal point. When obstacles are vertically distributed on the line connecting the AGV and the target point, the calculated direction of the resultant force may almost coincide with the direction of the line, causing the AGV to stagnate at a certain position during driving and unable to move forward. In addition, when multiple obstacles are densely distributed near the AGV's motion path, the path generated by the traditional artificial potential field method may experience "oscillations", causing the AGV to fall into a local optimal point in the middle of the path and fail to successfully complete the scheduled driving task.

[0008] In view of the above problems, the improvement of artificial potential field method should be aimed at improving the ability to find the optimal solution for the target node and avoiding falling into the local optimal solution. Whether these technical problems can be solved is an important criterion for reflecting the advancement and scientificity of the improved artificial potential field algorithm in the path planning of AGV in an unmanned warehouse environment.

[0009] In the application scenario, the challenges faced by the traditional artificial market method in global path planning are reflected in:

[0010] (1) Challenges of target node accessibility. In unmanned warehouse environments, AGVs usually need to accurately reach designated target points. However, the traditional artificial potential field method (APF) is easily disturbed by nearby obstacles when approaching the target point, resulting in the AGV being unable to accurately reach the target. When the environment near the target point becomes more complex, such as structures such as shelves and stacks, these situations require more sophisticated path planning strategies to ensure that the AGV can successfully complete the task.

[0011] (2) Local optimal solution trap. In unmanned warehouse environments, various obstacles are densely distributed. Traditional APF is prone to fall into the local optimal solution trap, which makes the AGV unable to find the global optimal solution path. The local optimal solution trap causes the AGV to frequently change direction during operation, reducing the overall operation time and efficiency of the system; (3) Computational efficiency and resource consumption. In unmanned warehouse environments, most operations and selection behaviors require the use of computing resources, which leads to high requirements for computer resources. The path planning of AGV, especially dynamic planning behavior, requires a large amount of computing resources. In actual application scenarios, the AGV path planning system needs to process multiple data and various static and dynamic scenarios at the same time. Efficient algorithms are crucial to improving the system's operating efficiency and performance.

[0012] In summary, the challenges faced by traditional artificial potential field methods in industrial application scenarios mainly include: target node accessibility challenges, local optimal solution traps, and computational efficiency and resource consumption. Whether these problems can be solved is particularly important for improving the efficiency and performance of AGV path planning systems. Summary of the invention

[0013] The present invention aims to solve the existing calculation problems and provides a storage AGV static path planning method based on an improved artificial potential field method.

[0014] In order to solve the path planning problem of AGV in unmanned warehousing based on the improved artificial potential field method, the technical solution adopted is as follows: a static path planning method for warehousing AGV based on the improved artificial potential field method,

[0015] The specific steps include:

[0016] Firstly, a grid map based on the grid method is established according to the actual storage environment to mark the specific location of static obstacles; secondly, the fitness value of the feasible area node is determined according to the improved repulsion function combined with the gravitational constant and the connectivity value; finally, reverse traversal is started from the target node to find the starting node (the feasible area node where the AGV is located), and the optimal node is stored. After the traversal is completed, the optimal path is generated.

[0017] Step 1: First, the unmanned warehouse environment is modeled according to the grid method, and the map size and the coordinates of static obstacles are set through field measurements. In order to improve the accuracy of modeling, all obstacles are divided according to their projected areas. Obstacle areas that are less than one grid are treated as full grids and marked as 1; passable areas are marked as 0;

[0018] Step 2: Before confirming the starting node and the target node, calculate the static repulsion of all feasible areas and save the static repulsion. The magnitude and direction of the repulsive force , which will greatly reduce the path planning calculation time after determining the target node and improve the path planning efficiency;

[0019] First, when there are multiple obstacles within the force range of the feasible area, determine whether the obstacles are adjacent. If they are adjacent, the multiple adjacent obstacles are regarded as an obstacle cluster, and the centroid of the merged obstacle cluster is used to calculate the direction of the repulsive force generated on the feasible area, and the magnitude of the repulsive force is calculated by the distance from the feasible area to the overall edge of the obstacle; when there are multiple obstacles within the force range of the feasible area, and the obstacles are not adjacent, the repulsive force of each obstacle on the feasible area is calculated separately; it should be noted that when there are multiple non-adjacent obstacles within the force range of the feasible area, the repulsive force is attenuated to ensure that the robot can pass through the complex environment; when there is a single obstacle within the force range of the feasible area (or there is no obstacle), the repulsive force of the single obstacle on the feasible area is calculated (or the repulsive force is directly set to 0 or the direction does not exist).

[0020] Improved static repulsion calculation formula:

[0021] ,

[0022] in, is the static repulsion, It is a feasible area The repulsive force vector under the action of static obstacles within the range, is the attenuation coefficient, and They are of Coordinates and coordinate, It is an obstacle For feasible area The magnitude of the repulsive force, It is an obstacle For feasible area The direction of repulsion, It is a feasible area The direction of the resultant repulsive force on The function converts point coordinates into polar coordinates for finding the angle of force. is the repulsive force constant, It is an obstacle and feasible region The straight-line distance between and The feasible area of Coordinates and coordinate, and The obstruction of material mind Coordinates and coordinate; Represents the feasible region The number of obstacles within the repulsion range.

[0023] Step 3: After confirming the starting node and the target node, calculate the gravitational force on all feasible areas. Direction of gravity Calculation of; Setting initial gravity is a fixed value greater than the repulsive force , the direction is from the starting node (AGV) to the target node. This constant value is set large enough to offset the repulsive force, but not too large to prevent the mobile target from being unable to control when approaching the obstacle danger zone.

[0024] The improved gravity calculation formula is as follows:

[0025] ,

[0026] The initial gravity Fixed value , is greater than the repulsive force, From the feasible region Point to target point Direction angle; and The target nodes of Coordinates and coordinate,

[0027] Each batch of connectivity value calculations will generate a route. The connectivity values ​​of the feasible area nodes in the feasible route exist, and the connectivity values ​​of the feasible area nodes in the infeasible route are all 0. By comparing the size of the connectivity value of the starting node, the probability of the high-quality path being selected during path planning is increased.

[0028] Step 4: First, use the BFS algorithm to reversely traverse and calculate the connectivity value of the feasible area between the target node and the starting node (AGV) , size, and the connectivity value of the feasible region nodes adjacent to the target node is initially defaulted to After the reverse traversal calculation, if the connectivity value of the feasible area node adjacent to the starting node (AGV) is obtained (this is considered as the end of a batch of connectivity value traversal calculation), then its connectivity value is assigned to the feasible area node adjacent to the target node (the starting point of this batch of reverse traversal), and the connectivity value of the feasible area in the batch of reverse traversal is retained. ; When the stop condition appears (the connectivity values ​​of the nodes in the adjacent areas of this batch are calculated, or the adjacent uncalculated areas are adjacent to obstacles, then the calculation of this batch is terminated), the feasible area nodes of this batch of reverse traversal are assigned a value of 0, indicating that the path cannot connect the starting node (AGV) and the target node. By adjusting the connectivity value, the probability of high-quality candidate nodes being selected when selecting the initial path node is increased.

[0029] If there is an obstacle between the feasible region node to be calculated and the target node, multiply the connectivity value of the previous node in the feasible region to be calculated by the coefficient , Set to 0.8 and assign it to the feasible region to be calculated; if the step length of the feasible region to be calculated from the target node exceeds 5, multiply the connectivity value of the node before the region to be calculated by the coefficient , Set it to 0.9 and assign it to the feasible area to be calculated; if the distance between the feasible area to be calculated and the target node is within the step size of 5 and there are no obstacles, then directly copy the connectivity value of the target node: 1; reflects the degree of connectivity between the node and the target node; greatly reduces the complexity of the calculation, making it easier to select path nodes with fewer inflection points and shorter distances, thereby improving the efficiency of the algorithm.

[0030] Secondly, how to determine whether there are obstacles between the feasible area to be calculated and the target node, use the Bresenham algorithm to draw a straight line from the area to be calculated to the target node; traverse all the grid cells along this line, and check whether these cells are obstacles (that is, whether they are marked as "1"); if all the grid cells on the straight line are "0", there is no line of sight obstruction. If one or more grid cells on the straight line are "1", it means that there is a line of sight obstruction.

[0031] The formula for calculating the connectivity value is as follows:

[0032] ,

[0033] in, is the connectivity value of the feasible region, is the connectivity value of the feasible region adjacent to the target node, which is set to 1; for forward The connectivity value of each node; is the obstacle attenuation coefficient, is the distance attenuation coefficient, is the obstacle between the area to be calculated and the target area quantity; It is expressed as the step distance between the target area and the area to be calculated.

[0034] Each batch of connectivity value calculations will generate a route. The connectivity values ​​of the feasible area nodes in the feasible route exist, and the connectivity values ​​of the feasible area nodes in the infeasible route are all 0. By comparing the size of the connectivity value of the starting node, the probability of high-quality paths being selected during path planning is increased, greatly improving the efficiency of path planning. This algorithm uses a reverse traversal method starting from the target node to calculate the connectivity value in order to ensure the target accessibility of the path. Therefore, in step 3, after a batch of reverse traversals is completed, the connectivity value of the adjacent feasible area of ​​the starting node (AGV) will be assigned to the target node. The purpose of doing this is to subdivide the connectivity values ​​of the nodes in the adjacent area of ​​the target node, increase the probability of high-quality paths being selected in the path planning stage, and greatly improve the optimization efficiency of the algorithm.

[0035] Step 5: Combine the gravity of each feasible area Connectivity value , and then with the static repulsion Calculate and obtain the resultant force size of the feasible area , combined with the gravitational direction of each feasible area The direction of the repulsive force Calculate reasonable direction , compare and select the best feasible area as the optimal path node. First, for each feasible area, combine the static obstacle force and the gravity after the connectivity coefficient correction to generate the final force size and direction of the feasible area; secondly, start with the target node and find the optimal path. When there are multiple feasible areas directly adjacent to the target node, select the feasible area with the largest force and push it into the stack; starting from the second area node pushed into the stack (the first coordinate in the stack is the target node), traverse its adjacent feasible area nodes, and select the node with the smallest angle difference between the force direction and the force direction of the top node of the stack (when there are multiple feasible area nodes with the same angle difference between the force direction and the top node of the stack, select the feasible area with the largest force and push it into the stack); and repeat the above operation until the starting node is pushed into the stack. Take out the nodes in the stack one by one, and use them as the best path.

[0036] The calculation formulas for the resultant force value and direction are as follows:

[0037] ,

[0038] in The feasible area The magnitude of the resultant force on ; The feasible area The direction of the combined force of attraction and repulsion; The feasible area The connectivity value of The feasible area The magnitude of the gravitational force; The feasible area The magnitude of the repulsive force; The feasible area The direction of the net gravitational force; The feasible area The direction of the resultant repulsive force.

[0039] During the operation of the AGV, if a dynamic obstacle appears, the trajectory of the dynamic obstacle is used to determine whether there is a path intersection. If a trajectory intersection occurs, its speed is identified and whether there is a collision risk is predicted, and then targeted obstacle avoidance strategies are selected, such as: deceleration, stopping, and route modification.

[0040] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, a static path planning method for a storage AGV based on an improved artificial potential field method is implemented.

[0041] A computer-readable storage medium stores computer instructions, which, when executed by a processor, implement a static path planning method for a storage AGV based on an improved artificial potential field method.

[0042] Compared with the prior art, the advantages of the present invention are as follows:

[0043] First, the present invention provides a solution that uses the feasible area as a force carrier in the standpoint. Before the path planning begins, the repulsion from the static obstacles in the feasible area is calculated and stored. After the target node is determined, the gravity and connectivity value are added to calculate the size and direction of the resultant force of each feasible area, and finally the feasible area with the best fitness is selected as the node of the optimal path. By pre-calculating the static repulsion of each feasible area, there is no need to repeatedly calculate the repulsion when the target is determined, which significantly reduces the amount of calculation, speeds up the path planning, optimizes the use of computing resources, and improves the overall response speed of the system; and using the feasible area as an alternative node for path planning can ensure the safety of the generated path, without worrying about collisions with static obstacles;

[0044] Second, the present invention provides a solution for processing multiple adjacent static obstacles within a range: when multiple adjacent obstacles appear, the adjacent obstacles are regarded as a whole obstacle, and the direction of the repulsive force brought by the obstacle to the feasible area is calculated by the centroid, and the magnitude of the repulsive force is calculated by the distance from the feasible area to the overall edge of the obstacle. This can not only simplify the calculation, but also more accurately simulate the impact of obstacles on path planning in the actual environment;

[0045] Third, the present invention provides a solution for dealing with multiple non-adjacent static obstacles within the range: when multiple non-adjacent obstacles appear, the effect of the static repulsion on the feasible area is calculated normally. To ensure that the AGV can pass through the complex environment smoothly, the superimposed static repulsion force is multiplied by the attenuation coefficient. ,generally , which can well deal with the unreachable target problem and "oscillation" behavior of the traditional artificial potential field method;

[0046] Fourth, the present invention provides a gravity function that is easy to converge quickly: the present invention presets gravity as a constant greater than the basic value of repulsion, and the direction is from the starting node (AGV) to the target node. The carriers of gravity and repulsion analyzed by the present invention are feasible areas, so there is no need to worry about the risk of collision caused by excessive gravity. In addition, the attenuation effect of the connectivity value on gravity will greatly reduce the size of gravity. A larger gravity value can speed up the convergence of the resultant force, so that the direction of the resultant force in each feasible area tends to the target node; by setting the gravity value as a constant, the amount of calculation can be greatly reduced, the speed of path planning can be accelerated, the utilization of computing resources can be optimized, and the overall response speed of the system can be improved;

[0047] Fifth, the present invention provides a solution for quickly finding the optimal path: a method that reflects the degree of connectivity between the feasible area and the target node.

[0048] According to the straight-line distance between the feasible area and the target node and whether there are obstacles, the BFS algorithm is used to quickly evaluate the connectivity of the grid. The larger the obstacles and distance, the lower the connectivity value. When a connectivity value calculation is completed or ended, the connectivity value of the feasible area node adjacent to the starting node (AGV) is assigned to the next feasible area node after the target node. The effect of this is to select the path node with a higher overall connectivity value and a larger resultant force as soon as possible when traversing the path in reverse from the target node. Therefore, according to the influence of the connectivity value of this feasible area on the force field force, the probability of the optimal path being selected in the final path planning is improved. If the connectivity value does not decay and the value is 1, the gravitational force is not changed when calculating the resultant force, and the repulsion is adjusted to 0, which means that the feasible area can quickly and directly reach the target node without considering the influence of the repulsion. And using the idea of ​​dynamic programming, the already calculated connectivity value nodes are stored to avoid repeated calculations. In this way, the unreasonable excessive repulsion can be adjusted, and the target node can be quickly reached when the optimal path exists. Through this simplification strategy, the calculation time can be significantly reduced, the algorithm efficiency can be improved, and the most important thing is to ensure the rationality and feasibility of path planning.

[0049] Sixth, the present invention proposes a method for generating a final path to deal with unreachable targets and local optimal solutions: after the calculation of the resultant force and direction of all feasible areas is completed, the path node is selected by comparing the magnitude of the resultant force and the angle difference with the direction of the resultant force of the target node, and the selected node is pushed into the stack until the starting node (the feasible area where the AGV is located) is pushed into the stack, which is considered to be the end of the path planning. By searching for the path through reverse traversal, it can be ensured that the final path must be a reachable route; by comparing the direction of the resultant force of the feasible area to be selected and the adjacent target node, it can be ensured that the generated path has the smallest change in direction, and the angle difference is used as the first selection criterion, that is, when the force field environment is complex and the target node is reachable, the influence of the force field is eliminated and the next path node is directly reached. To a certain extent, the deadlock situation of "oscillation" and unreachable targets when approaching the target node is avoided. At the same time, each iteration selects the feasible area node with a smaller direction difference as the path node, which improves the ability to find the global optimal solution. This method has significant advantages in path planning efficiency and calculates static repulsion in advance: by pre-calculating the static repulsion of all feasible areas before determining the target node, it reduces the computational burden during task runtime and improves the real-time performance of path planning.

[0050] The stack structure stores the path. After the combined force is calculated, the path nodes are pushed into the stack in reverse order. When the AGV executes the path, it only needs to take out the nodes from the stack in sequence to move them. This avoids redundant operations in real-time path calculation and greatly improves the operating efficiency of the AGV.

[0051] Simplify the calculation of connectivity value. By simplifying the connectivity value between adjacent nodes, accurate calculation is performed only on key nodes. For adjacent nodes that meet the conditions, the connectivity value is directly copied, which significantly reduces the amount of calculation and improves the calculation speed.

[0052] Secondly, the algorithm also has advantages in controlling complexity and optimizing the combined force calculation: it uses a combination of advance calculation and step-by-step screening to decompose complex real-time calculation tasks into multiple simple operation steps, thereby effectively reducing the overall calculation complexity of the system;

[0053] Direction consistency judgment determines the consistency of the path by comparing the direction angle difference between nodes, thereby realizing the pre-calculation of the path. This method not only ensures the continuity of the path, but also avoids the increase in complexity caused by frequent calculations.

[0054] Merging strategy for obstacle processing: Merging adjacent obstacles simplifies the computational difficulty in a multi-obstacle environment and reduces redundant computations.

[0055] Finally, this method also has broad application prospects, efficient path planning in unmanned warehousing environments: In unmanned warehousing environments, AGVs need to quickly and accurately find the optimal path to improve operating efficiency. Through a variety of optimization methods, AGVs can quickly plan paths in complex environments, which is particularly suitable for warehousing operations in high-density obstacle environments.

[0056] Intelligent robot navigation, this algorithm can effectively handle path planning problems in multi-obstacle scenarios and has a wide range of application scenarios, such as intelligent warehousing, logistics robots, autonomous driving and other fields. Especially in applications that require real-time decision-making and path adjustment, it shows good adaptability. Scalability and customization: Your algorithm has a clear structure, rigorous logic, good scalability, and can be customized and optimized according to actual application needs. This makes it not only suitable for the current warehousing environment, but also can be extended to path planning tasks in other complex environments; the path planning method based on the improved artificial potential field method optimizes path calculation, reduces algorithm complexity, and improves path selection efficiency. It shows broad application prospects in unmanned warehousing environments and intelligent robot navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a schematic diagram of a static path planning method for a storage AGV based on an improved artificial potential field method according to the present invention,

[0058] Figure 2 is a flow chart of a static path planning method for a storage AGV based on an improved artificial potential field method according to the present invention,

[0059] Figure 3is a diagram showing the effects of different conditions of static obstacle repulsion on the feasible area of ​​the present invention,

[0060] Figure 4 is the resultant repulsive force diagram of the static obstacles on the feasible area nodes in the simulation experiment of the present invention,

[0061] Figure 5 This is a comparison chart of the effects of the improved APF algorithm and the APF algorithm in path planning in the simulation experiment of the present invention.

[0062] Figure 6 It is a line chart comparing the time consumption of path planning between the improved APF algorithm of the present invention and the APF algorithm.

[0063] Figure 7 It is a line graph comparing the route lengths of the improved APF algorithm of the present invention and the APF algorithm in path planning. DETAILED DESCRIPTION

[0064] In order to make the purpose and technical solution of the present invention more clearly presented, the present invention is further described in conjunction with examples. The description herein is for explaining the present invention, but not for limiting the present invention.

[0065] The embodiment of the present invention is a static path planning for a storage AGV based on an improved artificial potential field method, and has broad application scenarios in the fields of AGV path planning and scheduling in unmanned warehousing scenarios.

[0066] Example: Global Path Planning of AGV in Unmanned Warehousing Environment In an unmanned warehousing environment, the storage environment is complex and the layout is changeable. There are usually multiple static obstacles (for example, shelves, stacks, fixed equipment, etc.). The AGV needs to plan the path from the starting point to the end point while avoiding collisions with obstacles and optimizing the driving path as much as possible.

[0067] (1) Warehouse environment coordinate system construction: The static obstacles in the warehouse environment are manually marked using the grid method. Each grid represents the occupancy status (idle or occupied) of a specific area. A two-dimensional warehouse coordinate system is constructed, and on this basis, the starting and end points of the AGV, as well as the initial direction, speed, and steering constraints are determined;

[0068] (2) Force calculation in the feasible area: The improved artificial potential field method is used to calculate the repulsive force generated by the static obstacles on the AGV, the attractive force generated by the target point, and the connectivity value of the feasible area to calculate the resultant force. If there is no obstacle on the straight path, the repulsive force is made to approach 0 through the attenuation and enhancement function of the connectivity value. If there is an obstacle, the attractive force is adjusted according to the connectivity value and the resultant force is calculated to ensure the accessibility and efficiency of the path planning;

[0069] (3) Path planning and AGV operation: Generate the optimal path and run it according to the size and direction of the combined force in the feasible area, monitor dynamic obstacles in real time, and select the corresponding strategy based on their speed and path intersection prediction.

[0070] In view of the problems existing in the prior art, the present invention provides a static path planning method for a storage AGV based on an improved artificial potential field method. The present invention is further described below in conjunction with the accompanying drawings:

[0071] like Figure 1 As shown, the path planning method based on the improved artificial potential field method provided by the embodiment of the present invention includes the following steps:

[0072] S101, based on the specific conditions of the unmanned storage environment, the environment is described using a grid method, including information such as the size of the environment, the location distribution of static obstacles, etc.;

[0073] S102, before determining the coordinates of the starting node and the target node, analyze and calculate the repulsion force from the static obstacles in all feasible areas. If there are multiple obstacles within the force range of the feasible area node, and these multiple obstacles are adjacent, then these multiple obstacles are regarded as one obstacle, and the direction of the repulsion force is calculated using the combined obstacle mass center, and the magnitude of the repulsion force is calculated by the distance from the obstacle edge; if there are multiple obstacles within the force range of the feasible area node, and these multiple obstacles are not adjacent, then the repulsion effect of each obstacle on the feasible area node is calculated separately; if there is a single obstacle within the force range of the feasible area node, the repulsion effect is calculated normally.

[0074] S103, after determining the coordinates of the starting node and the target node, calculate the direction of gravity according to the starting node and the target node, and take the gravity as a constant, which is greater than the repulsive force; calculate the connectivity value of each feasible area by means of breadth-first search diffusion, and use dynamic programming to record the connectivity value of each node during the diffusion process to avoid repeated calculations, and when the connectivity value of a node is significantly smaller than the connectivity value of nearby nodes, stop calculating its connectivity value in advance to reduce unnecessary calculations.

[0075] S104, combining the gravity, repulsion and connectivity value to calculate the magnitude and direction of the resultant force of the feasible area nodes, and reversely traversing the feasible area nodes from the target node, selecting the feasible area node with the smallest angle difference and the largest resultant force as the path node, until the starting node is pushed into the stack.

[0076] S105, popping the nodes in the stack, which is the final optimal path.

[0077] like Figure 2 As shown, the path planning method based on the improved artificial potential field method provided in the embodiment of the present invention specifically includes the following steps:

[0078] Step 1: Map modeling and static obstacle annotation. Use the grid method to scale the storage environment according to the actual storage situation and truly restore the position of static obstacles in the environment. Divide the area occupied by the obstacle into an integer grid, represented by 1; and represent the grid of the passable area in the warehouse with 0.

[0079] Step 2: First, identify the obstacles within the force range of the nodes in the feasible area, and calculate the repulsive force according to the continuous situation, especially considering the closely adjacent obstacles as a whole obstacle for calculation.

[0080] The improved repulsion function expression is as follows:

[0081] ,

[0082] in, is the static repulsion, It is a feasible area The repulsive force vector under the action of static obstacles within the range, is the attenuation coefficient, It is an obstacle For feasible area The magnitude of the repulsive force, It is an obstacle For feasible area The direction of repulsion, It is a feasible area The direction of the resultant repulsive force on The function converts point coordinates into polar coordinates for finding the angle of force. is the repulsive force constant, It is an obstacle and feasible region The straight-line distance between and The feasible area of Coordinates and coordinate, and The obstruction of material mind Coordinates and coordinate; Represents the feasible region The number of obstacles within the repulsive force range;

[0083] Step 3: After confirming the start and target nodes, calculate the gravity and connectivity values. First, set the initial gravity. is a fixed value greater than the repulsive force , the direction is from the starting node (AGV) to the target node. This constant value is set large enough to offset the repulsive force, but not too large to prevent the moving target from being unable to be controlled when approaching the obstacle danger zone. This gravity value can be adjusted later according to the specific situation through the connectivity value to make the gravity value of the better path node larger.

[0084] The improved gravity calculation formula is as follows:

[0085] ,

[0086] The initial gravity Fixed value , is greater than the repulsive force, From the feasible region Point to target point Direction angle; and The target nodes of Coordinates and coordinate,

[0087] Next, the connectivity value is calculated.

[0088] Step 1: Set the target node Put in queue In order to perform the diffusion calculation of breadth-first search (BFS). Set the connectivity value of the target node ,

[0089] Step 2: BFS algorithm and path priority calculation, from the queue Take a node from deal with.

[0090] Connectivity value (connectivity value indicates the ability of the node to reach the target node and adjusts the gravity) calculation: priority evaluation is performed directly in the connectivity value calculation process. Calculation node To the target node distance, and check the number of obstacles on the path. To the target node The initial score of the path is set to 20 points. With the target node The distance between the nodes (number of grids), 1 point is deducted for each grid; use Bresenham algorithm to check nodes With the target node Whether there are obstacles between the paths, 2 points will be deducted for each obstacle found; finally, the priority of the paths will be compared based on the total score after the deductions.

[0091] Priority embedding: The connectivity value of paths with shorter distances and fewer obstacles is increased (i.e., the competitiveness of being selected is improved); the connectivity value of paths with longer distances or more obstacles is reduced.

[0092] Connectivity value calculation formula:

[0093] If there are obstacles and the distance to the target node is short: ,

[0094] If there are obstacles and the distance to the target node is long: ,

[0095] If there are no obstacles and the distance to the target node is long: ,

[0096] If there are no obstacles and the distance to the target node is short: ,

[0097] in, is the connectivity value of the feasible region, is the connectivity value of the feasible region adjacent to the target node, which is set to 1; for forward The connectivity value of each node; is the obstacle attenuation coefficient, is the distance attenuation coefficient, is the obstacle between the area to be calculated and the target area quantity; It is expressed as the step distance between the target area and the area to be calculated.

[0098] Here, the priority judgment and connectivity value calculation are combined together. The priority of the path is evaluated while calculating the connectivity value to avoid repeated calculation.

[0099] Step 3: Diffusion connectivity value: Diffusion connectivity value The connectivity value Diffusion to the adjacent feasible area nodes; if an adjacent area node has been processed by a high priority path (i.e. the connectivity value has been determined), skip the node to avoid repeated calculations; if the adjacent node has not been processed, add it to the queue , waiting for further processing. When the connectivity value of the feasible region node adjacent to the starting node is obtained, it is assigned to the feasible region node adjacent to the target node (the first node of the batch of connectivity value calculations). The purpose of this is to subdivide the connectivity values ​​of the nodes adjacent to the target node, increase the probability of high-quality paths being selected in the path planning stage, and greatly improve the optimization efficiency of the algorithm.

[0100] The fourth step is dynamic update and storage: the nodes and their values ​​whose connectivity values ​​have been calculated are stored so that they can be directly referenced in subsequent calculations to avoid unnecessary repeated calculations.

[0101] Step 5, termination condition: When the connectivity values ​​of all nodes have been determined, or the set time limit has been reached, the connectivity value calculation algorithm can be terminated in advance; the final connectivity value result is output for path planning.

[0102] Step 4: Calculate the feasible area force.

[0103] First, combining the gravitational forces of each feasible region Connectivity value , and then with the static repulsion Calculate and obtain the resultant force size of the feasible area , combined with the gravitational direction of each feasible area The direction of the repulsive force Calculate reasonable direction , compare and select the best feasible area as the optimal path node. First, for each feasible area, the static obstacle force is combined with the gravity after the connectivity coefficient correction to generate the final force size and direction of the feasible area;

[0104] Secondly, starting from the target node, find the optimal path. When there are multiple feasible areas directly adjacent to the target node, select the feasible area coordinate with the largest combined force and push it into the stack; starting from the second coordinate node pushed into the stack (the first coordinate in the stack is the target node), traverse its adjacent feasible area nodes, and select the node with the smallest angle difference between the combined force direction and the combined force direction of the top node of the stack (when there are multiple feasible area nodes with the same angle difference between the combined force direction and the top node of the stack, select the feasible area coordinate with the largest combined force and push it into the stack); and repeat the above operation until the starting node is pushed into the stack. Take out the nodes in the stack one by one, and use them as the optimal path.

[0105] The calculation formulas for the resultant force value and direction are as follows:

[0106] ,

[0107] in The feasible area The magnitude of the resultant force on ; The feasible area The direction of the combined force of attraction and repulsion; The feasible area The connectivity value of The feasible area The magnitude of the gravitational force; The feasible area The magnitude of the repulsive force; The feasible area The direction of the net gravitational force; The feasible area The direction of the resultant repulsive force.

[0108] like Figure 3 As shown, it is an analysis diagram of the repulsive force generated by the static obstacles on the feasible area nodes of the present invention.

[0109] When there are multiple obstacles within the force range of the feasible area, determine whether the obstacles are adjacent. If they are adjacent, the multiple adjacent obstacles are regarded as an obstacle cluster, and the centroid of the merged obstacle cluster is used to calculate the direction of the repulsive force it generates on the feasible area, and the magnitude of the repulsive force is calculated by the distance from the feasible area to the overall edge of the obstacle. When there are multiple obstacles within the force range of the feasible area and the obstacles are not adjacent, the repulsive force of each obstacle on the feasible area is calculated separately. It should be noted that when there are multiple non-adjacent obstacles within the force range of the feasible area, the repulsive force is attenuated to ensure that the robot can pass through complex environments. When there is a single obstacle or no obstacle within the force range of the feasible area, the repulsive force of the single obstacle on the feasible area is calculated or the repulsive force is directly set to 0 and the direction does not exist.

[0110] like Figure 4 As shown in the figure, it is a diagram of the repulsive force generated by static obstacles on the nodes of the feasible area in the experimental code of the present invention. The passability of the feasible area: strong, medium and weak are respectively represented by red, yellow and blue colors; at the same time, the direction of the repulsive force on the feasible area is represented by an arrow. The longer the arrow is and the brighter the brightness is, the greater the repulsive force on the feasible area is.

[0111] like Figure 5 As shown in the figure, the improved APF algorithm and the APF algorithm are compared in path planning in the simulation experiment of the present invention. It can be seen from the experimental results that the method of the present invention is not only affected by the repulsive force when performing path planning, but also selects the feasible area with stronger repulsive force as the result of path planning according to the actual connectivity of the feasible area.

[0112] like Figure 6 As shown, it is a comparison of the time consumption of path planning between the improved APF algorithm of the present invention and the APF algorithm. It can be seen that the time efficiency of path planning of the improved APF algorithm is obviously better than the path planning efficiency of the APF algorithm before improvement.

[0113] like Figure 7As shown in the figure, the route length of path planning between the improved APF algorithm of the present invention and the APF algorithm is compared. It can be seen that the path planning route length of the improved APF algorithm of the present invention is not always better than that of the APF algorithm before improvement, but the advantage of the improved APF algorithm of the present invention is mainly reflected in the efficiency of the path planning task.

[0114] The optimization model and improved function provided by the present invention solve the problems existing in the traditional and improved artificial potential field methods.

[0115] (1) The challenge of target node reachability. In traditional path planning methods, especially when there are obstacles near the target node, the target node is often unreachable. The improved algorithm sets an initial gravity value greater than the repulsive force and corrects the gravity by combining the calculation of the connectivity value, so that the target node can be reached smoothly even in a complex obstacle environment. This method effectively solves the problem of target node unreachability in path planning and improves the reliability of path planning.

[0116] (2) Local optimal solution trap. Local optimal solution is a common problem in path planning, which often causes the path to fall into the local optimal solution and cannot be optimized further. To avoid this situation, the improved algorithm calculates the combined force of each feasible area, including gravity, repulsion and connectivity value, and gives priority to the path node with the largest combined force when selecting the path. Through this global optimization strategy, the algorithm can effectively bypass the local optimal solution and find the global optimal path;

[0117] (3) Computational efficiency and resource consumption. Most of the improved path planning methods often have too much computational effort when dealing with complex environments, resulting in serious time and resource consumption. The improved algorithm greatly reduces the amount of computation by using raster modeling and optimized repulsion calculation methods. At the same time, a dynamic update and storage mechanism for connectivity values ​​is introduced to avoid repeated calculations and improve computational efficiency. When diffusing connectivity values, processed nodes are skipped by priority judgment, further reducing resource consumption. Overall, the improved algorithm performs better in terms of computational efficiency and resource utilization, and is suitable for scenarios with high real-time requirements.

[0118] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A static path planning method for warehouse AGV based on improved artificial potential field method, characterized in that: Build a map based on the unmanned warehousing scenario and determine the path based on the repulsion of static obstacles and the gravity and connectivity values ​​of the target nodes, including the following steps: Step 1: Use the grid method to model the unmanned warehouse environment. Set the map size and static obstacle coordinates based on field measurements. The area occupied by static obstacles is marked as 1, and areas that are less than one grid are considered full. The grid of the passable area is marked as 0. Step 2: Before confirming the starting node and the target node, calculate the static repulsion of all feasible areas and save the static repulsion F i The magnitude and direction of the repulsive force θ Fi , Step 3: After confirming the starting node and the target node, the gravity G of all feasible areas i With the gravitational direction θ Gi Calculate the connectivity value C of the feasible area between the target node and the starting node AGV through the BFS algorithm. i The initial connectivity value of the target node is 1, indicating that the node has the highest degree of accessibility to the target node. Step 4: Combine the gravitational forces G of each feasible region i Connectivity value C with feasible region i , and then with the static repulsion F i Calculate the resultant force R in the feasible area i Magnitude and direction of the resultant force R θ , compare and select the best feasible area as the optimal path node, Step 5: During the operation of the AGV, if a dynamic obstacle appears, determine whether there is a path intersection based on the trajectory of the dynamic obstacle. If there is a trajectory intersection, identify its speed and predict whether there is a collision risk, and then select an obstacle avoidance strategy.

2. According to claim 1, a static path planning method for storage AGV based on improved artificial potential field method is characterized in that: in, Step 2 includes: 2-1. When there are multiple obstacles within the force range of the feasible area, determine whether the obstacles are adjacent. If so, treat the multiple adjacent obstacles as an obstacle cluster, use the merged obstacle mass centroid to calculate the direction of their repulsive force on the feasible area, and calculate the magnitude of the repulsive force by the distance from the feasible area to the edge of the obstacle cluster; 2-2. If they are not adjacent, the repulsive force of each obstacle on the feasible area is calculated separately; when there are multiple non-adjacent obstacles within the force range of the feasible area, the repulsive force is attenuated to ensure that the robot can pass through the complex environment; 2-3. When there is a single obstacle or no obstacle within the force range of the feasible area, the repulsive force of the single obstacle on the feasible area is calculated or the repulsive force is 0. Static repulsion calculation formula: Among them, F i is the static repulsion, ||F i || is the repulsive force vector of the feasible area i under the action of static obstacles within the range, α is the attenuation coefficient, F i，x and F i，y They are ||F i ||'s x- and y-coordinates, F ji is the repulsive force of obstacle j on feasible area i, θ ji is the direction of the repulsive force of obstacle j on feasible area i, θ Fi is the direction of the resultant repulsive force on the feasible region i. The atan2 function converts the point coordinates into polar coordinates for the angle of force. k is the repulsive force constant, d ji is the straight-line distance between obstacle j and feasible area i, pi x and pi y are the x-coordinate and y-coordinate of the feasible region i, respectively, pj x and pj y are the x- and y-coordinates of the obstacle mass center respectively; f i (j) represents the number of obstacles within the repulsive force range of feasible region i.

3. According to claim 2, a storage AGV static path planning method based on an improved artificial potential field method is characterized in that: in, Step 3 includes: 3-1. Set the initial gravity G o is a fixed value g greater than the repulsive force, and the direction is from the starting node AGv to the target node. The gravity calculation formula is: The initial gravitational force G o is a fixed value g, which is greater than the repulsive force, θ Gi is the direction angle from the feasible area i to the target point t; pt x and pt y are the x-coordinate and y-coordinate of the target node t, respectively. 3-2. If there is an obstacle between the feasible area to be calculated and the target node, multiply the connectivity value of the adjacent previous node by the obstacle attenuation coefficient γ1 and assign it to the area to be calculated, where γ1 is 0.8, and proceed to step 3-3; 3-3. If the step length of the feasible area to be calculated from the target node exceeds 5, the connectivity value of the adjacent previous node is multiplied by the distance attenuation coefficient γ2 and assigned to the node to be calculated. γ2 is 0.

9. 3-4. If the step distance between the feasible area i to be calculated and the target node t is within 5 and there are no obstacles, then the connectivity value of the target node is directly copied; when the calculation of a batch of connectivity values ​​is completed, that is, the adjacent nodes of the starting node are calculated, or the adjacent uncalculated areas are all obstacles, then the calculation of this batch is terminated, and the feasible area nodes of the reverse traversal of this batch are assigned a value of 0, indicating that the route cannot be reached, and the connectivity value of the feasible area adjacent to the target node of this batch, that is, the feasible area adjacent to the starting node, is compared with the connectivity value of the feasible area adjacent to the target node of the previous batch. If the connectivity value of this batch is larger than that of the previous batch, the connectivity values ​​of all feasible areas on the route are covered, otherwise it remains unchanged. Connectivity value calculation formula: Among them, C i is the connectivity value of the feasible area, C g is the connectivity value of the feasible region adjacent to the target node, set to 1; C g-i C g The connectivity value of the first i nodes; γ1 is the obstacle attenuation coefficient, γ2 is the distance attenuation coefficient, b it (j) is the number of obstacles j between the area to be calculated and the target area; d it It is expressed as the step distance between the target area and the area to be calculated.

4. According to claim 1, a static path planning method for storage AGV based on improved artificial potential field method is characterized in that: in, Step 4 includes: 4-1. For each feasible region i, the magnitude and direction of the final force are calculated by combining the initial force. The initial force is the resultant force of the repulsive force of the static obstacle and the gravitational constant. The final force is the connectivity value C added to the feasible region. i The magnitude and direction of the resultant force afterwards; 4-2. When there are multiple feasible region nodes directly pointing to the target node, select the node with the largest combined force and push it into the stack; starting from the first node pushed into the stack, traverse its adjacent feasible region nodes and select the feasible region with the combined force R i The largest candidate node, when R i When they are equal, compare the direction of the resultant force R θ , select the feasible area node with small direction difference, and repeat step 4-2 until the starting node is pushed into the stack, and then take out the nodes in the stack one by one and use them directly as the best path. The final resultant force calculation formula is: Where R i is the magnitude of the resultant force in the feasible area i; R θ is the direction of the combined force of attraction and repulsion on the feasible region i; C i is the connectivity value of the feasible region i; G i is the magnitude of the gravitational force on the feasible region i; F i is the magnitude of the repulsive force on the feasible region i; θ Gi is the direction of the resultant gravitational force on the feasible region i; θ Fi is the direction of the resultant repulsive force on feasible region i.

5. The static path planning method for storage AGV based on improved artificial potential field method according to claim 3 is characterized in that: in, Step 3-2 includes: 3-21, using the Bresenham algorithm to generate a straight line from the area to be calculated to the target node; 3-22. Traverse all the grid cells along this straight line and check whether these cells are obstacles. If so, mark them as "1"; 3-23. If all grid cells on the straight line are "0", there is no line of sight obstruction. If one or more grid cells on the straight line are "1", it means there is line of sight obstruction.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, a static path planning method for a warehouse AGV based on an improved artificial potential field method as described in any one of claims 1 to 5 is implemented.

7. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instruction is executed by the processor, a static path planning method for a warehouse AGV based on an improved artificial potential field method as described in any one of claims 1 to 5 is implemented.

Images