Multi-AGV path planning method and related device

Abstract

The embodiment of the invention provides a multi-AGV path planning method and a related device. The method comprises the steps that in the single AGV path searching stage, a dynamic weight coefficient, a steering punishment mechanism and a Bezier curve smoothing strategy are introduced on the basis of a traditional A * algorithm, the operation efficiency of the algorithm can be improved, the unnecessary turning frequency in a path can be reduced, and the smoothness of the path can be improved. In a multi-AGV path conflict coordination stage, most traditional avoidance strategies use a fixed priority strategy, and the relative importance of tasks among the AGVs is not considered, however, according to the method, the avoidance among the AGVs can be implemented according to the relative importance of the tasks by setting a dynamic priority strategy. In the multi-AGV online navigation stage, a traditional DWA algorithm is prone to falling into local optimum, and the situation can be avoided by adopting the improved DWA algorithm, so that local obstacle avoidance can be achieved, and falling into local optimum can also be avoided.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of path planning technology, and in particular to a multi-AGV path planning method and related devices. Background Art

[0002] Multi-AGV (Automated Guided Vehicle) path planning algorithms are crucial in modern logistics, manufacturing, and warehousing systems. Key challenges include ensuring efficient and safe movement of multiple AGVs within a shared environment, avoiding collisions, coordinating tasks, and adapting to dynamic environmental changes. For example, in automated warehouses and intelligent manufacturing systems, the introduction of multi-AGV systems has significantly improved automation and operational efficiency.

[0003] Currently, there are two main approaches to implementing multi-AGV path planning algorithms: Conflict-Based-Search (CBS)-based multi-AGV path planning and reinforcement learning (RL)-based multi-AGV path planning. However, both the CBS-based multi-AGV path planning and reinforcement learning-based multi-AGV path planning algorithms have the following drawbacks:

[0004] 1. Disadvantages of multi-AGV path planning algorithm based on conflict search

[0005] In the single AGV path planning layer of the algorithm, the A* algorithm is mainly used. Its disadvantages are: (1) there are too many meaningless turns in the path generated by the traditional algorithm and the path smoothness is low; (2) the A* algorithm is a static search algorithm. This algorithm is not adaptable to dynamic unknown environments and has limited real-time obstacle avoidance capabilities. Later, it was proposed to integrate the traditional DWA algorithm, but it is easy to fall into the local optimal situation.

[0006] During the multi-AGV path conflict coordination phase, traditional algorithms use a fixed priority strategy. This strategy specifies the priority of each AGV before the algorithm begins. However, this can lead to low multi-AGV system efficiency and AGV deadlock during avoidance. Later, researchers proposed using random priorities. While this improved the deadlock problem, it still leaves the multi-AGV system inefficient due to a failure to balance the timeliness of each AGV's tasks.

[0007] 2. Disadvantages of the multi-AGV path planning algorithm based on reinforcement learning

[0008] Multi-AGV path planning methods based on reinforcement learning are flexible and adaptable in dynamic environments, but they also face some challenges. First, the training process is time-consuming and sample-inefficient, especially in high-dimensional state spaces. In addition, reinforcement learning is difficult to guarantee a global optimal solution and suffers from convergence problems, especially when multiple AGVs interact in complex ways. Although it can cope with dynamic obstacles, the model's generalization ability is weak and may not perform well in untrained environments. At the same time, deep reinforcement learning has high requirements for computing resources, and its strategies are often difficult to explain, resulting in limitations in safety and real-time performance. These problems mean that although reinforcement learning has potential, it still has many shortcomings in actual multi-AGV path planning. Summary of the Invention

[0009] The embodiments of the present invention provide a multi-AGV path planning method and related devices, which can improve the operating efficiency of the multi-AGV system, reduce the number of unnecessary turns in the path and improve the smoothness of the path, and can achieve local obstacle avoidance and avoid falling into local optimality.

[0010] In a first aspect, an embodiment of the present invention provides a multi-AGV path planning method, comprising:

[0011] Step (1): In the single AGV path search phase, the improved A* algorithm is used to plan the global optimal path from the starting point to the target point in the static map, and the Bezier curve is used to perform secondary optimization on the global optimal path to generate a smooth path;

[0012] Step (2): In the multi-AGV path conflict coordination phase, if there is no solution, terminate or modify the path according to the settings; if there is a solution, traverse the paths of all AGVs to check whether there is a conflict;

[0013] Step (3): If there is no conflict, go directly to step (5);

[0014] Step (4): If there is a conflict, assign a dynamic priority to each AGV according to the dynamic priority strategy, and the AGV with a higher priority has the priority for path planning; according to the conflict situation, adjust the path of the conflicting AGV; the AGV with a higher priority retains its path, and the AGV with a lower priority waits or replans its path; if an AGV replans its path, update the path set and perform conflict detection again; if there is still a conflict, repeat steps (1) to (4) until all paths are conflict-free;

[0015] Step (5): Return the final path of all AGVs;

[0016] Step (6): In the multi-AGV online navigation phase, all AGVs are started and use sensors to obtain real-time information about the surrounding environment. The turning points of the global path generated by the improved A* algorithm are used as navigation points for the improved DWA algorithm. If new dynamic obstacles or environmental changes appear during the execution process, the improved DWA algorithm will make real-time local path adjustments based on the latest sensor data to ensure that all AGVs maintain safe movement.

[0017] In some embodiments, the conflict situations include opposite-direction conflicts, node conflicts, and placeholder conflicts;

[0018] If the conflict is a directional conflict, wait or re-plan the path in advance, and the AGV with lower priority will wait at the starting point in advance and set off after the AGV with higher priority passes the repeated path; or re-plan the path for the AGV with lower task priority, and the AGV with higher task priority will follow the original route;

[0019] In the case where the conflict situation is a node conflict, a method is adopted in which the AGV with a lower task priority is ordered to wait for a period of time before departing from the starting point to avoid the conflict;

[0020] When the conflict situation is a place occupation conflict, the path of the AGV with a lower task priority is replanned.

[0021] In some embodiments, planning a global optimal path from a starting point to a target point in a static map using an improved A* algorithm includes:

[0022] First, the improvement strategy for the A* algorithm is to add a turning penalty factor to the path time cost formula to encourage the AGV to choose a path with fewer turns;

[0023] The path cost formula of the A* algorithm is:

[0024] f(n)=g(n)+h(n)

[0025] After adding the steering penalty factor:

[0026] f(n)=g(n)+h(n)+p(n)

[0027] Where g(n) is the path time of the AGV from the starting point to the current node n, and h(n) is the estimated path time from the current node n to the target point;

[0028] Among them, p(n) is the steering penalty function, which is calculated as follows:

[0029]

[0030] in,

[0031] d current =(x current -x previous ,y current -y previous )

[0032] d previous =(x previous -x pre_previous ,y previous -y pre_previous )

[0033] Among them, when the penalty factor n∈(0,1), the steering penalty is increased;

[0034] Secondly, a dynamic weight strategy is adopted to add a dynamic weight coefficient ω to the path cost formula of the A* algorithm;

[0035] f(n)=g(n)+ωh(n)+p(n)

[0036] in,

[0037] ω=e τ

[0038] in,

[0039]

[0040] Among them, (x n ,y n ) is the horizontal and vertical coordinates of the current node; (x s ,y s ) is the horizontal and vertical coordinates of the starting point; (x g ,y g ) are the horizontal and vertical coordinates of the end point.

[0041] In some embodiments, performing secondary optimization on the path using a Bezier curve to generate a smooth path includes:

[0042] The optimization is performed using a third-order Bezier curve, which is represented by the following parametric equation:

[0043] P(t)=p0(1-t) 3 +3p1(1-t) 2 +3p2t 2 (1-t)+p3t 3 t∈[0,1]

[0044] Find the first-order derivative of P(t):

[0045]

[0046] That is

[0047] P′(t)=3(1-t) 2 (p1-p0)+6t(1-t)(p2-p1)+3t 2 (p3-p2)

[0048] The above formula is the expression of velocity. The velocity function is a quadratic function of t and has a smooth transformation.

[0049] Continuing to derive, we get:

[0050]

[0051] That is

[0052] P″(t)=6(1-t)(p2-2p1+p0)+6t(p3-p2+p1)

[0053] The above formula is the expression of acceleration. Acceleration is a linear function of t and is also smooth and continuous.

[0054] In some embodiments, the dynamic priority strategy includes:

[0055] The priority of the a-th AGV passing through node i when executing task j is evaluated, and an evaluation model is constructed based on the current energy consumption E, the current load W, the timeliness requirement U of the current task and the remaining distance S of the current task. To indicate the current load condition W To indicate whether the current AGV is loaded; the timeliness requirement of the current task U is used to measure the urgency of the current task, To express the remaining distance S of the current task, we set the AGV with smaller remaining path length to have a higher priority to complete the task first. To express;

[0056] According to the priority calculation formula

[0057] P=cE+dX+αU+βS

[0058] The priority evaluation model for the a-th AGV passing through node i when performing feeding task j is constructed as follows:

[0059]

[0060] Among them, E min is the minimum power threshold; represents the power consumption of the a-th AGV transport task j; To indicate whether the current AGV is loaded; represents the timeliness requirement of the a-th AGV transport task j, and U represents the maximum timeliness requirement; represents the distance traveled by the a-th AGV transport task j when it arrives at node i; represents the total distance planned for the a-th AGV transport task j; c, d, α, β, γ, and μ are all weight coefficients, where α, β, and γ ∈ (0, 1), which are determined based on the system operation history data using the analytic hierarchy process or fuzzy comprehensive evaluation decision-making method;

[0061] When a conflict occurs and there are AGVs with the same real-time priority, the order of the AGV numbers is used to determine the priority.

[0062] In some embodiments, the improved DWA algorithm includes:

[0063] The evaluation function of the DWA algorithm is:

[0064]

[0065] Introducing the heuristic function in the improved A* algorithm path cost formula into the evaluation function, we get:

[0066]

[0067] The path evaluation function of the DWA algorithm includes three factors: direction, speed and distance, which are expressed as velocity(v) and Among them, ωh(n)+p(n) is the heuristic function of the improved A* algorithm; among them, c, d, α, and β are the weight coefficients of the three factors.

[0068] In some embodiments, the improved A* algorithm includes:

[0069]

[0070] Among them, f(n) is the comprehensive cost value, h(n) is the path time from the current point to the target point, g(n) is the path time from the starting point to the current point, R is the distance from the starting point to the target point, and r is the distance between the current point and the target point.

[0071] In some embodiments, the improved A* algorithm further includes a redundant node deletion strategy, and the redundant node deletion strategy includes:

[0072] Traverse all nodes in the path in sequence. If the current node and the two previous and next nodes are on the same line, remove the current node.

[0073] After removing the redundant points on the same straight line, let the nodes in the path be {P k |k=1,2,…,n}, connect point P1 and point P3. If the distance between line segment P1P3 and the obstacle is greater than the set safety distance, connect P1 and P4 until P1Pk If the distance to the obstacle is less than the set safety distance, connect P1P k -1, remove the middle nodes, and repeat the above operation from point P2 until all nodes in the path are traversed.

[0074] In a second aspect, an embodiment of the present invention further provides an electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the multi-AGV path planning method as described in the first aspect when executing the computer program.

[0075] In a third aspect, an embodiment of the present invention further provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the multi-AGV path planning method as described in the first aspect.

[0076] According to the multi-AGV path planning method and related devices provided by the embodiment of the present invention, the multi-AGV path planning method includes: step (1): in the single AGV path search stage, the improved A* algorithm is used to plan the global optimal path from the starting point to the target point in the static map, and the Bezier curve is used to perform secondary optimization on the global optimal path to generate a smooth path; step (2): in the multi-AGV path conflict coordination stage, if there is no solution, the path is terminated or modified according to the setting; if there is a solution, the paths of all AGVs are traversed to check whether there is a conflict; step (3): if there is no conflict, step (5) is directly executed; step (4): if there is a conflict, a dynamic priority is assigned to each AGV according to the dynamic priority strategy, and the AGV with the higher priority has the priority of path planning; according to the conflict In case of a conflict, adjust the path of the conflicting AGV; the AGV with high priority retains its path, and the AGV with low priority waits or replans its path; if an AGV replans its path, update the path set and perform conflict detection again; if there is still a conflict, repeat steps (1) to (4) until all paths are conflict-free; step (5): return the final path of all AGVs; step (6): in the multi-AGV online navigation stage, all AGVs start, use sensors to obtain real-time information about the surrounding environment, and use the turning points of the global path generated by the improved A* algorithm as navigation points for the improved DWA algorithm; if new dynamic obstacles or environmental changes appear during the execution process, the improved DWA algorithm will perform real-time local path adjustments based on the latest sensor data to ensure that all AGVs maintain safe movement. In the single AGV path search stage, the embodiment of the present invention introduces a dynamic weight coefficient, a steering penalty mechanism, and a Bezier curve smoothing strategy based on the traditional A* algorithm, which can improve the algorithm's operating efficiency, reduce the number of unnecessary turns in the path, and improve the smoothness of the path. In the multi-AGV path conflict coordination stage, traditional avoidance strategies mostly use fixed priority strategies, which do not take into account the relative importance of tasks between AGVs. However, the present invention sets a dynamic priority strategy, which allows avoidance between AGVs to be implemented according to the relative importance of tasks. In the multi-AGV online navigation stage, the traditional DWA algorithm is prone to falling into local optimality, but the present invention can avoid this situation by adopting an improved DWA algorithm, which can achieve both local obstacle avoidance and avoid falling into local optimality. Based on this, the embodiment of the present invention can improve the operating efficiency of the multi-AGV system, reduce the number of unnecessary turns in the path, and improve the smoothness of the path, and can achieve both local obstacle avoidance and avoid falling into local optimality. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 is a flow chart of a multi-AGV path planning method provided by one embodiment of the present invention;

[0078] Figure 2 This is a schematic diagram of a directional collision provided by an embodiment of the present invention;

[0079] Figure 3 This is a schematic diagram of node conflicts provided by an embodiment of the present invention;

[0080] Figure 4 This is a schematic diagram of a placeholder conflict provided by an embodiment of the present invention;

[0081] Figure 5 This is a grid method mapping provided by an embodiment of the present invention;

[0082] Figure 6 This is a schematic diagram of a redundant node deletion strategy provided by an embodiment of the present invention;

[0083] Figure 7A This is a schematic diagram of a grid environment for a smart fishery scenario provided by an embodiment of the present invention;

[0084] Figure 7B This is a schematic diagram of a random scene grid environment provided by an embodiment of the present invention;

[0085] Figure 8 This is an initial path diagram planned by three AGVs provided by an embodiment of the present invention;

[0086] Figure 9 This is a path diagram under a fixed priority strategy provided by an embodiment of the present invention;

[0087] Figure 10 This is a path diagram under a dynamic priority strategy provided by an embodiment of the present invention;

[0088] Figure 11 This is a diagram showing the effect of randomly moving obstacles avoiding obstacles provided by an embodiment of the present invention;

[0089] Figure 12 FIG. 1 is a schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0090] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0091] It should be noted that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and so on, used in the specification, claims, and accompanying drawings are used to distinguish similar items and are not necessarily intended to describe a specific sequence or precedence.

[0092] In the embodiments of the present invention, words such as "further," "exemplarily," or "optionally" are used to indicate examples, illustrations, or explanations and should not be interpreted as being more preferred or advantageous over other embodiments or designs. The use of words such as "further," "exemplarily," or "optionally" is intended to present related concepts in a concrete manner.

[0093] First, some terms involved in this invention are analyzed:

[0094] Automated Guided Vehicle (AGV): AGVs are driverless vehicles typically used to automate material handling and transportation in industrial environments. They use pre-defined paths, sensors, and navigation systems to achieve autonomous movement and are widely used in warehouses, manufacturing plants, distribution centers, and more.

[0095] A multi-AGV path planning algorithm based on conflict search: This algorithm is used to plan paths for multiple automated guided vehicles (AGVs) in a shared environment. It performs initial path planning, detects potential conflict points along the path, generates constraints for each conflict point, and iteratively replans the paths to ensure efficient and conflict-free operation of multiple AGVs.

[0096] A* algorithm: A graph search algorithm used to find the shortest path from a starting point to a destination in a graph. The A* algorithm combines heuristic search and shortest path search to improve search efficiency by prioritizing the expansion of nodes with the lowest estimated total cost. The cost function of the traditional A* algorithm is the sum of the actual path cost (the cost from the starting point to the current node) and the heuristic estimated cost (the estimated cost from the current node to the target node).

[0097] The DWA algorithm is a local path planning algorithm for dynamic obstacle avoidance and path planning in robots. Within a given time window, the DWA algorithm samples the robot's feasible velocity space, evaluates the trajectory of each velocity combination within that time window, and scores the trajectory based on metrics such as safety, target distance, and speed to select the optimal velocity combination. Its primary purpose is to adjust the robot's motion path in real time in a dynamic environment to avoid obstacles and advance toward the target.

[0098] Dynamic Priority Obstacle Avoidance: A multi-robot obstacle avoidance method that prioritizes AGVs to prevent collisions when paths conflict between them. The strategy dynamically prioritizes AGVs based on their battery life, load, mission timeliness, and target location. High-priority robots are prioritized in path planning and obstacle avoidance, ensuring that all robots in the system can complete their tasks safely and efficiently.

[0099] In order to more conveniently describe the working principle of the embodiments of the present invention later, an introduction to relevant technical scenarios is first given below.

[0100] Multi-AGV path planning algorithms are crucial in modern logistics, manufacturing, and warehousing systems. Key challenges include ensuring the efficient and safe movement of multiple AGVs within a shared environment, avoiding collisions, coordinating tasks, and adapting to dynamic environmental changes. For example, in automated warehouses and smart manufacturing systems, the introduction of multi-AGV systems has significantly improved automation and operational efficiency.

[0101] Currently, there are two main approaches to implementing multi-AGV path planning algorithms: Conflict-Based-Search (CBS)-based multi-AGV path planning and reinforcement learning (RL)-based multi-AGV path planning. However, both the CBS-based multi-AGV path planning and reinforcement learning-based multi-AGV path planning algorithms have the following drawbacks:

[0102] 1. Disadvantages of multi-AGV path planning algorithm based on conflict search

[0103] In the single AGV path planning layer of the algorithm, the A* algorithm is mainly used. Its disadvantages are: (1) there are too many meaningless turns in the path generated by the traditional algorithm and the path smoothness is low; (2) the A* algorithm is a static search algorithm. This algorithm is not adaptable to dynamic unknown environments and has limited real-time obstacle avoidance capabilities. Later, it was proposed to integrate the traditional DWA algorithm, but it is easy to fall into the local optimal situation.

[0104] During the multi-AGV path conflict coordination phase, traditional algorithms use a fixed priority strategy. This strategy specifies the priority of each AGV before the algorithm begins. However, this can lead to low multi-AGV system efficiency and AGV deadlock during avoidance. Later, researchers proposed using random priorities. While this improved the deadlock problem, it still leaves the multi-AGV system inefficient due to a failure to balance the timeliness of each AGV's tasks.

[0105] 2. Disadvantages of the multi-AGV path planning algorithm based on reinforcement learning

[0106] Multi-AGV path planning methods based on reinforcement learning are flexible and adaptable in dynamic environments, but they also face some challenges. First, the training process is time-consuming and sample-inefficient, especially in high-dimensional state spaces. In addition, reinforcement learning is difficult to guarantee a global optimal solution and suffers from convergence problems, especially when multiple AGVs interact in complex ways. Although it can cope with dynamic obstacles, the model's generalization ability is weak and may not perform well in untrained environments. At the same time, deep reinforcement learning has high requirements for computing resources, and its strategies are often difficult to explain, resulting in limitations in safety and real-time performance. These problems mean that although reinforcement learning has potential, it still has many shortcomings in actual multi-AGV path planning.

[0107] Based on this, the present invention provides a multi-AGV path planning method and related devices based on an unsupervised industrial large model. Among them, the multi-AGV path planning method includes: step (1): in the single AGV path search stage, the improved A* algorithm is used to plan the global optimal path from the starting point to the target point in the static map, and the Bezier curve is used to perform secondary optimization on the global optimal path to generate a smooth path; step (2): in the multi-AGV path conflict coordination stage, if there is no solution, terminate or modify the path according to the setting; if there is a solution, traverse the paths of all AGVs to check whether there is a conflict; step (3): if there is no conflict, directly execute step (5); step (4): if there is a conflict, assign a dynamic priority to each AGV according to the dynamic priority strategy, and the AGV with a high priority has priority in path planning; according to the conflict situation, adjust the path of the conflicting AGV path; the AGV with high priority retains its path, and the AGV with low priority waits or re-plans its path; if an AGV re-plans its path, the path set is updated and conflict detection is performed again; if there is still a conflict, repeat steps (1) to (4) until all paths are conflict-free; step (5): return the final path of all AGVs; step (6): in the multi-AGV online navigation stage, all AGVs start, use sensors to obtain real-time information about the surrounding environment, and use the turning points of the global path generated by the improved A* algorithm as navigation points for the improved DWA algorithm; if new dynamic obstacles or environmental changes appear during the execution process, the improved DWA algorithm will make real-time local path adjustments based on the latest sensor data to ensure that all AGVs maintain safe movement. In the single AGV path search stage, the embodiment of the present invention introduces a dynamic weight coefficient, a steering penalty mechanism, and a Bezier curve smoothing strategy based on the traditional A* algorithm, which can improve the algorithm's operating efficiency, reduce the number of unnecessary turns in the path, and improve the smoothness of the path. In the multi-AGV path conflict coordination stage, traditional avoidance strategies mostly use fixed priority strategies, which do not take into account the relative importance of tasks between AGVs. However, the present invention sets a dynamic priority strategy, which allows avoidance between AGVs to be implemented according to the relative importance of tasks. In the multi-AGV online navigation stage, the traditional DWA algorithm is prone to falling into local optimality, but the present invention can avoid this situation by adopting an improved DWA algorithm, which can achieve both local obstacle avoidance and avoid falling into local optimality. Based on this, the embodiment of the present invention can improve the operating efficiency of the multi-AGV system, reduce the number of unnecessary turns in the path, and improve the smoothness of the path, and can achieve both local obstacle avoidance and avoid falling into local optimality.

[0108] The embodiments of the present invention are further described below with reference to the accompanying drawings.

[0109] like Figure 1 As shown, Figure 1This is a flowchart of a multi-AGV path planning method provided by an embodiment of the present invention. The multi-AGV path planning method may include but is not limited to steps (1) to (6).

[0110] Step (1): In the single AGV path search phase, the improved A* algorithm is used to plan the global optimal path from the starting point to the target point in the static map, and the Bezier curve is used to perform secondary optimization on the global optimal path to generate a smooth path;

[0111] Step (2): In the multi-AGV path conflict coordination phase, if there is no solution, terminate or modify the path according to the settings; if there is a solution, traverse the paths of all AGVs to check whether there is a conflict;

[0112] Step (3): If there is no conflict, go directly to step (5);

[0113] Step (4): If there is a conflict, assign a dynamic priority to each AGV according to the dynamic priority strategy, and the AGV with a higher priority has the priority for path planning; according to the conflict situation, adjust the path of the conflicting AGV; the AGV with a higher priority retains its path, and the AGV with a lower priority waits or replans its path; if an AGV replans its path, update the path set and perform conflict detection again; if there is still a conflict, repeat steps (1) to (4) until all paths are conflict-free;

[0114] Step (5): Return the final path of all AGVs;

[0115] Step (6): In the multi-AGV online navigation phase, all AGVs are started and use sensors to obtain real-time information about the surrounding environment. The turning points of the global path generated by the improved A* algorithm are used as navigation points for the improved DWA algorithm. If new dynamic obstacles or environmental changes appear during the execution process, the improved DWA algorithm will make real-time local path adjustments based on the latest sensor data to ensure that all AGVs maintain safe movement.

[0116] It can be understood that the multi-AGV path planning method of the present invention can be used for but not limited to smart fishery feeding AGVs, automated warehousing robots, logistics distribution robots, manufacturing AGVs, cleaning robots in shopping malls and public areas, hospitals and medical facilities AGVs, etc.

[0117] It is understandable that conflict situations are mainly divided into three categories, and their corresponding avoidance strategies are as follows:

[0118] (1) Opposite conflict: Figure 2As shown, wait in advance or re-plan the path. The lower priority AGV waits at the starting point in advance and sets off after the higher priority AGV passes the repeated path; or the path is re-planned for the AGV with lower task priority, and the AGV with higher task priority follows the original route.

[0119] (2) Node conflict (intersection conflict): Figure 3 As shown in FIG, a method is adopted to instruct the AGV with lower task priority to wait for a period of time before departing from the starting point to avoid conflicts.

[0120] (3) Placeholder conflict: e.g. Figure 4 As shown in Figure 1, the path of the AGV with lower task priority is replanned.

[0121] It is understandable that when constructing the operating space of the mobile automatic feeding AGV (automatic guided vehicle), the present invention introduces a unitized representation method, namely the grid method strategy, which decomposes the environment around the AGV into a series of cell matrices carrying binary feature codes. This method is not only suitable for static scenes, but also can flexibly respond to dynamically changing environments. For example, in order to achieve accurate depiction of elements such as fish ponds and obstacles in the smart fishery environment, the present invention adopts this grid division strategy to model the operating environment of the AGV (automatic guided vehicle).

[0122] Furthermore, the present invention uses this gridding method to subdivide the operating space shared by multiple AGVs, defining the horizontal direction as the X-axis and the vertical direction as the Y-axis, forming a clear coordinate system. In this system, the AGV is abstracted as a point in space, and its trajectory is composed of a sequence of X and Y coordinate points as it traverses each grid. This representation is similar to discretizing space into a set of points in a two-dimensional coordinate system.

[0123] In order to distinguish between obstacles and non-obstacle areas in the environment, the present invention adopts an intuitive coding mechanism: the grids occupied by obstacles, fish ponds, etc. are represented by dark colors (such as black) and assigned the identification code "1", while the free-travel areas are represented by light-colored (such as white) grids and assigned "0" as the code. This differentiated marking method makes the storage and processing of environmental information more efficient and intuitive. Figure 5 As shown in the figure, the black part represents the fish pond and the white part represents the area where the AGV can drive.

[0124] It is understandable that the present invention utilizes the improved conflict search algorithm to perform path planning for each AGV.

[0125] The traditional conflict search algorithm uses the A* algorithm at the single AGV planning layer to perform global path planning for each AGV. However, due to the low efficiency, numerous inflection points, and poor path smoothness of the traditional A* algorithm in practice, the following strategy is used to improve the A* algorithm at the single AGV planning layer.

[0126] First, the improvement strategy for the A* algorithm is to add a turning penalty factor to the path time cost formula to encourage the robot to choose a path with fewer turns.

[0127] Path cost formula of A* algorithm

[0128] f(n)=g(n)+h(n)

[0129] After adding the steering penalty factor

[0130] f(n)=g(n)+h(n)+p(n)

[0131] Where g(n) is the path time of the AGV from the starting point to the current node n, and h(n) is the estimated path time from the current node n to the target point.

[0132] Among them, p(n) is the steering penalty function, which is calculated as follows

[0133]

[0134] in,

[0135] d current =(x current -x previous ,y current -y previous )

[0136] d previous =(x previous -x pre_previous ,y previous -y pre_previous )

[0137] When the penalty factor n∈(0,1), the steering penalty can be increased to allow the algorithm to expand along a straight line, but the penalty value will not be too large and the algorithm will not choose an expansion point with a larger distance cost in order to maintain the straight line expansion.

[0138] Secondly, to address the low efficiency of the traditional A* algorithm, a dynamic weight strategy is adopted.

[0139] The traditional A* algorithm uses a 1:1 weight ratio between g(n) and h(n). When h(n) is always zero, the A* algorithm degenerates into the Dijkstra path search algorithm. While it can find the optimal path, the algorithm's search efficiency is low. When h(n) is always greater than g(n), the A* algorithm approaches the depth-first search algorithm. While the algorithm is efficient, it struggles to find the optimal path. Therefore, this paper adds a dynamic weight coefficient, ω, to the A* algorithm's path cost formula.

[0140] f(n)=g(n)+ωh(n)+p(n)

[0141] in,

[0142] ω=e τ

[0143] in,

[0144]

[0145] Among them, (x n ,y n ) is the horizontal and vertical coordinates of the current node; (x s ,y s ) is the horizontal and vertical coordinates of the starting point; (x g ,y g ) are the horizontal and vertical coordinates of the end point.

[0146] To solve the problem of uneven path, the third-order Bezier curve is used for optimization.

[0147] Because the third-order Bezier curve is not only continuous in position, but also ensures the continuity of the first-order derivative (velocity) and the second-order derivative (acceleration), this is very important for AGV motion because it can avoid sudden changes in velocity or acceleration along the curve, which can lead to unstable or uncontrollable motion.

[0148] The third-order Bezier curve can be expressed by the following parametric equation

[0149] P(t)=p0(1-t) 3 +3p1(1-t) 2 +3p2t 2 (1-t)+p3t 3 t∈[0,1]

[0150] Find the first-order derivative of P(t)

[0151]

[0152] Right now

[0153] P′(t)=3(1-t) 2(p1-p0)+6t(1-t)(p2-p1)+3t 2 (p3-p2)

[0154] The above formula is the expression of velocity. It can be seen that the velocity function is a quadratic function of t and has a smooth transformation.

[0155] Continue to seek guidance, there is

[0156]

[0157] Right now

[0158] P″(t)=6(1-t)(p2-2p1+p0)+6t(p3-p2+p1)

[0159] The above formula is the expression of acceleration. It can be seen that acceleration is a linear function of t and is also smooth and continuous.

[0160] Understandably, the use of fixed priorities during the multi-AGV conflict coordination phase of the conflict search algorithm can easily lead to AGV deadlock. Later, researchers used randomized priority strategies, which, while somewhat improving the deadlock situation, still failed to account for the relative importance of each AGV's tasks. Therefore, this paper proposes the use of a dynamic priority strategy.

[0161] Dynamic priority avoidance strategies are often used to coordinate path planning in multi-AGV systems, especially when paths intersect or resources compete. To evaluate the priority of the a-th AGV passing through node i when performing task j, an evaluation model can be constructed based on the following dimensions:

[0162] (1) Current energy consumption E

[0163] Check the AGV power to ensure that the vehicle will not be unable to operate due to insufficient power, which will eventually cause system failure. To express.

[0164] (2) Current load condition W

[0165] Compared with the unloaded AGV, the loaded AGV consumes more energy for driving and starting and stopping. Therefore, in order to reduce energy consumption, when the loaded AGV and the unloaded AGV have a path conflict, the unloaded AGV needs to avoid it. To indicate whether the current AGV is loaded.

[0166] (3) Current task timeliness requirement U

[0167] This indicator can measure the urgency of the current task. To express.

[0168] (4) The remaining distance S of the current task

[0169] This indicator can set the AGV with a smaller remaining path length to a higher priority to complete the task first and reduce the probability of causing conflicts. To express.

[0170] According to the priority calculation formula

[0171] P=cE+dX+αU+βS

[0172] The priority evaluation model for the a-th AGV passing through node i when performing feeding task j can be constructed as follows:

[0173]

[0174] Among them, E min is the minimum power threshold (the power threshold in this article is 15%);

[0175] represents the power consumption of the a-th AGV transport task j;

[0176] To indicate whether the current AGV is loaded.

[0177] represents the timeliness requirement of the a-th AGV transport task j, and U represents the maximum timeliness requirement;

[0178] represents the distance traveled by the a-th AGV transport task j when it arrives at node i;

[0179] represents the total distance planned for the a-th AGV transport task j;

[0180] Among them, c, d, α, β, γ, and μ are all weight coefficients, α, β, γ∈(0,1), which can be determined based on the historical data of system operation using decision-making methods such as hierarchical analysis method and fuzzy comprehensive evaluation method.

[0181] When a conflict occurs and there are AGVs with the same real-time priority, the order of the AGV numbers is used to determine the priority.

[0182] The basic idea of DWA is to construct an AGV's speed space, simulate multiple speed trajectories, and use an evaluation function to select the speed corresponding to the optimal trajectory to drive the AGV. The improvement strategy for the DWA algorithm is to redesign the evaluation function.

[0183] The evaluation function of the DWA algorithm is

[0184]

[0185] Introducing the heuristic function in the improved A* algorithm path cost formula into the evaluation function

[0186]

[0187] The path evaluation function of traditional DWA usually includes three factors: direction, speed and distance, which are expressed as velocity(v) and Where ωh(n)+p(n) is the heuristic function of the improved A* algorithm, and c, d, α, and β are the weight coefficients of the three factors.

[0188] The reason for introducing the heuristic function of the improved A* algorithm is to make the AGV drive along the path planned by the improved A* algorithm as much as possible, so as to avoid falling into the local optimum.

[0189] Based on this, compared with the traditional multi-AGV path planning algorithm based on conflict search, the present invention has the following advantages:

[0190] 1. Single AGV path search stage

[0191] Introducing dynamic weight coefficients, turning penalty mechanisms, and Bezier curve smoothing strategies based on the traditional A* algorithm can improve the algorithm's operating efficiency, reduce the number of unnecessary turns in the path, and improve the smoothness of the path.

[0192] 2. Multi-AGV path conflict coordination stage

[0193] Traditional avoidance strategies mostly use fixed priority strategies, which do not consider the relative importance of tasks between AGVs. However, this patent sets a dynamic priority strategy, which can make avoidance between AGVs be implemented according to the relative importance of tasks.

[0194] 3. Multi-AGV online navigation stage

[0195] The traditional DWA algorithm is prone to falling into local optimality, but the improved DWA algorithm proposed in this patent can avoid this situation. It can achieve local obstacle avoidance and avoid falling into local optimality.

[0196] It is understandable that the A* algorithm can be improved by using other solutions, as follows:

[0197] This paper discusses the optimization strategy for the heuristic evaluation function in the traditional A* algorithm. When the heuristic function's predicted cost is lower than the actual cost, the algorithm tends to explore more nodes, increasing computational complexity but ensuring the possibility of finding the global optimal path. Conversely, if the predicted cost is too high, the algorithm converges quickly, improving computational efficiency but potentially missing the optimal path. Ideally, the heuristic function precisely matches the actual cost, achieving the highest search efficiency. Since the present invention uses Euclidean distance as the heuristic basis, its output naturally does not exceed the shortest distance between points in reality. Therefore, when far from the target, the heuristic estimate significantly underestimates the actual distance, forcing the algorithm to search extensively and limiting efficiency. To optimize this situation, the present invention considers increasing the weight of the predicted value to accelerate the search process. As the search progresses, the predicted value gradually approaches the true value. To avoid missing the optimal path due to overestimation, the weight should be appropriately reduced. In summary, this paper proposes a new cost function construction method, as shown in the following formula. This method aims to dynamically adjust the weight of the heuristic function to balance search efficiency and path optimality.

[0198]

[0199] Where f(n) is the comprehensive cost; h(n) is the path time from the current point to the target point; g(n) is the path time from the starting point to the current point. R is the distance from the starting point to the target point, and r is the distance between the current point and the target point.

[0200] While the aforementioned optimization strategy for the A* algorithm effectively improves search performance, the generated path still contains several non-critical redundant nodes, resulting in an unsmooth path. Therefore, a redundant node removal mechanism is proposed to accurately identify and remove redundant parts of the path, retaining only the key turning points that constitute the necessary direction of the path. The specific redundant node removal strategy is shown in the figure below. Figure 6 shown.

[0201] 1) Traverse all nodes in the path in sequence. If the current node is on the same line as the two previous and next nodes, remove the current node.

[0202] 2) After removing the redundant points on the same straight line, let the nodes in the path be {P k |k=1,2,…,n}, connect point P1 and point P3. If the distance between line segment P1P3 and the obstacle is greater than the set safety distance, connect P1 and P4 until P1P k If the distance to the obstacle is less than the set safety distance, connect P1P k -1, remove the middle nodes, and then repeat the above operation from point P2 until all nodes in the path are traversed; after the redundant point removal strategy operation, the path planned by the A* algorithm only contains the starting point, end point and necessary nodes, effectively reducing the path length.

[0203] This paper first conducted comparative experiments using the traditional A* algorithm and the improved A* algorithm in two scenarios of varying complexity. These experiments are shown in Figures 7(a) and 7(b). The black area represents fixed obstacles, and the white area represents the area where the automated feeding AGV can operate. The traditional A* algorithm and the improved A* algorithm were run separately. In Figures 7(a) and 7(b), Curve 1 represents the path planned by the traditional A* algorithm, while Curve 2 represents the path planned by the improved A* algorithm.

[0204] As can be seen from Figure 7(a) and Figure 7(b), the path of the improved A* algorithm is smoother than that of the traditional A* algorithm, and the inflection points are reduced.

[0205] To verify the versatility of the improved algorithm, a 23×23 random grid map was selected for simulation experiments. Each time, 40% of the grid map was randomly generated with random fixed obstacles. The three algorithms were used to solve the problem 100 times in a row. The algorithm performance is summarized in Table 1. Algorithm 1 is the traditional A* algorithm, and Algorithm 2 is the improved A* algorithm.

[0206] Table 1 Algorithm performance table

[0207] algorithm Average running time Average number of turns Traditional A* 0.198s 10.7 times Improving A* 0.0083s 4.2 times

[0208] The experimental results in Table 1 show that the improved A* algorithm reduces the average number of turns by 50.6% over the traditional A* algorithm in 100 experimental results, resulting in a smoother path. In terms of algorithm time, the improved A* algorithm is 58.2% faster.

[0209] For the dynamic programming experiment of multiple AGVs, we take three AGVs as an example. First, we randomly generate tasks for each of the three AGVs. The task information table of the three AGVs is shown in Table 2 below.

[0210] Table 2 Task information of three AGVs

[0211]

[0212]

[0213] For the convenience of description, the row-first storage principle is used to convert the two-dimensional coordinates and index values in the raster image. Assuming that the coordinates of a point in the raster image are (x, y), its index value can be obtained according to the following formula.

[0214] index=y×Width+x

[0215] Where Width represents the total number of rows in the grid map.

[0216] like Figure 8The improved A* algorithm plans the initial paths for the three AGVs. Calculations confirm that there are conflicts among the three AGVs. At the 5th second, AGV2 and AGV3 collide at the node with an index of 55.5. At the 10th second, AGV1 and AGV3 collide in opposite directions between nodes with indexes of 116.5 and 117.5.

[0217] When the traditional conflict resolution strategy (fixed priority) is used to resolve path conflicts between AGVs, the priority order among the three AGVs is: AGV1>AGV2>AGV3.

[0218] First, at the 5th second, AGV2's priority is higher than AGV3's, so AGV2 continues to run along its original route, while AGV3 waits for a while or replans its route. If AGV3 adopts the waiting strategy, it only needs to wait for 1 second at the starting point to avoid conflict with AGV2. At this time, the total time required for AGV3 to complete the feeding task is 17 seconds. If AGV3 replans its route, the new route of AGV3 is: 12.5-11.5-10.5-30.5-50.5-70.5-90.5-110.5-130.5-150.5-170.5-190.5-210.5-230.5-231.5-232.5-233.5

[0219] -234.5-235.5-255.5-275.5, from this we can know that the time it takes for AGV3 to complete the feeding task is 20s.

[0220] Secondly, at the 10th second, AGV1 and AGV3 collide in opposite directions between nodes with index values 116.5 and 117.5. According to the fixed priority conflict resolution rule, AGV1's priority is higher than AGV3's, so AGV1 continues to operate along its original route. AGV3 chooses to wait for a while or replan its route. If AGV3 adopts the strategy of waiting for a while, the waiting time for AGV3 is 13 seconds, and the time required for AGV3 to complete the feeding task is 30 seconds. If the path of AGV3 is replanned, the new path of AGV3 is: 12.5 (waiting for 1 second) - 32.5–33.5–34.5–35.5–55.5–75.5–95.5–115.5–135.5–155.5–175.5–195.5–215.5–235.5–255.5–275.5. At this time, the total time for AGV3 to complete the task is 19 seconds (it needs to wait for 1 second at the starting point to avoid conflict with AGV2).

[0221] Therefore, a fixed priority strategy is used to resolve conflicts between the three feeding AGVs. The specific path diagram is as follows: Figure 9 As shown. Among them, AGV1 and AGV2 continue to maintain the initially planned paths, and AGV3 waits at the starting point for 1 s and re-plans its path.

[0222] According to Table 2, it can be calculated that at the 5th s, AGV2 and AGV3 have an intersection conflict at the node with an index value of 55.5. At this time, P_AGV2 = 1.5111 < P_AGV3 = 1.5175. Therefore, AGV3 continues to run along its original route, and AGV2 adopts a strategy of waiting for a period of time or re-planning its path. If AGV2 adopts the waiting strategy, AGV2 only needs to wait at the starting point for 1 s to avoid conflict with AGV3. At this time, the total time required for AGV2 to complete the feeding task is 10 s; if AGV2's path is re-planned, the new path of AGV2 at this time is: 50.5–30.5–31.5–32.5–33.5–34.5–35.5–36.5–37.5–57.5–77.5–78.5. It can be seen from this that the time for AGV2 to complete the feeding task is 11 s.

[0223] At the 10th s, AGV1 and AGV3 will have a head-on conflict between the nodes with index values of 116.5 and 117.5. According to the dynamic priority conflict resolution rule, at this time, the priority of AGV3 is greater than that of AGV1. Therefore, AGV3 continues to run along its original route. AGV1 adopts a strategy of waiting for a period of time or re-planning its path. If AGV1 adopts the strategy of waiting for a period of time, the time that AGV1 needs to wait at this time is 13 s, and the time required for AGV3 to complete the feeding task is 29 s; if AGV1's path is re-planned, the new path of AGV1 is: 230.5–231.5–232.5–233.5–234.5–235.5–236.5–216.5–196.5–176.5–156.5–136.5–116.5–96.5. At this time, the total time for AGV1 to complete the task is 13 s.

[0224] Therefore, if the dynamic priority strategy is adopted to resolve conflicts for the three feeding AGVs, the specific path diagram after resolution is as Figure 10 shown. Among them, AGV3 continues to maintain the initially planned path, AGV2 waits at the starting point for 1 s and then runs along the original route, and AGV1 re-plans its path.

[0225] The analysis of the experimental results is shown in Table 3 of the experimental results analysis below. The objective function for the multi-AGV path planning optimization in this patent is expressed by the following formula

[0226]

[0227] Among them, Sum_T represents the sum of the travel time and waiting time of all AGVs to complete all package transportation tasks. It represents the driving time of the ath AGV to complete the task package. Indicates the time the a-th AGV needs to wait to complete the task-th task.

[0228] Table 3 Experimental results analysis table

[0229]

[0230] As can be seen from Table 3 above, the three feeding tasks in the smart fishery scenario are completed by adopting the dynamic priority model proposed in this patent, and the efficiency of all automatic feeding AGVs in completing tasks is improved by about 5%.

[0231] The experimental verification of the improved online navigation is as follows Figure 11 As shown. Random moving dynamic obstacles (indicated by dots) are added to the three automatic feeding AGVs respectively. The dotted line is the motion trajectory of the random moving obstacles. Figure 12 The experimental results show that when the automatic feeding AGV detects that it is approaching these obstacles, the improved DWA algorithm replans the local path for the automatic feeding AGV and does not fall into the local optimum.

[0232] In addition, if Figure 12 As shown, an embodiment of the present invention also discloses an electronic device, including: at least one processor 210; at least one memory 220, for storing at least one program; when the at least one program is executed by the at least one processor 210, a multi-AGV path planning method as in any of the previous embodiments is implemented.

[0233] In addition, an embodiment of the present invention further discloses a computer-readable storage medium storing computer-executable instructions for executing the multi-AGV path planning method in any of the foregoing embodiments.

[0234] The system architecture and application scenarios described in the embodiments of the present invention are intended to more clearly illustrate the technical solutions of the embodiments of the present invention and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. Those skilled in the art will appreciate that with the evolution of the system architecture and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present invention are equally applicable to similar technical problems.

[0235] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0236] In a hardware implementation, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, as is well known to those skilled in the art, communication media typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0237] As used in this specification, the terms "component," "module," "system," and the like are used to refer to computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, or a computer. By way of illustration, both applications running on a computing device and a computing device can be components. One or more components can reside in a process or execution thread, and a component can be located on one computer or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures stored thereon. Components can communicate, for example, through local or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component on a local system, a distributed system, or a network, such as the Internet interacting with other systems via signals).

Claims

1. A multi-AGV path planning method, comprising: Step (1): In the single AGV path search phase, the improved A* algorithm is used to plan the global optimal path from the starting point to the target point in the static map, and the Bezier curve is used to perform secondary optimization on the global optimal path to generate a smooth path; Step (2): In the multi-AGV path conflict coordination phase, if there is no solution, terminate or modify the path according to the settings; if there is a solution, traverse the paths of all AGVs to check whether there is a conflict; Step (3): If there is no conflict, go directly to step (5); Step (4): If there is a conflict, assign a dynamic priority to each AGV according to the dynamic priority strategy, and the AGV with a higher priority has the priority for path planning; according to the conflict situation, adjust the path of the conflicting AGV; the AGV with a higher priority retains its path, and the AGV with a lower priority waits or replans its path; if an AGV replans its path, update the path set and perform conflict detection again; if there is still a conflict, repeat steps (1) to (4) until all paths are conflict-free; Step (5): Return the final path of all AGVs; Step (6): In the multi-AGV online navigation phase, all AGVs are started and use sensors to obtain real-time information about the surrounding environment. The turning points of the global path generated by the improved A* algorithm are used as navigation points for the improved DWA algorithm. If new dynamic obstacles or environmental changes appear during the execution process, the improved DWA algorithm will make real-time local path adjustments based on the latest sensor data to ensure that all AGVs maintain safe movement.

2. The method according to claim 1, characterized in that The conflict situations include opposite conflicts, node conflicts and position conflicts; If the conflict is a directional conflict, wait or re-plan the path in advance, and the AGV with lower priority will wait at the starting point in advance and set off after the AGV with higher priority passes the repeated path; or re-plan the path for the AGV with lower task priority, and the AGV with higher task priority will follow the original route; In the case where the conflict situation is a node conflict, a method is adopted in which the AGV with a lower task priority is ordered to wait for a period of time before departing from the starting point to avoid the conflict; When the conflict situation is a place occupation conflict, the path of the AGV with a lower task priority is replanned.

3. The method according to claim 1, characterized in that The improved A* algorithm is used to plan the global optimal path from the starting point to the target point in the static map, including: First, the improvement strategy for the A* algorithm is to add a turning penalty factor to the path time cost formula to encourage the AGV to choose a path with fewer turns; The path cost formula of the A* algorithm is: f(n)=g(n)+h(n) After adding the steering penalty factor: f(n)=g(n)+h(n)+p(n) Where g(n) is the path time of the AGV from the starting point to the current node n, and h(n) is the estimated path time from the current node n to the target point; Among them, p(n) is the steering penalty function, which is calculated as follows: in, d current =(x current -x previous ,y current -y previous ) d previous =(x previous -x pre_previous ,y previous -y pre_previous ) Among them, when the penalty factor n∈(0,1), the steering penalty is increased; Secondly, a dynamic weight strategy is adopted to add a dynamic weight coefficient ω to the path cost formula of the A* algorithm; f(n)=g(n)+ωh(n)+p(n) in, ω=e τ in, Among them, (x n ,y n ) is the horizontal and vertical coordinates of the current node; (x s ,y s ) is the horizontal and vertical coordinates of the starting point; (x g ,y g ) are the horizontal and vertical coordinates of the end point.

4. The method according to claim 1, wherein The method of performing secondary optimization on the path using the Bezier curve to generate a smooth path includes: The optimization is performed using a third-order Bezier curve, which is represented by the following parametric equation: P(t)=p0(1-t) 3 +3p1(1-t) 2 +3p2t 2 (1-t)+p3t 3 t∈[0,1] Find the first-order derivative of P(t): That is P′(t)=3(1-t) 2 (p1-p0)+6t(1-t)(p2-p1)+3t 2 (p3-p2) The above formula is the expression of velocity. The velocity function is a quadratic function of t and has a smooth transformation. Continuing to derive, we get: That is P″(t)=6(1-t)(p2-2p1+p0)+6t(p3-p2+p1) The above formula is the expression of acceleration. Acceleration is a linear function of t and is also smooth and continuous.

5. The method according to claim 1, characterized in that The dynamic priority strategy includes: The priority of the a-th AGV passing through node i when executing task j is evaluated, and an evaluation model is constructed based on the current energy consumption E, the current load W, the timeliness requirement U of the current task and the remaining distance S of the current task. The current energy consumption E is used to To indicate the current load condition W To indicate whether the current AGV is loaded; the timeliness requirement of the current task U is used to measure the urgency of the current task, To express the remaining distance S of the current task, we set the AGV with smaller remaining path length to have a higher priority to complete the task first. To express; According to the priority calculation formula P=cE+dX+αU+βS The priority evaluation model for the a-th AGV passing through node i when performing feeding task j is constructed as follows: Among them, E min is the minimum power threshold; represents the power consumption of the a-th AGV transport task j; To indicate whether the current AGV is loaded; represents the timeliness requirement of the a-th AGV transport task j, and U represents the maximum timeliness requirement; represents the distance traveled by the a-th AGV transport task j when it arrives at node i; represents the total distance planned for the a-th AGV transport task j; c, d, α, β, γ, and μ are all weight coefficients, where α, β, and γ ∈ (0, 1), which are determined based on the system operation history data using the analytic hierarchy process or fuzzy comprehensive evaluation decision-making method; When a conflict occurs and there are AGVs with the same real-time priority, the order of the AGV numbers is used to determine the priority.

6. The method according to claim 1, characterized in that The improved DWA algorithm includes: The evaluation function of the DWA algorithm is: Introducing the heuristic function in the improved A* algorithm path cost formula into the evaluation function, we get: The path evaluation function of the DWA algorithm includes three factors: direction, speed and distance, which are expressed as velocity(v) and Among them, ωh(n)+p(n) is the heuristic function of the improved A* algorithm; among them, c, d, α, and β are the weight coefficients of the three factors.

7. The method according to claim 1, characterized in that The improved A* algorithm includes: Among them, f(n) is the comprehensive cost value, h(n) is the path time from the current point to the target point, g(n) is the path time from the starting point to the current point, R is the distance from the starting point to the target point, and r is the distance between the current point and the target point.

8. The method according to claim 1, characterized in that The improved A* algorithm also includes a redundant node deletion strategy, which includes: Traverse all nodes in the path in sequence. If the current node and the two previous and next nodes are on the same line, remove the current node. After removing the redundant points on the same straight line, let the nodes in the path be {P k |k=1,2,…,n}, connect point P1 and point P3. If the distance between line segment P1P3 and the obstacle is greater than the set safety distance, connect P1 and P4 until P1P k If the distance to the obstacle is less than the set safety distance, connect P1P k -1, remove the middle nodes, and repeat the above operation from point P2 until all nodes in the path are traversed.

9. An electronic device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the multi-AGV path planning method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the multi-AGV path planning method according to any one of claims 1 to 7.

Images