An AGV transportation method and robot

Abstract

The present invention belongs to the technical field of AGV, and provides an AGV transportation method and a robot. The method includes: obtaining a task list and the initial positions of each AGV car; using tabu search to optimize the task execution order to minimize the weighted cost of the last task completion time, average completion time, and number of stops; constructing a spatio-temporal graph representing time and spatial positions of nodes for each AGV car according to the optimized order; using a time enhancement algorithm to plan the paths of each AGV car on the spatio-temporal graph, regarding the paths of other AGV cars as time obstacles, and using a path cost function; allowing the AGV cars to wait in place during path planning to avoid collisions, and updating the paths of each AGV car in real time; smoothing the planned discrete paths with cubic Bezier curves to generate a continuous and controllable driving trajectory; finally, controlling each AGV car to execute tasks along the trajectory. The present invention can solve the problems of low task scheduling efficiency and frequent path conflicts of multiple AGV cars in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of AGV, and in particular to an AGV transportation method and a robot. Background Art

[0002] Automated guided vehicles (AGVs), also known as transport robots, are widely used in warehousing, logistics, and production workshops to achieve flexible material handling and task scheduling as the level of intelligent manufacturing and industrial automation continues to improve. In a multi-AGV transport robot system, how to coordinate multiple AGVs to perform tasks efficiently and orderly in a limited space has become a key issue affecting the overall performance of the system.

[0003] Existing multi-AGV path planning methods usually adopt static or semi-dynamic scheduling modes. Paths and task assignments are uniformly generated before the system is started. They lack the ability to respond to dynamic changes (such as task updates, traffic congestion, and path conflicts), which can easily lead to collisions, deadlocks, or task delays between AGVs. On the other hand, some path planning algorithms do not fully consider the time dimension, resulting in overlapping occupancy in time even if the paths do not conflict in space, thus affecting the efficiency of multi-vehicle collaboration.

[0004] In addition, existing task scheduling strategies mostly use heuristic methods or simple rules (such as shortest distance first, first come, first served, etc.), which make it difficult to globally optimize the task allocation order. Especially in scenarios with a large number of tasks and complex vehicle conditions, the system operation efficiency and stability are insufficient.

[0005] Therefore, there is an urgent need for an AGV transport robot that has both dynamic scheduling capabilities and spatiotemporal collision avoidance path planning capabilities, which can combine time-enhanced path search with task sorting optimization to minimize task execution time and path conflicts, thereby improving the overall operating efficiency and safety of the system. Summary of the invention

[0006] In view of the above technical problems, the present invention provides an AGV transportation method and a robot to solve the problems of low task scheduling efficiency and frequent path conflicts of multiple AGV vehicles in the prior art.

[0007] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0008] According to one aspect of the present invention, an AGV transportation method is disclosed, the method comprising:

[0009] Receive a task list containing multiple tasks and the initial position of each AGV;

[0010] Use the tabu search algorithm to optimize the execution order of tasks in the task list to minimize the weighted cost function composed of the time taken by the last AGV to complete the task, the average of the task completion times of all AGVs, and the total number of stops of all AGVs.

[0011] Construct a spatio-temporal graph for each AGV according to the optimized task order. The nodes of the spatio-temporal graph represent the spatial positions of the AGV at specific time steps, and the edges between the nodes represent the AGV moving from the current position to the next position or staying in place between adjacent time steps.

[0012] Use the time enhancement algorithm to search for paths for each AGV on a three-dimensional spatio-temporal graph that includes the time dimension. Consider the planned paths of other AGVs as time obstacles, and adopt the path cost function f(j,k)=αg(j,k)+βh(j,k), where g is the actual cost of the current path, h is the heuristic cost to the target position, and α and β are preset weights.

[0013] During the path search process, allow the AGV to wait in place when necessary to avoid collisions with other AGVs, and update the path planning of each AGV in real time in an online manner.

[0014] Perform cubic Bezier curve smoothing on the discrete paths obtained by the time enhancement algorithm to generate a continuous and controllable driving trajectory.

[0015] Control each AGV to execute its corresponding task according to the driving trajectory.

[0016] Furthermore, in the tabu search algorithm, a predetermined heuristic algorithm is used to generate an initial solution for the task execution order, and a tabu list is established by setting the tabu length to record the visited solution sequences within the tabu length range to avoid the search process falling into a local optimum.

[0017] Furthermore, the constructed spatio-temporal graph adopts a layer structure with discrete time steps, and each layer corresponds to a discrete time step.

[0018] Furthermore, in the path cost function f(j,k) of the time enhancement algorithm, g(j,k) is defined as the cumulative actual cost from the starting position to the node (j,k), h(j,k) is defined as the heuristic estimated cost from the node (j,k) to the target position, and α and β are preset weights used to balance the actual cost and the heuristic cost.

[0019] Further, during the path search process, the adjacent nodes of the AGV at each node include nodes that move to passable positions adjacent in space at the next time step or stay in place at the current node. For adjacent nodes where there are static obstacles or positions occupied by other AGVs at the same time, they are regarded as time obstacles and not considered.

[0020] Further, when performing cubic Bezier curve smoothing on the discrete path, the control point generation rules include:

[0021] For three consecutive path points in the discrete path, calculate the midpoints of the lines connecting the previous path point and the middle path point and the middle path point and the subsequent path point, and translate the line segment connecting the two midpoints to the middle path point. Use the two endpoints of the translated line segment as the control points of the adjacent Bezier curve segments at the corresponding middle path point to make the tangent directions of the adjacent curve segments consistent at the connection to ensure the smooth continuity of the trajectory.

[0022] Further, a waiting-in-place mechanism is provided during the path search process. When the next move of the AGV will cause a conflict with the planned paths of other AGVs, it is determined that the waiting action of the corresponding AGV is to wait for one time step at the current position, and the waiting action is used as an optional adjacent node for path expansion, and a time cost equivalent to one move or a preset waiting penalty cost is set for it.

[0023] Further, the path planning includes a path dynamic update and a scheduling feedback mechanism. When path conflicts, delays, or task changes occur during the execution of tasks by multiple AGVs, an online replanning of the remaining paths of the relevant AGVs is triggered in real time, and the updated path planning results are collected and fed back to adjust the task execution order or the path planning of other AGVs to achieve dynamic coordination of the scheduling of multiple AGVs.

[0024] According to another aspect of the present invention, an AGV transport robot is disclosed, including an AGV, and the AGV includes an intelligent control system integrated with or separated from it, and the intelligent control system is used to execute the method as described above.

[0025] The technical solution of the present disclosure has the following beneficial effects:

[0026] Through the introduction of a time enhancement algorithm and a tabu search scheduling mechanism, the present invention realizes the collaborative optimization of task allocation and path planning of multiple AGVs, can effectively reduce the total task execution duration, reduce the vehicle waiting times, improve the system operation efficiency and safety, and meet the real-time scheduling requirements in a dynamic environment. Description of the Drawings

[0027] Figure 1 This is a structural block diagram of an AGV transport robot and an intelligent control system in an embodiment of this specification;

[0028] Figure 2 This is a flowchart of an AGV transport method in an embodiment of this specification. Specific embodiments

[0029] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be more thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of the embodiments of the present disclosure. However, those skilled in the art will realize that the technical solutions of the present disclosure can be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be used. In other cases, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0030] In addition, the accompanying drawings are only schematic illustrations of the present disclosure. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0031] As Figure 1 shown, an embodiment of this specification provides an AGV transport robot. The AGV cart includes an intelligent control system integrated with or separated from it. The intelligent control system is used to execute the AGV transport method. As Figure 2 shown, the AGV transport method includes steps S101 - S107:

[0032] In step S101, a task list containing multiple tasks and the initial positions of each of the AGV carts are received.

[0033] Among them, the task list consists of multiple tasks to be executed. For example, each task (or a "task group") can include several consecutive working phases: for example, the AGV cart starts from the starting position, travels to the designated picking location to complete picking up goods, then transports the materials to the target placement station to complete unloading, and finally returns to the predetermined rest position. The initial position of each AGV cart above is input as the starting point of path planning, and together with the task list, it constitutes the input information of the scheduling and path planning module. Based on this, the system obtains all task requirements and vehicle starting states for subsequent task allocation, path planning, and coordinated control.

[0034] In step S102, the tabu search algorithm is used to optimize the execution order of tasks in the task list to minimize the weighted cost function composed of the time taken for the last AGV cart to complete the tasks, the average value of the task completion times of all AGV carts, and the total number of stops of all AGV carts.

[0035] Among them, the tabu search algorithm is used to optimize the execution order of tasks in the task list. Tabu search is a heuristic algorithm based on iterative improvement. It explores new feasible solutions by continuously making small perturbations to the current scheduling scheme and uses a "tabu list" to record recent search state changes to avoid cycles in the solution space during the search process. In this embodiment, each task execution order scheme is evaluated by a weighted cost function, which comprehensively considers three target indicators for multi-AGV scheduling: (1) the time taken for the last AGV cart to complete all its tasks (i.e., the total time taken to complete all tasks); (2) the average value of the time taken for each AGV cart to complete its tasks; and (3) the total number of stops of all AGV carts during the execution process. The above three indicators are respectively assigned preset weight parameters λ1, λ2, and λ3, and are weighted and summed to form the overall target cost value c. The tabu search algorithm can first use heuristic methods such as the nearest neighbor to generate an initial task allocation and execution sequence, and then generate neighborhood solutions in each iteration by swapping task orders, adjusting task allocations, etc., and calculate the above cost function values for each neighborhood solution to evaluate its quality. The algorithm selects the solution with a lower cost function value (i.e., a better scheduling scheme) as the new current solution and records this adjustment in the tabu list to prohibit reverse operations in subsequent iterations, thereby avoiding repeated search of the previously explored solution space. Repeating this iteration continuously updates the task order scheme until a predetermined stop condition is met (such as the number of iterations reaches the upper limit or no better solution has been found for multiple consecutive generations). Finally, the tabu search outputs an approximately optimal task execution order scheme, making the value of the above weighted cost function as small as possible, that is, the comprehensive indicators such as the time to complete the last task, the average completion time, and the number of stops are close to the optimal.

[0036] In step S103, a spatio-temporal graph is constructed for each of the AGV vehicles according to the optimized task sequence. The nodes of the spatio-temporal graph represent the spatial positions of the AGV vehicles at specific time steps, and the edges between the nodes represent the movement of the AGV vehicles from the current position to the next position or staying in place between adjacent time steps.

[0037] Among them, a corresponding spatio-temporal graph is constructed for each of the AGV vehicles according to the optimized task sequence. The spatio-temporal graph extends the motion space of the AGV vehicle with a time dimension, forming a three-dimensional grid graph model. Specifically, the nodes of the spatio-temporal graph represent the spatial positions of the AGV vehicles at specific time steps, that is, associating discrete time and spatial positions: for each possible discrete time k, a set of nodes corresponding to each reachable position on the map is established. Connection edges are defined between the nodes to represent the state transition of the AGV vehicle moving from one time step to the next time step. For example, when there is a road connecting two spatial nodes A and B, a directed edge is established between the node A at time k and the node B at time k + 1, indicating that the AGV can move from position A to position B within one time unit. If the AGV chooses to wait in place at a certain time step, a self-loop edge is established between the same position nodes corresponding to time k and time k + 1, indicating that the vehicle stays in place during this time interval. In this way, the constructed spatio-temporal graph covers all possible position sequences of the AGV vehicle from the starting position to the target position at each discrete time point. The initial position of each AGV vehicle corresponds to the starting node of the initial time layer (e.g., k = 0) in the spatio-temporal graph, and the task target position corresponds to the target node of the termination time layer. With the help of this spatio-temporal graph model, the AGV path planning algorithm can search for spatial paths considering time factors, laying a foundation for the collision avoidance path planning in the subsequent steps.

[0038] In step S104, a time enhancement algorithm is used to search for paths for each of the AGV vehicles on the three-dimensional spatio-temporal graph including the time dimension, regarding the planned paths of other AGV vehicles as time obstacles, and adopting the path cost function f(j,k)=αg(j,k)+βh(j,k), where g is the actual cost of the current path, h is the heuristic cost to the target position, and α and β are preset weights.

[0039] Specifically, an improved A search algorithm that introduces the time dimension (i.e., the time-enhanced A algorithm) is adopted to plan paths for each AGV vehicle in sequence according to the optimal execution order obtained in step S102. When searching for a path for a certain AGV vehicle, the paths already planned by other AGV vehicles are regarded as time obstacles: that is to say, for those position nodes in the spatio-temporal graph that will be occupied by other vehicles at a specific time, they are all marked as occupied states and cannot be used as the feasible moving positions of the current AGV vehicle. By taking the running trajectories of other vehicles as dynamic obstacles that change with time as input, this algorithm ensures that different AGV vehicles will not compete for the same spatial position at the same time, thus naturally avoiding collision conflicts. During the path search process, the heuristic evaluation function f(j,k)=α·g(j,k)+β·h(j,k) of the A algorithm is used to score the candidate paths, where g(j,k) represents the actual cumulative cost from the starting point to the current node j (at time layer k), such as the travel time or path distance, h(j,k) represents the heuristic estimated cost from this node to the target, and α and β are preset weights used to balance the relative influence of the actual cost and the heuristic cost in the evaluation function. The algorithm selects nodes with smaller costs for priority expansion based on this evaluation function for search until the optimal or sub-optimal path from the starting point to the target node is found. Through the time-enhanced A algorithm, each AGV vehicle can find a driving path that is spatially reachable and avoids conflicts with other AGV vehicles in time.

[0040] In step S105, during the path search process, the AGV vehicle is allowed to wait in place when necessary to avoid collisions with other AGVs, and the path planning of each AGV is updated in real time in an online manner. That is to say, if at a certain time step, all potential moving positions of the current AGV vehicle are occupied by other vehicles or will cause conflicts, the AGV vehicle can choose to stay at this position for one time unit and wait for the front road to become free, rather than being forced to detour a long distance. Incorporating the "wait in place" action into the path planning enables the algorithm to flexibly adjust the traveling rhythm of the AGV vehicle and stagger the possible conflicts between vehicles in time. In this way, even if multiple AGV vehicles share some of the same spatial routes, they can achieve safe passage through time-based error correction. In addition, the path planning adopts a strategy of real-time update in an online manner: during the execution process, the system can re-plan or correct the paths of each AGV in real time according to environmental changes or dynamic adjustments of the task queue. When the actual driving progress of a certain AGV vehicle deviates from the original plan, or when a new task is added, the scheduling and path planning module will immediately update the driving trajectory of the affected AGV to ensure that the entire control system always remains collision-free, coordinated, and efficient during operation.

[0041] In step S106, the discrete path obtained through the time enhancement algorithm is smoothed using cubic Bézier curves to generate a continuous and controllable driving trajectory.

[0042] Specifically, a series of discrete inflection points obtained from path planning are connected by cubic Bézier curve segments to smooth the turning angles of the broken lines on the path and ensure the smoothness of the AGV's turning transition. A cubic Bézier curve is defined by four control points, where the starting and ending points of the curve correspond to two adjacent nodes of the original discrete path, and the middle two control points are set according to the desired trajectory shape to adjust the curvature and inclination of the curve so that it connects with the path direction at the starting and ending points. By moving along the curve with a parameter variable (ranging from 0 to 1), the coordinate positions of each point on the curve can be easily calculated, thus converting the discrete path into a parameterized continuous trajectory for the controller to reference. After being smoothed by cubic Bézier curves, the driving trajectory of the AGV is continuous in space and curvature, without sharp turning changes, which ensures the operability and stability of the AGV's motion control and reduces pauses and mechanical losses caused by discontinuous paths.

[0043] The cubic Bézier curve can be expressed as:

[0044] ;

[0045] ;

[0046] represents an integer value between 0 and 1 indicating the position of the AGV in the curve and the progress of the AGV's movement along this side. When = 0, (x, y) = (x0, y0), where (x0, y0) is the starting point of the curve. When = 1, (x, y) reaches the end point. As continuously changes within the interval, the control points ( , ) move smoothly on the curve, and the coefficients , are determined by the control points of the path and are used to adjust the shape and curvature of the curve. Specifically, ( , ) describes the continuous motion trajectory of the AGV in the plane. As the parameter increases, the position of the AGV changes continuously along the curve, which means the AGV moves from one node to the next in a smooth manner without discrete jumps.

[0047] In step S107, each AGV is controlled to execute its corresponding task according to the driving trajectory.

[0048] Among them, the control system uses the smoothed trajectory as the driving plan for each AGV vehicle, and issues specific motion instructions (such as speed setting and steering angle) to guide the AGV vehicle to travel along the predetermined trajectory. After each AGV vehicle starts from its initial position, it sequentially passes through each station specified in the task list according to the planned trajectory, and performs the predetermined task operations (such as material picking and placing, loading and unloading, etc.) when reaching the corresponding position, and then continues to move along the trajectory to the next target point. Since the overall coordination and obstacle avoidance verification of the vehicle trajectories of all AGVs have been carried out in the previous planning stage, each AGV vehicle can complete its respective tasks without interference, and even on the shared road section, it can avoid conflicts by staggering the traffic peaks. Finally, all AGV vehicles efficiently complete the tasks assigned to them according to the optimized task sequence and planned path, and multiple AGV vehicles cooperate to complete all operation objectives in the task list.

[0049] In one embodiment, in the tabu search algorithm, a predetermined heuristic algorithm is used to generate an initial solution of the task execution order, and a tabu list is established by setting a tabu length to record the visited solution sequence within the tabu length range to avoid the search process falling into a local optimum. For example, through the nearest neighbor heuristic method, according to the position where each AGV vehicle was when it last completed a task, the task with the shortest travel distance among the remaining tasks is preferentially assigned to that AGV vehicle. Then, a tabu list is established by setting a predetermined tabu length, and the visited solution sequence or the characteristic changes of the solution are recorded within this tabu length range. When the candidate solution generated in the subsequent iteration matches the solution sequence recorded in the tabu list, the selection of this solution is skipped and it is regarded as tabu to avoid repeated search, thereby preventing the search process from returning to the checked solution and falling into a local optimum.

[0050] In one embodiment, the constructed spatio-temporal graph adopts a layer structure with discrete time steps, and each layer corresponds to a discrete time step. Specifically, the task environment is abstracted into a graph model with a time dimension, the time axis is discretized into equally spaced steps k (k = 0, 1, 2,..., kmax), and each k value corresponds to a layer of spatial topology structure. In each time layer, it contains all the feasible walking position nodes corresponding to the actual map and their connectivity relationships, and marks whether each node is occupied or has obstacles according to the situation at that moment. With the above multi-layer spatio-temporal graph representation, the path planning of the AGV vehicle can be carried out synchronously in two dimensions of spatial position and time, so that the path is composed of a sequence of nodes connected in chronological order, ensuring that the planning process fully considers the time sequence relationship and the dynamic avoidance requirements.

[0051] In one embodiment, in the path cost function f(j,k) of the time enhancement algorithm, g(j,k) is defined as the cumulative actual cost from the starting position to node (j,k), h(j,k) is defined as the heuristic estimated cost from node (j,k) to the target position, and α and β are preset weights used to balance the actual cost and the heuristic cost. Among them, by appropriately selecting the values of α and β, the relative weights of the actual cost such as the driving distance and the heuristic estimated cost of the algorithm can be adjusted, thereby improving the path search efficiency and maintaining the rationality of the path planning result.

[0052] In one embodiment, during the path search process, the adjacent nodes of the AGV vehicle at each node include the nodes that move to the passable positions adjacent in space at the next time step or stay in place at the current node. For the adjacent nodes where there are static obstacles or the positions are occupied by other AGV vehicles at the same time, they are regarded as time obstacles and not considered.

[0053] Among them, during the path search process, the optional adjacent nodes of the AGV vehicle at each node include two categories: one is the node corresponding to moving to the passable position adjacent in space at the next time step, and the other is the node corresponding to staying in place at the current node for one time step. For the adjacent nodes where there are static obstacles or the positions are occupied by other AGV vehicles at the same time, they are regarded as obstacles in the time dimension (time obstacles) and not considered, that is, this state will not be added to the candidate set during path expansion for evaluation. By the above method, it can be ensured that multiple AGV vehicles do not occupy the same position at any discrete time step, thus avoiding collision conflicts in path planning.

[0054] In one embodiment, when performing cubic Bezier curve smoothing on the discrete path, the control point generation rules include:

[0055] For three consecutive path points in the discrete path, calculate the midpoints of the lines connecting the previous path point and the middle path point and the middle path point and the subsequent path point, translate the line segment connecting the two midpoints to the middle path point, and use the two endpoints of the translated line segment as the control points of the adjacent Bezier curve segments at the corresponding middle path point, so that the tangent directions of the adjacent curve segments are consistent at the connection, to ensure the smooth continuity of the trajectory.

[0056] Among them, through the above rules, the tangent directions of the adjacent curve segments can be kept consistent at the connection, thereby ensuring the smooth continuity of the generated trajectory.

[0057] In one embodiment, a waiting-in-place mechanism is provided during the path search process. When the next move of the AGV cart will cause a conflict with the planned paths of other AGV carts, it is determined that the waiting action for the corresponding AGV cart is to wait for one time step at the current position, and this waiting action is taken as an optional adjacent node for path expansion, and a time cost equivalent to one move or a preset waiting penalty cost is set for it.

[0058] Among them, when it is predicted that the next move of the AGV cart will cause a conflict with the planned paths of other AGV carts, it is determined that the corresponding AGV cart needs to wait for one discrete time step at the current position, and this waiting action is added as an optional adjacent node during path expansion for search. For the waiting action, a time cost equivalent to one move is given in the path cost evaluation, or a preset waiting penalty cost is given as appropriate. By introducing the above waiting mechanism, the AGV cart can actively pause at the moment of potential conflict, sacrificing time for space to avoid collisions, so that the overall path planning can avoid collisions through staggered scheduling.

[0059] In one embodiment, the path planning includes a path dynamic update and a scheduling feedback mechanism. When path conflicts, delays or task changes occur during the execution of tasks by multiple AGV carts, an online replanning of the remaining paths of the relevant AGV carts is triggered in real time, and the updated path planning results are collected and fed back to adjust the task execution order or the path planning of other AGV carts, so as to achieve dynamic coordination of the scheduling of multiple AGV carts.

[0060] Among them, through the above dynamic update and scheduling feedback mechanism, real-time dynamic coordination of the scheduling of multiple AGV carts is achieved, that is, during the task execution process, the driving routes and task assignments of each AGV cart are optimized in a timely manner according to the changes in the actual situation to ensure the efficient and orderly execution of tasks in the entire multi-AGV system.

[0061] Beneficial effects:

[0062] By introducing the time enhancement algorithm and the tabu search scheduling mechanism, the present invention realizes the collaborative optimization of the task assignment and path planning of multiple AGV carts, can effectively reduce the total task execution duration, reduce the number of vehicle waiting times, improve the system operation efficiency and safety, and meet the real-time scheduling requirements in a dynamic environment.

[0063] Those skilled in the art will readily conceive of other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include known common general knowledge or conventional technical means in the technical field not disclosed in the present disclosure. The specification and embodiments are only to be considered as exemplary, and the true scope and spirit of the present disclosure are pointed out by the claims.

Claims

1. An AGV transportation method, characterized in that, The method includes: Receiving a task list containing multiple tasks and the initial positions of each AGV vehicle; Using a tabu search algorithm to optimize the execution order of the tasks in the task list to minimize a weighted cost function composed of the time taken by the last AGV vehicle to complete the tasks, the average of the task completion times of all the AGV vehicles, and the total number of stops of all the AGV vehicles; Constructing a spatio-temporal graph for each AGV vehicle according to the optimized task order, where the nodes of the spatio-temporal graph represent the spatial positions of the AGV vehicle at specific time steps, and the edges between the nodes represent the AGV vehicle moving from the current position to the next position or staying in place between adjacent time steps; Using a time enhancement algorithm to search for paths for each AGV vehicle on a three-dimensional spatio-temporal graph including the time dimension, regarding the planned paths of other AGV vehicles as time obstacles, and adopting a path cost function f(j,k)=αg(j,k)+βh(j,k), where g is the actual cost of the current path, h is the heuristic cost to the target position, and α and β are preset weights; in the path cost function f(j,k) of the time enhancement algorithm, g(j,k) is defined as the cumulative actual cost from the starting position to the node (j,k), h(j,k) is defined as the heuristic estimated cost from the node (j,k) to the target position, and α and β are preset weights for balancing the actual cost and the heuristic cost; During the path search process, allowing the AGV vehicle to wait in place when necessary to avoid collisions with other AGV vehicles, and updating the path planning of each AGV in an online manner in real time; Performing cubic Bezier curve smoothing on the discrete path obtained by the time enhancement algorithm to generate a continuous and controllable driving trajectory; when performing cubic Bezier curve smoothing on the discrete path, the control point generation rule includes: for three consecutive path points in the discrete path, calculating the midpoints between the previous path point and the middle path point and between the middle path point and the subsequent path point, translating the line segment connecting the two midpoints to the middle path point, and using the two endpoints of the translated line segment as the control points of the adjacent Bezier curve segments at the corresponding middle path point to make the tangent directions of the adjacent curve segments consistent at the connection to ensure the smooth continuity of the trajectory; Controlling each AGV vehicle to execute its corresponding task according to the driving trajectory.

2. The AGV transportation method according to claim 1, wherein In the tabu search algorithm, a predetermined heuristic algorithm is used to generate an initial solution of the task execution order, and a tabu list is established by setting a tabu length to record the visited solution sequences within the tabu length range to avoid the search process falling into a local optimum.

3. The AGV transportation method according to claim 1, wherein The constructed spatio-temporal graph adopts a layer structure with discrete time steps, and each layer corresponds to a discrete time step.

4. The AGV transportation method according to claim 1, wherein The method further includes: during the path search process, the adjacent nodes of each node of the AGV vehicle include nodes that move to accessible positions adjacent in space at the next time step or stay in place at the current node. For adjacent nodes where there are static obstacles or positions occupied by other AGV vehicles at the same time, they are regarded as time obstacles and not considered.

5. The AGV transportation method according to claim 1, wherein, The method further includes: there is an in-situ waiting mechanism during the path search process. When the next move of the AGV vehicle will cause a conflict with the planned paths of other AGV vehicles, it is determined that the waiting action corresponding to the AGV vehicle is to wait for one time step at the current position, and the waiting action is used as an optional adjacent node for path expansion, and a time cost equivalent to one move or a preset waiting penalty cost is set for it.

6. The AGV transportation method according to claim 1, wherein The path planning includes a path dynamic update and a scheduling feedback mechanism. When path conflicts, delays or task changes occur during the task execution of multiple AGV vehicles, the online replanning of the remaining paths of the relevant AGV vehicles is triggered in real time, and the updated path planning results are collected and fed back to adjust the task execution order or the path planning of other AGV vehicles, so as to achieve the dynamic coordination of the scheduling of multiple AGV vehicles.

7. An AGV transport robot, characterized in that, It includes an AGV vehicle, and the AGV vehicle includes an intelligent control system integrated with or separated from it. The intelligent control system is used to execute the method according to any one of claims 1-6.

Images