An automated guided vehicle (AGV) path replanning method, apparatus, equipment, and storage medium

Abstract

This invention discloses a method, apparatus, device, and storage medium for replanning the path of an automated guided vehicle (AGV). The method includes: identifying areas affected by obstructions based on coordinates, dimensions, and safety distances to obtain areas unaffected by obstructions; abstracting grid regions within the unaffected areas, traversing all feasible grid nodes within the grid regions, and calculating the actual cost and estimated cost of the current feasible grid node and the next feasible grid node; selecting the feasible grid node with the lowest cost based on the sum of the actual cost and the estimated cost to obtain a detour path; simplifying the detour path to key nodes and fitting it as a spline curve; creating at least one temporary node and at least one temporary segment based on the spline curve to form a dynamic temporary map; and performing path replanning for the AGV based on the dynamic temporary map. This invention solves the problem of automatically creating temporary paths when there is no passable path in the scheduling system, avoiding system-level deadlock.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, and in particular to an automated guided vehicle (AGV) path replanning method, apparatus, equipment, and storage medium. Background Technology

[0002] In modern warehouses, factories, and other material handling and workshop operations, the application of Automated Guided Vehicles (AMRs) is becoming increasingly widespread, significantly improving the level of automation and production efficiency. During AMR scheduling, point-segmented directed topology maps are currently the mainstream map usage method. This type of map can effectively improve space utilization while simplifying the complexity of AMR path planning, adapting to the operational needs of enclosed indoor environments.

[0003] However, point-to-point directed topology maps have inherent drawbacks. In actual operation, if an unknown static obstacle appears on the path, such as goods dropped during operation or temporarily piled materials, or if an AMR suddenly malfunctions during task execution, it will directly block traffic on that section of the path. This will prevent all subsequent AMRs traveling along that path from moving normally, leading to a large-scale deadlock in the entire scheduling system and causing a complete paralysis of operational efficiency.

[0004] To address the problem of traffic congestion along routes, three common solutions have emerged in existing technologies. The first is local replanning by the AMR (Autonomous Mobile Controller). After detecting an obstacle, the AMR relies on its sensors and computing power to search for alternative paths within a local area. However, in complex network maps, locally feasible paths are extremely rare, often making it difficult to find an effective solution. Even if a feasible path is found, the AMR cannot know the positions and task status of other vehicles, making it prone to conflicts with other AMRs in operation, leading to a deadlock. Furthermore, the decentralized nature of this method means that the planned path is not globally optimal, and there are also problems such as difficulty in sharing replanning data, leading to subsequent vehicles repeatedly planning routes and causing new congestion.

[0005] Secondly, the scheduling system simply bypasses obstacles. After receiving an obstacle report, the scheduling system attempts to plan a path around the obstacle area for the affected AMR. However, the map on which the system relies remains unchanged. When obstacles block critical nodes, it is often impossible to find a feasible alternative path, and the deadlock problem still cannot be solved.

[0006] Thirdly, the AMR transmits the route back to the scheduling system after detouring. Although this method solves the limitations of multi-vehicle shared replanning of routes, other defects such as local conflicts and non-globally optimal routes in AMR local replanning have not been improved. Summary of the Invention

[0007] The purpose of this invention is to provide an automated guided vehicle (AMR) path replanning method, apparatus, device, and storage medium, aiming to solve problems such as system deadlock and inefficiency caused by AMR path congestion.

[0008] In a first aspect, embodiments of the present invention provide an automated guided vehicle (AGV) route replanning method, comprising:

[0009] Receive obstruction information and the current mission route of the automated guided vehicle reported by the vehicle;

[0010] Obtain the existing map and determine whether to plan a detour route based on the existing map, the obstacle information, and the current mission route of the automated guided vehicle;

[0011] When planning a detour route, obtain the coordinates of the obstruction, the size of the obstruction, and the preset safe distance;

[0012] Based on the coordinates, dimensions, and safety distance, the area affected by the obstruction is identified on the existing map, and the area unaffected by the obstruction is obtained.

[0013] Abstract a grid region within the unobstructed area, use a path search algorithm to traverse all feasible grid nodes within the grid region, and calculate the pheromone concentration, turning angle weight value, and static map configuration of the current feasible grid node and the next feasible grid node to obtain the actual cost.

[0014] The estimated cost is obtained by calculating the cost from the next feasible grid node to the end point of the task route using a heuristic function.

[0015] Based on the sum of the actual cost and the estimated cost, the feasible grid nodes with the lowest cost are selected in sequence, and the selected feasible grid nodes are combined to obtain the detour path;

[0016] The detour path is simplified to key nodes and fitted as a spline curve;

[0017] Based on the spline curve, at least one temporary node and at least one temporary segment are created to form a dynamic temporary map;

[0018] Based on the dynamic temporary map, rerouting is performed for the blocked automated guided vehicles (AGVs) and all AGVs whose paths will pass through the obstruction.

[0019] In a second aspect, embodiments of the present invention provide an automated guided vehicle (AGV) route replanning device, comprising:

[0020] The receiving unit is used to receive obstruction information reported by the automated guided vehicle and the current mission route of the automated guided vehicle.

[0021] The judgment unit is used to acquire an existing map and determine whether to plan a detour route based on the existing map, the obstacle information, and the current task route of the automated guided vehicle.

[0022] The acquisition unit is used to acquire the coordinates of the obstruction, the size of the obstruction, and the preset safe distance when planning a detour route;

[0023] The marking unit is used to mark the area affected by the obstruction on the existing map according to the coordinates, the size and the safety distance, so as to obtain the area unaffected by the obstruction.

[0024] The traversal unit is used to abstract the grid area within the unobstructed area. It uses a path search algorithm to traverse all feasible grid nodes within the grid area and calculates the pheromone concentration, turning angle weight value and static map configuration of the current feasible grid node and the next feasible grid node to obtain the actual cost.

[0025] The computation unit is used to calculate the cost from the next feasible grid node to the end point of the task route using a heuristic function, and obtain the estimated cost.

[0026] The selection unit is used to select the feasible grid node with the lowest cost in sequence based on the sum of the actual cost and the estimated cost, and combine the selected feasible grid nodes to obtain the detour path;

[0027] A fitting unit is used to simplify the detour path to key nodes and fit it to a spline curve;

[0028] A creation unit is used to create at least one temporary node and at least one temporary segment based on the spline curve, forming a dynamic temporary map;

[0029] The replanning unit is used to replan the routes for blocked automated guided vehicles (AGVs) and all AGVs whose paths will pass through obstacles, based on the dynamic temporary map.

[0030] Thirdly, embodiments of the present invention provide a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the automated guided vehicle path replanning method described in the first aspect.

[0031] Fourthly, embodiments of the present invention also provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program that, when executed by a processor, implements the automated guided vehicle path replanning method described in the first aspect.

[0032] This invention discloses a method, apparatus, device, and storage medium for replanning the path of an automated guided vehicle (AGV). The method includes: receiving obstruction information reported by the AGV and the AGV's current mission route; acquiring an existing map and determining whether to plan a detour path based on the existing map, the obstruction information, and the AGV's current mission route; when planning a detour path, acquiring the coordinates, size, and preset safety distance of the obstruction; marking the area affected by the obstruction on the existing map based on the coordinates, size, and safety distance to obtain an area unaffected by the obstruction; abstracting a grid region within the unaffected area, and traversing all feasible grid nodes within the grid region using a path search algorithm. The invention calculates the pheromone concentration, turning angle weight, and static map configuration of the current and next feasible grid nodes to obtain the actual cost. It then uses a heuristic function to calculate the cost from the next feasible grid node to the end point of the task route, obtaining an estimated cost. Based on the sum of the actual and estimated costs, the feasible grid nodes with the lowest costs are selected sequentially, and the selected feasible grid nodes are combined to obtain a detour path. The detour path is simplified to key nodes and fitted as a spline curve. Based on the spline curve, at least one temporary node and at least one temporary segment are created to form a dynamic temporary map. Based on the dynamic temporary map, path replanning is performed for blocked Automated Guided Vehicles (AMRs) and all AMRs whose paths will pass through obstacles. This invention receives obstruction or fault information reported by AMRs, combines multi-dimensional data such as task and node pheromone concentration, plans the optimal detour path in areas unaffected by obstructions, and then simplifies key nodes and fits them as spline curves to create temporary nodes and temporary road segments to form a dynamic temporary map. This invention fundamentally avoids system deadlock, achieves globally optimal path planning, and possesses adaptive and self-healing capabilities. It requires no manual intervention, exhibits good compatibility with existing AMR systems, effectively reduces vehicle conflicts, and significantly improves the operational throughput and stability of the scheduling system. This invention also provides an automated guided vehicle (AMR) path replanning device, a computer-readable storage medium, and a computer device, all possessing the aforementioned beneficial effects, which will not be elaborated further here. Attached Figure Description

[0033] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 A flowchart illustrating the path replanning method for automated guided vehicles;

[0035] Figure 2This is a schematic diagram of an AMR encountering obstacles.

[0036] Figure 3 A schematic diagram illustrating the principle of the method for calculating the coordinates of the shortest path point;

[0037] Figure 4 A scene diagram generated for the detour route;

[0038] Figure 5 A simplified schematic diagram of the detour path nodes;

[0039] Figure 6 A schematic diagram of a scenario where key nodes are fitted with spline curves;

[0040] Figure 7 This is a schematic block diagram of an automated guided vehicle (AGV) path replanning device. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] It should be understood that, when used in this specification and the appended claims, the terms “comprising” and “including” indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more of its features, integrals, steps, operations, elements, components and / or collections thereof.

[0043] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0044] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the relevant listed items and all possible combinations, and includes such combinations.

[0045] Please see Figure 1 This embodiment provides an automated guided vehicle (AGV) route replanning method, including:

[0046] S101: Receive obstruction information and the current mission route of the automated guided vehicle reported by the automated guided vehicle;

[0047] For details, please refer to Figure 2 When an Automated Guided Vehicle (AVR) performing a material transfer task detects an obstruction ahead, it uses its onboard laser and vision sensors to detect the obstruction. The AVR then simultaneously reports the obstruction information and its current task route to the dispatch system. The obstruction information includes its specific coordinates, actual size, and detection timestamp. The task route includes the starting point, all nodes along the route, the destination, the estimated time to reach each node, and the task's priority level. Upon receiving this information, the dispatch system immediately retrieves the existing electronic map corresponding to the warehouse area and matches the reported obstruction information with the node distribution and road segment attributes on the map. Simultaneously, it combines this information with key nodes along the task route to provide comprehensive data support for subsequent decisions on whether to plan a detour and for accurately identifying areas affected by the obstruction.

[0048] S102: Obtain the existing map and determine whether to plan a detour route based on the existing map, the obstacle information, and the current mission route of the automated guided vehicle;

[0049] Specifically, obtaining the existing map and determining whether to plan a detour route based on the existing map, obstacle information, and the current mission route of the automated guided vehicle includes:

[0050] Determine whether the obstruction is an automated guided vehicle based on the obstruction information;

[0051] If the obstruction is an automated guided vehicle (AGV), obtain the fault code reported by the AGV and determine whether the fault repair time exceeds the predetermined time based on the fault code.

[0052] If the repair time for the fault exceeds the predetermined time, it is determined that the current automated guided vehicle cannot be restored in the short term;

[0053] If the repair time for the fault does not exceed the scheduled time, then wait for the automated guided vehicle to be repaired;

[0054] If the obstruction is not an automated guided vehicle, then the obstruction is determined to be a faulty object;

[0055] When the obstruction is a faulty object or the current automated guided vehicle cannot be restored in the short term, determine whether a new path has been found on the existing map;

[0056] If no new route is found on the existing map, a detour route will be planned;

[0057] If a new path is found on the existing map, the travel cost of the current mission route and the travel cost of the new path of the automated guided vehicle are calculated to obtain the original travel cost and the new travel cost, respectively.

[0058] Determine whether the difference between the new passage cost and the original passage cost exceeds a predetermined difference;

[0059] If the difference between the new toll cost and the original toll cost exceeds a predetermined difference, a detour route will be planned.

[0060] If the difference between the new passage cost and the original passage cost does not exceed a predetermined difference, the automated guided vehicle will be controlled to travel along the new path.

[0061] More specifically, the Automated Guided Vehicle (AMR1) is performing a material handling task from point E, through point Z, to point F along a preset task route. When the AMR1 reaches the vicinity of point E, its onboard sensors detect an obstruction in the area ahead at point Z. It immediately reports the obstruction information to the dispatch system, including its coordinates, size, and timestamp, and simultaneously transmits details of the current task route.

[0062] The scheduling system first acquires the stored existing map data and, combined with the received obstruction information and the AMR1 mission route, initiates a detour path planning and judgment process. Based on the obstruction's coordinates and features, the system determines whether the obstruction belongs to another automated guided vehicle (AGV). If identified as an AGV, it retrieves the fault code reported by that vehicle and assesses whether the fault repair time exceeds a predetermined time (e.g., 30 minutes) based on the code type. If the repair time exceeds the limit, the vehicle is deemed unable to be restored in the short term; if not, the system enters a waiting state to monitor the fault repair progress. For example, if it is a critical component failure, it is determined that recovery is not possible in the short term. If it is an operational fault, such as a temporary loss of SLAM navigation, it is determined that recovery is possible in the short term. A fault monitoring timer is activated to monitor the fault duration. If the fault duration exceeds a system-set threshold, it is also considered that recovery is not possible in the short term. If the fault disappears within this time, the fault monitoring timer is cleared.

[0063] If the obstruction is not an automated guided vehicle (AGV) (such as fallen cargo), it is directly identified as a static malfunction. When the obstruction is a malfunction or the malfunctioning vehicle cannot be restored in the short term, the system attempts to find a new path around the obstruction area on the existing map. If no feasible path is found, the system immediately triggers the detour path planning mechanism.

[0064] If a new path is found, the system calculates the travel cost of the original mission route and the travel cost of the new path for AMR1. The travel costs of the original mission route and the new path are calculated using the following formulas:

[0065] ;

[0066] Where C is the total cost of the path, the path consists of n nodes, and i is the node number. Let be the pheromone concentration of the i-th node; Let R be the distance between the (i-1)th node and the ith node, and let R be the turning coefficient. It is the turning angle from node i-1 to node i.

[0067] Compare the difference between the new cost and the original cost. If the difference exceeds a preset ratio (e.g., 2 times), it is determined that a detour is necessary and path planning is initiated. If the difference does not exceed the limit, AMR1 is controlled to continue traveling directly along the new path.

[0068] The entire decision-making process relies on the global data integration capabilities of the scheduling system to ensure that decisions are based on real-time environmental conditions and task parameters, avoiding local suboptimal solutions. By dynamically evaluating path feasibility and economy, the system minimizes interference with the overall workflow while ensuring traffic efficiency.

[0069] In some embodiments, determining whether an obstruction is an automated guided vehicle based on obstruction information includes:

[0070] The system receives spatial coordinate data and visual recognition feature vectors of obstacles reported by automated guided vehicles (AGVs). The visual recognition feature vectors contain information on the obstacle's outline, surface texture, and reflectivity distribution. It queries the real-time location database of all AGVs in the scheduling system to obtain a set of candidate vehicles with matching spatial coordinates. The candidate vehicle set includes all AGVs with position deviations less than a preset threshold. Standard physical feature parameters for each AGV are extracted from the candidate vehicle set. These standard physical feature parameters include vehicle size, shape factor, and surface material reflectivity. The visual recognition feature vector of the obstacle is compared with the standard physical feature parameters of each AGV in the candidate vehicle set using cosine similarity calculation to generate a feature matching score sequence. The feature matching score sequence is sorted, and the highest score is selected. If the highest score exceeds a preset similarity threshold and the status information of the corresponding AGV in the candidate vehicle set contains a motion anomaly flag, the obstacle is determined to be an AGV; otherwise, the obstacle is determined to be a static obstacle. The determination result is associated with and stored with the obstacle's spatial coordinate data to provide a basis for subsequent path planning.

[0071] S103: When planning a detour route, obtain the coordinates of the obstruction, the size of the obstruction, and the preset safe distance;

[0072] Specifically, the system extracts the precise coordinates of obstacles from the information reported by the AMR1. These coordinates are determined based on the warehouse's global coordinate system and include X and Y axis position information. Simultaneously, it acquires the obstacle's three-dimensional dimensions, including length, width, and height. This data is generated through the AMR1's multi-sensor fusion system. The system then calls preset safety distance parameters, which are dynamically set based on the AMR1's vehicle dimensions, motion characteristics, and operating environment to ensure safety redundancy during obstacle avoidance.

[0073] S104: Based on the coordinates, the dimensions, and the safety distance, mark the area affected by the obstruction on the existing map to obtain the area unaffected by the obstruction;

[0074] Specifically, AMRs are abstracted into circular models, and obstacles are abstracted into rectangular models. Based on coordinate data and size parameters, spatial occupancy models of obstacles are accurately constructed in electronic maps.

[0075] Next, the AABB-Circle Collision Detection algorithm is used to calculate spatial relationships. The algorithm first determines the coordinates of the point closest to the center of the AMR circle within the rectangular area of ​​the obstacle, calculated as the Euclidean distance between the nearest point and the center of the circle. If this distance is less than the sum of the AMR radius and the safety distance, the area is determined to be a collision zone.

[0076] Next, using the obstacle's coordinates as the center and combining size parameters, a basic influence range is expanded, and then a safety distance buffer zone is superimposed to form the final influence area boundary. In the electronic map, all path nodes and road segments intersecting with this boundary are marked as areas affected by the obstruction, and the system updates their status to impassable.

[0077] After identifying the affected areas, a safe zone unaffected by obstructions is calculated using map differential. This zone includes all nodes and road segments that do not intersect with the affected boundary and meet the physical space requirements for AMR passage. The safe zone is then gridded, with the grid size set according to the AMR's motion accuracy, providing a structured spatial basis for subsequent path searching.

[0078] The entire process is completed automatically through the geometric calculation module of the scheduling system, ensuring the accuracy and real-time nature of the affected area identification and providing reliable spatial constraints for obstacle avoidance path planning.

[0079] The detailed process of the AABB-circle collision detection algorithm is as follows:

[0080] Based on the center point and length of the AMR, construct a circle. Find the nearest point within the rectangle of the obstacle. If the distance from this nearest point to the center of the circle is less than the radius of the circle, then...

[0081] ;

[0082] This is considered a collision. The coordinates of the nearest point (x1, y1) are calculated based on the different positions of the circle around the rectangle, thus obtaining the coordinates of the shortest path point.

[0083] Specifically, such as Figure 3As shown, the input parameters are explicitly defined: circle is the coordinate of the center A of the circle (circle.x is the X coordinate of point A, circle.y is the Y coordinate of point A); rect is the coordinate of the top left corner B of the rectangle (rect.x is the X coordinate of point B, rect.y is the Y coordinate of point B), and rect implicitly contains the width rect.w (i.e., the horizontal length of the rectangle from left to right, corresponding to the horizontal distance from B to E in the figure) and the height rect.h (i.e., the vertical length of the rectangle from top to bottom, corresponding to the vertical distance from B to C in the figure).

[0084] Then, the X and Y coordinates of the nearest point (closestPoint) are calculated in two steps:

[0085] When calculating the X-coordinate, compare the X-coordinate of the center A (circle.x) with the left and right boundaries of the rectangle: If the X-coordinate of point A is less than the left boundary of the rectangle (rect.x, i.e., the X-coordinate of point B), then the X-coordinate of the nearest point is taken as the left boundary value (rect.x); if the X-coordinate of point A is greater than the right boundary of the rectangle (rect.x + rect.w, i.e., the X-coordinate of the lower right corner D or upper right corner E of the rectangle), then the X-coordinate of the nearest point is taken as the right boundary value; if the X-coordinate of point A is between the left and right boundaries of the rectangle, then the X-coordinate of the nearest point is the same as the X-coordinate of point A.

[0086] When calculating the Y-coordinate, compare the Y-coordinate of the center A (circle.y) with the upper and lower boundaries of the rectangle: if the Y-coordinate of point A is less than the upper boundary of the rectangle (rect.y, i.e., the Y-coordinate of point B), then the Y-coordinate of the nearest point is taken as the upper boundary value (rect.y); if the Y-coordinate of point A is greater than the lower boundary of the rectangle (rect.y + rect.h, i.e., the Y-coordinate of the lower left corner C or the lower right corner D of the rectangle), then the Y-coordinate of the nearest point is taken as the lower boundary value; if the Y-coordinate of point A is between the upper and lower boundaries of the rectangle, then the Y-coordinate of the nearest point is the same as the Y-coordinate of point A.

[0087] The final returned `closestPoint` is the point on the rectangle that is closest to the center A of the circle. Considering the position of the circle on the left side of the rectangle in the diagram, this closest point may fall on the left edge of the rectangle (such as near point H). The specific coordinates are determined by the X and Y coordinate calculation results mentioned above.

[0088] S105: Abstract the grid area within the unobstructed area, use the path search algorithm to traverse all feasible grid nodes within the grid area, and calculate the pheromone concentration, turning angle weight value and static map configuration of the current feasible grid node and the next feasible grid node to obtain the actual cost.

[0089] Specifically, starting from the current position E of AMR1 and ending at the target position F, the A* path search algorithm is initiated within the safe area. The algorithm meshes the environment to a precision of 1mm × 1mm and systematically traverses all feasible grid nodes by maintaining open and closed lists. For each currently feasible grid node, the algorithm explores its next adjacent feasible grid node and calculates the actual cost of moving from the current feasible grid node to the next feasible grid node.

[0090] The calculation of actual costs comprehensively considers three dynamic factors: pheromone concentration P(t), which is dynamically updated based on historical vehicle traffic frequency through a evaporation mechanism; steering angle weight Sr, which is calculated using the minimum steering angle formula combined with the current orientation of the AMR; and map static configuration Ss, which reads preset special road segment weights. Specifically, the actual costs are obtained by calculating the pheromone concentration, steering angle weight, and map static configuration of the current feasible grid node and the next feasible grid node, including:

[0091] The actual cost of the current feasible grid node and the next feasible grid node is calculated using the following formula:

[0092] G = Sr * P(t) * Ss * Distance;

[0093] Where G represents the actual cost; Sr represents the turning angle weight value; P(t) represents the pheromone concentration of the next feasible grid node; Ss represents the static map configuration of the next feasible grid node; and Distance represents the distance between the current feasible grid node and the next feasible grid node.

[0094] The calculated actual cost is updated to the path cost matrix in real time, providing a quantitative basis for global path decision-making and ensuring that AMR can efficiently and safely bypass obstacle areas.

[0095] In some embodiments, the AMR needs to rotate less during its orbital maneuver to achieve a smoother travel path. Therefore, the turning angle required for the AMR to reach the next point is calculated. Specifically, the method for calculating the weighted value of the turning angle includes:

[0096] Based on the existing map, obtain the coordinates of the next feasible grid node and the angle between the current feasible grid node and the positive x-axis;

[0097] Based on the coordinates of the next feasible grid node, calculate the angle between the next feasible grid node and the positive x-axis using the following formula;

[0098] ;

[0099] Where x and y represent the x-coordinate and y-coordinate of the next feasible grid node, respectively, and θ represents the angle between the next feasible grid node and the positive x-axis;

[0100] Based on the angle between the next feasible grid node and the positive x-axis and the current feasible grid node, the minimum turning angle from the current feasible grid node to the next feasible grid node is calculated using the following formula:

[0101] ;

[0102] in, This represents the minimum turning angle from the current feasible grid node to the next feasible grid node; This represents the angle between the currently feasible grid node and the positive x-axis direction;

[0103] The steering angle weight value is calculated based on the minimum steering angle using the following formula:

[0104] ;

[0105] Where k represents the steering weight coefficient; Sr represents the steering angle weight value.

[0106] This embodiment uses atan2(y, x) to solve for the angle between the next feasible grid node and the positive x-axis. Compared to the traditional tan(y / x) calculation method, it can cover the entire coordinate system quadrant and accurately locate the angular orientation of the node, laying a reliable foundation for subsequent calculations. Simultaneously, the minimum steering angle calculation closely follows actual steering patterns. It uses a dedicated formula to solve for the minimum steering angle from the current node to the next node, rather than simply calculating the angle difference. This accurately reflects the physical characteristic of the automated guided vehicle (AGV) prioritizing the selection of the minimum angle during actual steering, avoiding deviations caused by direct calculation of angle differences and realistically reflecting the actual cost of steering operations.

[0107] In some embodiments, during detours, it is necessary to avoid frequently passing nodes as much as possible. In the scheduling system, each time a vehicle passes a node, it increases the concentration of that node. Simultaneously, the concentration naturally dissipates over time. When calculating obstacle avoidance paths, if the obstacle avoidance point is near a certain node, the concentration of that node will affect the obstacle avoidance path. Therefore, the influence of pheromone concentration needs to be considered.

[0108] Furthermore, the formula for calculating pheromone concentration is as follows:

[0109] P(t)=P(t-1)*(1-ρ)+ΔP;

[0110] Where P(t) represents the pheromone value at the current time; P(t-1) represents the pheromone value at the previous time; ΔP represents the concentration value added in the current time window; and ρ represents the evaporation rate, a constant between 0 and 1.

[0111] In some embodiments, if the node is configured as a special road segment point in the existing map, it will also increase the cost of obstacle avoidance at that node. Therefore, the weight coefficient of the special road segment point (i.e., the static map configuration) is set to Ss.

[0112] S106: Calculate the estimated cost from the next feasible grid node to the end of the task route using a heuristic function;

[0113] Specifically, the coordinates of the next feasible grid node and the coordinates of the endpoint of the task route are obtained. The Euclidean distance between the two points is calculated as a basic estimate. The static configuration database of the existing map is queried to obtain the special weight coefficient Sg of the endpoint area. The special weight coefficient Sg is set with different values ​​depending on whether the endpoint area is a workstation, charging area, or high-traffic area. Then, the average pheromone concentration Pavg(t) of each node on the path from the next feasible grid node to the endpoint is obtained. The average pheromone concentration Pavg(t) is obtained by calculating the weighted average of the pheromone concentrations of all nodes on the path, with the weight being the reciprocal of the distance from each node to the endpoint. It is then checked whether the straight path from the next feasible grid node to the endpoint intersects with a special road segment. If... If intersections occur, the congestion prediction coefficient Tc of the intersecting special road segment is obtained. The congestion prediction coefficient Tc is calculated based on the special road segment creation time, the number of allocated AMRs, and the geometric length of the special road segment. Then, the base estimate, special weight coefficient Sg, average pheromone concentration Pavg(t), and congestion prediction coefficient Tc are substituted into the heuristic function H=Dbase×Sg×(1+α×(1-Pavg(t)))×(1+β×Tc) to calculate the estimated cost, where Dbase is the base estimate, and α and β are system-preset adjustment parameters. Subsequently, the calculated estimated cost is added to the actual cost G to obtain the total cost F, and the nodes in the open list are prioritized based on the total cost F.

[0114] Specifically, the average pheromone concentration Pavg(t) of each node on the path from the next feasible grid node to the destination is obtained. The average pheromone concentration Pavg(t) is obtained by calculating the weighted average of the pheromone concentrations of all nodes on the path, with the weights being the reciprocal of the distance from each node to the destination, including:

[0115] Extract real-time pheromone concentration data of all intermediate and terminal nodes along the path, and simultaneously obtain the straight-line distance or topological distance data from each node to the end point of the task route. Based on the distance data from each node to the end point of the task route, calculate the weight coefficient corresponding to each node, which is the reciprocal of the distance from the corresponding node to the end point. Multiply the real-time pheromone concentration data of each node with the corresponding weight coefficient, and sum all the product results to obtain a weighted sum. At the same time, calculate the sum of the weight coefficients of all nodes. By dividing the weighted sum by the sum of the weight coefficients, obtain the average pheromone concentration Pavg(t) of each node on the path from the next feasible grid node to the end point of the task route.

[0116] In some embodiments, the congestion prediction coefficient Tc is calculated based on the special road segment creation time, the number of allocated AMRs, and the geometric length of the special road segment, including:

[0117] Obtain the spatial coordinates of the next feasible grid node and the location coordinates of the task route endpoint. Construct a straight path equation connecting the two points. Traverse the set of all line segments marked as special road segments in the existing map. Calculate the geometric intersection of the straight path equation with each special road segment line segment. When an intersection exists and is located within the range of a special road segment line segment, it is considered an intersection. Obtain the creation timestamp ts, current timestamp current_time, number of allocated automated guided vehicles n, and geometric length L of the intersecting special road segment. Calculate the existence duration Δt = current_time - ts of the special road segment. Substitute the existence duration Δt, the number of allocated automated guided vehicles n, and the geometric length L into the following formula to calculate the congestion prediction coefficient:

[0118] ;

[0119] Wherein, α, β, and γ are preset system parameters, α controls the basic congestion level, β controls the time decay rate, and γ is a constant offset.

[0120] Taking an AMR traveling from the current node to the mission endpoint as an example, the system first obtains the spatial coordinates of the next feasible grid node, assuming it is point C (x1, y1), and the position coordinates of the mission endpoint, assuming it is point D (x2, y2). Based on these two coordinates, the system constructs a straight-line path equation connecting points C and D, for example, using the two-point equation form y - y1 = m(x - x1), where the slope m is calculated through the coordinate difference, m = (y2 - y1) / (x2 - x1).

[0121] The system then iterates through the set of all line segments marked as special road segments in the existing map. Special road segments may include narrow passages or high-traffic areas. For each special road segment, the geometric intersection point of the straight path equation and the line segment is calculated. The coordinates of the intersection point are obtained by solving the system of equations defined by the straight line equation and the endpoints of the line segment. When an intersection point exists and is within the parameterized range of the special road segment, that is, when the intersection point coordinates satisfy the boundary conditions of the start and end points of the line segment, the system determines that the two segments intersect.

[0122] For intersecting special road segments, the system obtains their creation timestamp ts, current timestamp current_time, number of assigned automated guided vehicles n, and geometric length L. The duration of the special road segment is calculated as Δt = current_time - ts. Then, Δt, n, and L are substituted into the congestion prediction coefficient formula to obtain the congestion prediction coefficient.

[0123] S107: Based on the sum of the actual cost and the estimated cost, select the feasible grid node with the lowest cost in sequence, and combine the selected feasible grid nodes to obtain the detour path;

[0124] For details, please refer to Figure 4 From the open list, select the feasible raster node with the smallest F-value as the current expansion node. For this node, check if it is the target node (i.e., the endpoint); if not, generate all neighboring nodes, calculate the actual cost and estimated cost of each neighboring node, and update its F-value. By comparing the F-values ​​of all candidate nodes, select the node with the smallest cost in sequence to add to the path sequence.

[0125] The selected nodes are combined into a node sequence based on the parent pointer backtracking relationship, forming an initial detour path. This path ensures that obstacle areas are avoided while minimizing the total travel cost. Through continuous iteration and cost comparison, the final output is an optimized detour path composed of nodes with the minimum cost, providing a feasible navigation solution for the automated guided vehicle.

[0126] The entire process demonstrates the efficiency of cost-driven path search, ensuring path quality and planning efficiency through dynamic evaluation and selection.

[0127] S108: Simplify the detour path to key nodes and fit it as a spline curve;

[0128] Specifically, the detour path is simplified to key nodes and fitted as a spline curve, including:

[0129] Iterate through each node in the detour path;

[0130] Based on all nodes in the detour path, calculate the slope and distance between the current node and the previous node;

[0131] When the absolute value of the slope exceeds the preset angle tolerance threshold or the distance exceeds the preset distance tolerance threshold, the current node is marked as a critical node.

[0132] Retain the key nodes and assign the current node's value to the previous node;

[0133] The key nodes that were retained were integrated into simplified polylines;

[0134] When determining whether the number of key nodes in a simplified polyline exceeds a preset node number threshold;

[0135] When the number of key nodes exceeds a preset threshold, the simplified polyline will be divided into multiple sub-paths.

[0136] Multi-segment paths are converted into multi-segment spline curves using natural spline curve equations.

[0137] Specifically, after obtaining the initial detour path, the path simplification and curve fitting operations begin. First, the algorithm traverses each node in the detour path, processing each path point sequentially. For each current node, the system calculates the slope change and Euclidean distance between it and the previous node. The slope is obtained from the difference between the coordinates of the two points, and the distance is calculated using standard geometric formulas.

[0138] When the calculated absolute value of the slope exceeds a preset angle tolerance threshold, or the node spacing exceeds a preset distance tolerance threshold, the system marks the current node as a critical node. After the critical node is retained, the current node is assigned to the previous node variable for subsequent comparisons. After traversal is complete, all critical nodes are integrated into a simplified polyline.

[0139] The system then determines whether the number of critical nodes in the simplified polyline exceeds a preset threshold. If it does, the simplified polyline is divided into multiple sub-path segments, each containing an appropriate number of critical nodes. Finally, a cubic natural spline curve equation is applied to each sub-path segment to generate a smooth spline curve path. This curve ensures that the automated guided vehicle can smoothly transition between critical points, improving path tracking performance.

[0140] The entire process involves using threshold control to filter key nodes, then using piecewise curve fitting to optimize the path, and finally outputting a smooth trajectory suitable for AMR motion.

[0141] In one specific embodiment, please refer to Figure 5 and Figure 6After the A* algorithm generates the original obstacle avoidance path E→G→H→I→J→K→L→M→N→O→P→Q→R→S→T→U→V→W→F, the path simplification process is initiated. Initial parameters are set as follows: angle tolerance threshold is 20 degrees, corresponding to a slope tolerance of 0.36; distance tolerance threshold is 3 meters; the previous node pointer `prevNode` is initialized to the starting point E; and the key node list `keyNodes` is initialized to empty.

[0142] Then, the traversal begins from the second node G:

[0143] Calculate the slope k1=0.25 between G and E, and the distance d1=3.2 meters. Since the distance exceeds the distance tolerance threshold, G is retained. Mark G as a key node, add it to keyNodes, and update prevNode to G.

[0144] Next, the slopes of H and G, k1 = 0.12 and the distance d1 = 1.8 meters, were calculated. Since neither exceeded the threshold, H was not retained.

[0145] Then, the slopes of I and H were calculated as k1=0.32 and the distance d1=1.8 meters. Since neither exceeded the threshold, H was not retained.

[0146] Upon reaching point J, the slope k3 of J and I is 0.45, which exceeds the slope threshold of 0.36. Therefore, J is marked as a key node, added to keyNodes, and prevNode is updated to J.

[0147] Following this logic, the key nodes that are ultimately retained are E, G, J, N, R, S, U, V, W, and F.

[0148] The simplified path forms a simplified polyline. Next, the number of key nodes is calculated to be 10, while the preset node number threshold is 9. Since 10 > 9, the path needs to be segmented. The intermediate key node R is selected as the split point, dividing the curve into two segments. Then, the polyline is transformed into a cubic spline curve using the cubic natural spline curve equation. The formula for calculating the cubic natural spline curve equation is:

[0149] ;

[0150] Where x(t) and y(t) are cubic polynomials in the x and y directions, respectively; , , , , , , and is the control coefficient of the curve, which is solved through boundary conditions and continuity constraints; t is the parameter variable (usually representing the curve progress or time).

[0151] Furthermore, the cubic spline curve is divided into two segments: ER is the first segment, with G, J, and N as control points; RF is the second segment, with S, U, V, and W as control points. At this point, a temporary loop allowing AMR to flow from ERF is created.

[0152] This simplification and fitting process significantly reduces path complexity while maintaining path smoothness and obstacle avoidance effectiveness, enabling AMRs to efficiently and safely bypass obstacle areas.

[0153] S109: Create at least one temporary node and at least one temporary segment based on the spline curve to form a dynamic temporary map;

[0154] Specifically, creating at least one temporary node and at least one temporary segment based on spline curves to form a dynamic temporary map includes:

[0155] Based on the curvature variation characteristics of spline curves, at least one temporary node is set at curve inflection points, curvature abrupt changes, and the positions where the original map paths connect at both ends;

[0156] Obtain the node in front of the obstacle on the mission route;

[0157] Starting from the node in front of the obstacle on the mission route and ending at the temporary node, generate at least one temporary segment;

[0158] Temporary nodes and segments are associated with the existing map topology to form a dynamic temporary map.

[0159] This embodiment sets temporary nodes based on the curvature characteristics of spline curves, allowing nodes to accurately match the actual shape of the path. Curve inflection points and abrupt curvature changes are critical locations for vehicle turning and path changes during driving. Setting nodes at these locations not only meets the actual traffic needs of the vehicle and avoids driving stutters caused by abrupt path connections, but also makes the node distribution of the temporary path more reasonable, avoiding increased computational burden due to redundant nodes. Secondly, temporary segments are generated starting from nodes in front of obstacles in the task route, enabling precise connection between the temporary path and the original task route. There is no need to replan the entire task's starting path, ensuring that the vehicle can naturally enter the detour path from the original route, reducing the computational load of path planning, and improving response efficiency. Finally, temporary nodes and temporary segments are topologically associated with the existing map, so that the temporary path does not exist independently of the original map, but forms a continuous passage link with the nodes and paths of the original map. This maintains the topological consistency of the map, avoids passage breaks, and allows the dynamic temporary map to adapt to the current obstacle scene while being compatible with the existing path resources of the original map, without the need to reconstruct the complete map, saving system computational and storage costs.

[0160] In one specific embodiment, based on the curvature change characteristics of the two spline curves, a temporary node T1 is set at J (curvature abrupt change point) of the first curve segment, and a temporary node T2 is set at R where the second curve connects to the original map path.

[0161] Next, retrieve the task route data and determine that the node in front of the obstacle (material accumulation in the workshop) in the route is G. With G as the starting point and temporary node T1 as the ending point, generate temporary segment G-T1; with T1 as the starting point and temporary node T2 as the ending point, generate temporary segment T1-T2.

[0162] Finally, temporary nodes T1 and T2, as well as temporary segments G-T1 and T1-T2, are topologically connected with the original nodes in the existing map (such as the predecessor node of G and the successor node of T2), so that the temporary segments form a passable node link with the existing map, forming a dynamic temporary map adapted to the current obstacle scenario.

[0163] In some embodiments, after creating at least one temporary node and at least one temporary segment based on spline curves to form a dynamic temporary map, the following steps are included:

[0164] Calculate the estimated time interval for each automated guided vehicle to pass through the temporary section;

[0165] Construct a spatiotemporal conflict detection matrix, and use the spatiotemporal conflict detection matrix to record the time interval overlap of all automated guided vehicles applying to pass through the same temporary segment;

[0166] When overlapping time intervals are detected, the passage priority coefficient of the automated guided vehicle is obtained and the time allocation is adjusted according to the passage priority coefficient, and a unique passage time window is assigned to each automated guided vehicle.

[0167] Bind the passage time window to the spatial topology data of the temporary segment to form a passage plan;

[0168] The automated guided vehicles are guided to start and stop according to the reserved time slots, following the traffic plan.

[0169] Specifically, after the scheduling system successfully creates a dynamic temporary map containing temporary nodes and temporary segments, the system immediately initiates the spatiotemporal resource allocation process. First, based on the current position, speed model, and temporary segment path length of each automated guided vehicle (AGV), the estimated time interval for each vehicle to pass through the temporary segment is calculated. This time interval includes two key timestamps: entry time and departure time.

[0170] Next, a spatiotemporal conflict detection matrix is ​​constructed. This matrix uses temporary segments as the row dimension and time as the column dimension, recording the overlap of time intervals for all automated guided vehicles (AGVs) requesting passage through the same temporary segment through matrix cells. When overlapping time intervals of multiple vehicles are detected, the passage priority coefficient of each vehicle is obtained. This coefficient is calculated based on attributes such as task urgency and cargo weight.

[0171] Based on the traffic priority coefficient, the time allocation strategy is dynamically adjusted to allocate earlier and more generous traffic time windows to high-priority vehicles, while shifting or compressing the time windows of low-priority vehicles to ensure that each automated guided vehicle (AGV) obtains a unique traffic time window. After the adjustment is completed, the system binds the traffic time windows with the spatial topology data of the temporary sections to form a complete spatiotemporal joint traffic scheme.

[0172] Finally, the dispatching system guides the automated guided vehicles (AGVs) to start and stop during the reserved time slots according to the traffic plan. As the vehicles approach the temporary section, they adjust their speed based on system instructions to ensure exclusive passage through the temporary section within the designated time window. This completely avoids vehicle conflicts in both time and space, achieving lock-free passage. The entire process, through precise time window management, maximizes the utilization efficiency of the temporary road sections while ensuring traffic safety.

[0173] In some embodiments, obstacle information in the operating environment of the automated guided vehicle is acquired, including the obstacle's location, size, estimated duration, and affected traffic area. Then, based on the estimated duration and affected traffic area, and combined with a preset three-level temporary topology division standard, the corresponding temporary topology level of the obstacle is determined. The three-level temporary topology includes an instantaneous layer, a short-term layer, and a long-term layer. The instantaneous layer corresponds to scenarios where the estimated duration of the obstacle is a first preset duration (e.g., within 5 minutes), the short-term layer corresponds to scenarios where the estimated duration of the obstacle is a second preset duration (e.g., 5-60 minutes), and the long-term layer corresponds to scenarios where the estimated duration of the obstacle is a third preset duration (e.g., 60 minutes). (The estimated duration of the obstacle corresponding to the instantaneous, short-term, and long-term layers can be determined according to the actual situation; this embodiment does not impose any limitations on this.) Then, a differentiated resource allocation strategy is executed according to the determined temporary topology level. If it is an instantaneous layer... The system creates necessary temporary nodes only at the inflection points of the spline curves along the detour path and at the junctions with the existing map. For short-term layers, temporary passing zones are added on top of the core temporary nodes. For long-term layers, temporary charging points and bidirectional passing zones are created in addition to all temporary nodes. Then, the system collects the actual obstacle clearance time, level determination results, and path operation data. This data is then input into a preset reinforcement learning model. The model is iteratively trained to dynamically adjust the level thresholds of the three-level temporary topology. Based on the adjusted level thresholds, the system optimizes the temporary topology layering strategy, associating the temporary nodes, auxiliary facilities, and temporary segments corresponding to each level with the existing map topology to form a dynamic temporary map. Finally, based on the dynamic temporary map, the system performs path replanning for the blocked Automated Guided Vehicles (AUVs) and their paths through the obstacle areas, clarifying the passage rules for each AUV in the temporary topology.

[0174] By establishing a three-tiered temporary topology hierarchy, the system can accurately match corresponding resource allocation schemes based on the estimated duration of obstacles. The transient layer adopts a minimal resource allocation strategy, creating only necessary nodes to avoid resource waste; the short-term layer adds passing zones to meet the basic passage requirements of medium-duration obstacles; and the long-term layer establishes a complete temporary infrastructure network. This differentiated strategy ensures a precise match between resource input and obstacle impact, significantly improving system resource utilization efficiency.

[0175] In some embodiments, the collected historical obstacle clearance times, level determination results, and path operation data are preprocessed to remove abnormal records caused by sensor malfunctions. The historical obstacle clearance times, level determination results, and path operation data are then converted into standardized numerical indicators to construct a uniformly formatted training dataset. Based on the preprocessed dataset, the data distribution of each category is analyzed. Synthetic minority class oversampling is used to augment the long-term and short-term layer samples, while the proportion of instantaneous layer samples is adjusted through random sampling to achieve a balanced data volume across all levels. Subsequently, during model training, changes in data distribution are continuously monitored. When obstacle clearance times at each level are detected... When the mean deviation exceeds a preset threshold, an automatic training dataset update mechanism is triggered to incorporate the latest scene data to maintain model adaptability. Subsequently, a proximal policy optimization algorithm is selected as the framework for the reinforcement learning model, constructing a lightweight network structure containing an input layer, two hidden layers, and an output layer. The parameter update magnitude is controlled by pruning the objective function. A random deactivation mechanism and L2 regularization constraint are introduced into the network hidden layer, and an early stopping training strategy based on the validation set accuracy is set to prevent model overfitting. Furthermore, domain knowledge constraints are applied to the output layer of the reinforcement learning model to limit the reasonable range of threshold values ​​for each level and maintain the logical order relationship between thresholds, ensuring that the model output meets the needs of practical applications.

[0176] In one specific embodiment, relevant data on 420 historical obstacle events in the park over the past 12 months are collected, including the actual clearing time of each obstacle, the initial determination of the temporary topology level (instantaneous layer, short-term layer, long-term layer), and the traffic efficiency, map update frequency, and number of path conflicts in the path operation data.

[0177] Next, an outlier detection algorithm was used to filter out 18 abnormal records. Among them, 12 were incorrect records of clearing time caused by sensor malfunction (e.g., 2 minutes were mistakenly recorded as 20 minutes), and 6 were data with an impact range that exceeded the reasonable range (e.g., an obstacle of 10 square meters was mistakenly reported as 100 square meters). These abnormal records were removed. Then, the unstructured data was transformed into standardized numerical indicators: the hierarchical judgment results were encoded as instantaneous layer 1, short-term layer 2, and long-term layer 3. The impact range was quantified according to the number of channels (1 channel is recorded as 1, 2 channels as 2). The data format and units were unified, so that all time data was in minutes and distance data was in meters. Finally, a training dataset with a unified format and standardized data was constructed.

[0178] Further analysis of the class distribution based on the preprocessed dataset revealed a significant data imbalance: 70% of the samples were in the transient layer, 20% in the short-term layer, and only 10% in the long-term layer. To address this, synthetic minority class oversampling was employed to augment the long-term and short-term samples. Algorithms generated virtual samples that matched the original data characteristics, adding 84 samples to the long-term layer and 42 to the short-term layer. Simultaneously, 168 core samples were randomly selected from the original 294 samples in the transient layer. After these adjustments, the data proportions across each layer became more balanced: 40% for the transient layer, 35% for the short-term layer, and 25% for the long-term layer.

[0179] Subsequently, the distribution changes of the dataset were monitored in real time, and the mean deviation threshold for obstacle clearing time at each level was set at 20%. After one month of operation, it was detected that the mean actual clearing time of instantaneous layer obstacles changed from 3.2 minutes to 4.1 minutes, with a mean deviation of 28.1%, exceeding the preset threshold. This automatically triggered the training dataset update mechanism, incorporating 35 scenario data points from the latest month (including instantaneous layer obstacle data caused by the increase in temporary loading and unloading obstacles during peak season). The updated dataset maintained the data balance at each level, ensuring the model's adaptability.

[0180] When constructing the reinforcement learning model, the proximal policy optimization algorithm was selected as the core framework of the reinforcement learning model, and a lightweight network structure was designed: the input layer receives the processed standardized numerical index, the hidden layer is set with 2 layers and each layer contains 64 neurons, and the output layer outputs the level threshold of the three-level temporary topology; by pruning the objective function, the parameter update amplitude is controlled within ±0.1, avoiding model oscillation caused by excessive update step size.

[0181] To prevent overfitting, a random deactivation mechanism is introduced in the hidden layers of the network, with a dropout rate of 0.2, to randomly deactivate some neurons to avoid the model becoming overly reliant on specific features. L2 regularization is used to constrain the model weight parameters, with the regularization coefficient set to 0.001 to prevent overfitting caused by excessively large parameters. At the same time, an early stopping training strategy is set, with the layer decision accuracy on the validation set as the monitoring metric. When this metric shows no improvement for 5 consecutive iterations, training is automatically stopped and the current optimal model parameters are retained.

[0182] In addition, domain knowledge constraints are imposed on the output layer of the reinforcement learning model: the threshold of the instantaneous layer is limited to no more than 10 minutes and the threshold of the long-term layer is limited to no less than 30 minutes. The logical order of the short-term layer threshold being greater than the instantaneous layer threshold and the long-term layer threshold being greater than the short-term layer threshold is clearly defined. These constraints are transformed into mathematical expressions and incorporated into the model loss function, forcing the model to output layer thresholds that meet the needs of actual applications, and ensuring the feasibility of the model results.

[0183] S110: Based on the dynamic temporary map, perform path replanning for the blocked automated guided vehicles and all automated guided vehicles whose paths will pass through the obstruction.

[0184] Specifically, based on this dynamic temporary map, the scheduling system immediately performs global route replanning for the blocked AMR1 and all other automated guided vehicles (AGVs) whose routes pass through the Z-point obstacle area. The system first retrieves the task lists of all affected vehicles, including AMR2 and AMR3, whose original routes all involve the Z-point area. For each vehicle, a new route is calculated using the improved A* algorithm based on the dynamic temporary map, incorporating cost factors such as pheromone concentration and steering weights into the algorithm.

[0185] After replanning, AMR1's path was updated from EZF to ERF, bypassing obstacles via a temporary node R. AMR2's original path was also adjusted to avoid passing through point Z, potentially introducing other temporary segments to maintain efficiency. Traffic control rules were generated simultaneously to allocate passage time windows for temporary segments and prevent multi-vehicle conflicts.

[0186] After the planned route was issued, AMR1 immediately turned from point E to temporary node R, and then headed towards point F. Other affected AMRs also followed the updated route. The dispatching system monitored the entire process to ensure smooth traffic flow. Through the replanning mechanism of the dynamic temporary map, deadlocks were effectively avoided, and overall throughput efficiency was improved.

[0187] Simultaneously, the dispatch system begins monitoring. Each subsequent AMR passing through node T1 will have its front-end sensors scan the original obstacle area. The dispatch system analyzes this transmitted data. Assuming that none of the 20 AMRs passing through within 10 consecutive dispatch cycles report an obstacle, the decision control module determines that the obstacle has disappeared and immediately instructs the map management module to delete temporary node T1 and its two associated temporary segments, restoring the existing map to its original state.

[0188] This embodiment breaks the deadlock problem by dynamically expanding the map topology, overcoming the deadlock caused by blocked roads. Simultaneously, centralized dynamic route planning by the scheduling system optimizes the paths of all AMRs from a global perspective, preventing AMRs from affecting communication with other AMRs when executing new paths, thus avoiding conflicts and suboptimal solutions caused by local decision-making. Furthermore, through a periodic monitoring mechanism, the system can automatically detect environmental changes and promptly restore the original state after obstacles disappear, achieving self-healing capabilities without manual intervention. Moreover, this method does not place high demands on the intelligence of the AMRs themselves; the main intelligence is concentrated in the scheduling system, facilitating upgrades and modifications to existing AMR systems.

[0189] Please see Figure 7 This embodiment provides an automated guided vehicle (AGV) route replanning device 200, comprising:

[0190] The receiving unit 201 is used to receive obstruction information reported by the automated guided vehicle and the current mission route of the automated guided vehicle;

[0191] Judgment unit 202 is used to obtain an existing map and determine whether to plan a detour route based on the existing map, the obstacle information and the current task route of the automated guided vehicle;

[0192] The acquisition unit 203 is used to acquire the coordinates of the obstruction, the size of the obstruction, and the preset safety distance when planning a detour path;

[0193] The marking unit 204 is used to mark the area affected by the obstruction on the existing map according to the coordinates, the size and the safety distance, so as to obtain the area unaffected by the obstruction;

[0194] Traversal unit 205 is used to abstract grid areas within areas unaffected by obstructions. It uses a path search algorithm to traverse all feasible grid nodes within the grid area and calculates the pheromone concentration, turning angle weight value, and static map configuration of the current feasible grid node and the next feasible grid node to obtain the actual cost.

[0195] The calculation unit 206 is used to calculate the cost from the next feasible grid node to the end point of the task route using a heuristic function, and obtain the estimated cost.

[0196] Selection unit 207 is used to select the feasible grid node with the lowest cost in sequence based on the sum of the actual cost and the estimated cost, and combine the selected feasible grid nodes to obtain the detour path;

[0197] Fitting unit 208 is used to simplify the detour path to key nodes and fit it to a spline curve;

[0198] Creation unit 209 is used to create at least one temporary node and at least one temporary segment based on the spline curve to form a dynamic temporary map;

[0199] The replanning unit 210 is used to replan the routes of blocked automated guided vehicles and all automated guided vehicles whose paths will pass through obstacles based on the dynamic temporary map.

[0200] Furthermore, the determination unit 202 includes:

[0201] An obstacle determination subunit is used to determine whether the obstacle is an automated guided vehicle based on the obstacle information.

[0202] The encoding acquisition subunit is used to acquire the fault code reported by the automated guided vehicle if the obstruction is an automated guided vehicle, and determine whether the fault repair time exceeds the predetermined time based on the fault code.

[0203] The short-term determination subunit is used to determine that the current automated guided vehicle cannot be restored in the short term if the repair time of the fault exceeds a predetermined time.

[0204] A waiting subunit is used to wait for the automated guided vehicle to be repaired if the repair time of the fault does not exceed a predetermined time.

[0205] The fault determination subunit is used to determine that the obstruction is a faulty object if the obstruction is not an automated guided vehicle.

[0206] The path determination subunit is used to determine whether a new path has been found on the existing map when the obstruction is a faulty object or the current automated guided vehicle cannot be restored in the short term.

[0207] The first planning subunit is used to plan a detour route if no new route is found on the existing map.

[0208] The cost calculation subunit is used to calculate the travel cost of the current task route and the travel cost of the new route of the automated guided vehicle if a new path is found in the existing map, and obtain the original travel cost and the new travel cost respectively.

[0209] The difference judgment subunit is used to determine whether the difference between the new passage cost and the original passage cost exceeds a predetermined difference.

[0210] The second planning subunit is used to plan a detour route if the difference between the new passage cost and the original passage cost exceeds a predetermined difference.

[0211] The control subunit is configured to control the automated guided vehicle to travel along the new route if the difference between the new passage cost and the original passage cost does not exceed a predetermined difference.

[0212] Furthermore, the method for calculating the steering angle weight value includes:

[0213] Based on the existing map, obtain the coordinates of the next feasible grid node and the angle between the current feasible grid node and the positive x-axis;

[0214] Based on the coordinates of the next feasible grid node, the angle between the next feasible grid node and the positive x-axis is calculated according to the following formula;

[0215] ;

[0216] Where x and y represent the x-coordinate and y-coordinate of the next feasible grid node, respectively, and θ represents the angle between the next feasible grid node and the positive x-axis;

[0217] Based on the angle between the next feasible grid node and the positive x-axis and the current feasible grid node and the positive x-axis, the minimum turning angle from the current feasible grid node to the next feasible grid node is calculated according to the following formula:

[0218] ;

[0219] in, This represents the minimum turning angle from the current feasible grid node to the next feasible grid node; This represents the angle between the currently feasible grid node and the positive x-axis direction;

[0220] The steering angle weight value is calculated based on the minimum steering angle using the following formula:

[0221] ;

[0222] Where k represents the steering weight coefficient; Sr represents the steering angle weight value.

[0223] Furthermore, the traversal unit 205 includes:

[0224] The actual cost calculation sub-unit calculates the actual cost of the current feasible grid node and the next feasible grid node according to the following formula:

[0225] G = Sr * P(t) * Ss * Distance;

[0226] Where G represents the actual cost; Sr represents the turning angle weight value; P(t) represents the pheromone concentration of the next feasible grid node; Ss represents the static map configuration of the next feasible grid node; and Distance represents the distance between the current feasible grid node and the next feasible grid node.

[0227] Furthermore, the fitting unit 208 includes:

[0228] The node traversal sub-unit is used to traverse each node in the bypass path;

[0229] The slope calculation subunit is used to calculate the slope and distance between the current node and the previous node based on all nodes in the detour path;

[0230] The node marking subunit is used to mark the current node as a critical node when the absolute value of the slope exceeds a preset angle tolerance threshold or the distance exceeds a preset distance tolerance threshold.

[0231] The assignment subunit is used to retain the key node and assign the current node to the previous node;

[0232] The node integration subunit is used to integrate the retained key nodes into a simplified polyline;

[0233] The node count determination subunit is used to determine whether the number of key nodes in the simplified polyline exceeds a preset node count threshold.

[0234] The polyline segmentation subunit is used to divide the simplified polyline into multiple sub-paths when the number of key nodes exceeds a preset threshold.

[0235] The path transformation sub-unit is used to convert multi-segment sub-paths into multi-segment spline curves using natural spline curve equations.

[0236] Furthermore, the creation unit 209 includes:

[0237] The node setting subunit is used to set at least one temporary node at curve inflection points, curvature abrupt changes, and positions where the two ends connect to the original map path, based on the curvature change characteristics of the spline curve.

[0238] The node acquisition subunit is used to acquire nodes located in front of obstacles along the task route;

[0239] A temporary segment generation subunit is used to generate at least one temporary segment, starting from a node in front of an obstacle on the task route and ending at the temporary node.

[0240] The topology association subunit is used to associate the temporary node, the temporary segment, and the existing map topology to form a dynamic temporary map.

[0241] Furthermore, the creation unit 209 also includes:

[0242] The time calculation subunit is used to calculate the estimated time interval for each automated guided vehicle to pass through the temporary section;

[0243] The matrix construction subunit is used to construct the spatiotemporal conflict detection matrix and record the time interval overlap of all automated guided vehicles applying to pass through the same temporary segment through the spatiotemporal conflict detection matrix;

[0244] The time window allocation subunit is used to obtain the passage priority coefficient of the automated guided vehicle when overlapping time intervals are detected, and adjust the time allocation according to the passage priority coefficient to allocate a unique passage time window to each automated guided vehicle.

[0245] The data binding subunit is used to bind the passage time window to the spatial topology data of the temporary segment to form a passage scheme.

[0246] The passage guidance subunit is used to guide the automated guided vehicle to start and stop during the reserved time period according to the passage plan.

[0247] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0248] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, can implement the methods provided in the above embodiments. The storage medium may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0249] The present invention also provides a computer device, which may include a memory and a processor. The memory stores a computer program, and when the processor calls the computer program in the memory, it can implement the methods provided in the above embodiments. Of course, the computer device may also include various network interfaces, power supplies, and other components.

[0250] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make various improvements and modifications to this invention without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this invention.

[0251] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusivity.

[0252] The term "comprises" implies that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. An automated guided vehicle path replanning method, characterized by, The method comprises the following steps: receiving the blocking object information reported by the automatic guided vehicle and the current task route of the automatic guided vehicle; acquiring an existing map and determining whether to plan a detour path according to the existing map, the blocking object information and the current task route of the automatic guided vehicle; when planning the detour path, acquiring the coordinates of the blocking object, the size of the blocking object and a preset safety distance; identifying the area affected by the blocking object on the existing map according to the coordinates, the size and the safety distance, and obtaining the area not affected by the blocking object; abstracting a grid area in the area not affected by the blocking object, traversing all feasible grid nodes in the grid area by using a path search algorithm, and calculating the pheromone concentration, turning angle weight value and map static configuration of the current feasible grid node and the next feasible grid node to obtain an actual cost; calculating the cost of the next feasible grid node to the end point of the task route by using a heuristic function to obtain an estimated cost; selecting the feasible grid node with the minimum cost in turn based on the sum of the actual cost and the estimated cost, and combining the selected feasible grid nodes to obtain the detour path; simplifying the detour path into key nodes and fitting the key nodes into a spline curve; creating at least one temporary node and at least one temporary segment based on the spline curve to form a dynamic temporary map; replanning the path for the blocked automatic guided vehicle and all automatic guided vehicles that will pass through the blocking object based on the dynamic temporary map.

2. The method of claim 1, wherein, The step of acquiring the existing map and determining whether to plan the detour path according to the existing map, the blocking object information and the current task route of the automatic guided vehicle comprises the following steps: determining whether the blocking object is an automatic guided vehicle according to the blocking object information; if the blocking object is an automatic guided vehicle, acquiring the fault code reported by the automatic guided vehicle and determining whether the repair time of the fault exceeds a predetermined time according to the fault code; if the repair time of the fault exceeds the predetermined time, determining that the current automatic guided vehicle cannot be restored in a short period of time; if the repair time of the fault does not exceed the predetermined time, waiting for the automatic guided vehicle to be repaired; if the blocking object is not an automatic guided vehicle, determining that the blocking object is a fault object; when the blocking object is a fault object or the current automatic guided vehicle cannot be restored in a short period of time, determining whether a new path is found on the existing map; if a new path is not found on the existing map, planning the detour path; if a new path is found on the existing map, calculating the passing cost of the current task route of the automatic guided vehicle and the passing cost of the new path to obtain an original passing cost and a new passing cost, respectively; determining whether the difference between the new passing cost and the original passing cost exceeds a predetermined difference value; if the difference between the new passing cost and the original passing cost exceeds the predetermined difference value, planning the detour path; if the difference between the new passing cost and the original passing cost does not exceed the predetermined difference value, controlling the automatic guided vehicle to pass according to the new path.

3. The method of claim 1, wherein, The calculation method of the turning angle weight value comprises the following steps: acquiring the coordinates of the next feasible grid node and the angle between the current feasible grid node and the positive direction of the x-axis based on the existing map; Based on the coordinates of the next feasible grid node, an angle of the next feasible grid node and a positive direction of an x-axis is calculated according to the following formula: ； Wherein, x and y respectively represent an x-coordinate and a y-coordinate of the next feasible grid node, and θ represents the angle of the next feasible grid node and the positive direction of the x-axis. Based on the angle of the next feasible grid node and the positive direction of the x-axis and an angle of the current feasible grid node and the positive direction of the x-axis, a minimum steering angle from the current feasible grid node to the next feasible grid node is calculated according to the following formula: ； wherein, represents the minimum steering angle from the current feasible grid node to the next feasible grid node; represents the angle of the current feasible grid node with the positive direction of the x-axis; According to the minimum steering angle, a steering angle weight value is calculated according to the following formula: ； Wherein, k represents a weight coefficient of steering, and Sr represents the steering angle weight value.

4. The method of claim 1, wherein, The calculation of the pheromone concentration, the steering angle weight value and the map static configuration of the current feasible grid node and the next feasible grid node to obtain the actual cost includes: The actual cost of the current feasible grid node and the next feasible grid node is calculated according to the following formula: G = Sr*P(t)*Ss*Distance Wherein, G represents the actual cost, Sr represents the steering angle weight value, P(t) represents the pheromone concentration of the next feasible grid node, Ss represents the map static configuration of the next feasible grid node, and Distance represents the distance between the current feasible grid node and the next feasible grid node.

5. The method of claim 1, wherein, The simplification of the detour path into key nodes and the fitting into a spline curve includes: Traversing each node in the detour path; Based on all the nodes in the detour path, the slope and the distance between the current node and the previous node are calculated; When the absolute value of the slope exceeds a preset angle tolerance threshold or the distance exceeds a preset distance tolerance threshold, the current node is marked as a key node; The key nodes are retained and the current node is assigned to the previous node; The retained key nodes are integrated into a simplified polyline; It is judged whether the number of key nodes in the simplified polyline exceeds a preset node number threshold; When the number of key nodes exceeds the preset threshold, the simplified polyline is divided into multiple sub-paths; The multiple sub-paths are converted into multiple spline curves through a natural spline curve equation.

6. The method of claim 1, wherein, The creation of at least one temporary node and at least one temporary segment based on the spline curve to form a dynamic temporary map includes: Based on the curvature variation characteristics of the spline curve, at least one temporary node is set at the curve inflection point, the curvature mutation and the position of the original map path at both ends; A node in front of an obstacle where the task route is located is obtained; At least one temporary segment is generated with the node in front of the obstacle where the task route is located as the starting point and the temporary node as the ending point; The temporary node, the temporary segment and the existing map topology are associated to form a dynamic temporary map.

7. The method of claim 1, wherein, After the creation of at least one temporary node and at least one temporary segment based on the spline curve to form a dynamic temporary map, the following includes: The expected time interval of each automated guided vehicle passing through the temporary segment is calculated; A space-time conflict detection matrix is constructed, and the time interval overlap of all automated guided vehicles applying to pass through the same temporary segment is recorded through the space-time conflict detection matrix; When detecting the time interval overlap, a passing priority coefficient of the automated guided vehicle is acquired and the time distribution is adjusted according to the passing priority coefficient, and a unique passing time window is distributed to each automated guided vehicle; The passing time window is bound to the spatial topology data of the temporary period to form a passing scheme; The automated guided vehicle is guided to pass according to the passing scheme in the appointment period.

8. An automated guided vehicle path replanning device, characterized by, Comprise: A receiving unit configured to receive the blocking information reported by the automated guided vehicle and the current task route of the automated guided vehicle; A judging unit configured to acquire an existing map and determine whether to plan a detour path according to the existing map, the blocking information and the current task route of the automated guided vehicle; An acquiring unit configured to acquire the coordinates of the blocking, the size of the blocking and a preset safety distance when planning the detour path; An identifying unit configured to identify the area affected by the blocking on the existing map according to the coordinates, the size and the safety distance to obtain an area not affected by the blocking; A traversing unit configured to abstract a grid area in the area not affected by the blocking, traverse all feasible grid nodes in the grid area by using a path search algorithm, and calculate the pheromone concentration, turning angle weight value and map static configuration of the current feasible grid node and the next feasible grid node to obtain an actual cost; A calculating unit configured to calculate the cost of the next feasible grid node to the end point of the task route by using a heuristic function to obtain an estimated cost; A selecting unit configured to select the feasible grid node with the minimum cost in sequence based on the sum of the actual cost and the estimated cost, and combine the selected feasible grid nodes to obtain a detour path; A fitting unit configured to simplify the detour path into key nodes and fit the key nodes into a spline curve; A creating unit configured to create at least one temporary node and at least one temporary period based on the spline curve to form a dynamic temporary map; A re-planning unit configured to re-plan the path of the blocked automated guided vehicle and the automated guided vehicles that will pass through the blocking based on the dynamic temporary map.

9. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the automated guided vehicle path re-planning method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program causes the processor to execute the automated guided vehicle path re-planning method according to any one of claims 1 to 7 when the processor executes the computer program.

Images