Multi-target collaborative optimization intelligent multi-robot carrying scheduling system and method

Abstract

The invention provides a multi-target collaborative optimization intelligent multi-robot carrying scheduling system and method. The method comprises four core parts: a multi-robot task assignment algorithm, a single-robot walking path optimization algorithm, a multi-robot traffic control algorithm and a multi-robot deadlock problem solving algorithm. The robot task assignment algorithm is used for selecting an optimal task assignment scheme according to the comprehensive evaluation strategy; a single-robot walking path optimal algorithm: searching an optimal walking path of each robot on a grid map by adopting an A * algorithm; according to a multi-robot traffic control algorithm, anti-collision control among multiple robots is achieved through a grid locking mechanism; according to the multi-robot deadlock solution algorithm, a random and polling timeout waiting mode is adopted, and a path recalculation strategy is combined. Task assignment, path planning, traffic control and deadlock solution strategies are comprehensively optimized, and the overall operation efficiency and safety of the multi-robot system in carrying operation are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of intelligent robot technology and automated dispatching system, and in particular to a multi-robot handling dispatching system and method based on multi-objective optimization. Background Art

[0002] With the rapid development of Industry 4.0 and intelligent manufacturing, multi-robot systems are increasingly being used in automated warehouses, production lines, logistics and distribution. These systems can efficiently complete tasks such as handling, sorting, and assembly by working together, significantly improving production efficiency and reducing costs.

[0003] However, existing multi-robot handling and scheduling systems often face problems such as uneven task allocation, unreasonable path planning, traffic conflicts between robots, and deadlocks, resulting in low overall system operation efficiency and even safety hazards.

[0004] The limitations of the existing system are as follows:

[0005] 1. Single task assignment: Existing systems usually adopt fixed task assignment strategies, which are difficult to adapt to changing business scenarios and demand changes, resulting in some robots being overloaded while other robots are idle, resulting in low resource utilization.

[0006] 2. Poor path planning: Traditional path planning algorithms often have difficulty finding the optimal path in complex environments, resulting in long walking distances and long walking times for the robot, affecting handling efficiency.

[0007] 3. Insufficient traffic control: In a robot-intensive operating environment, the lack of an effective traffic control mechanism can easily lead to collisions between robots and traffic jams, affecting system stability and safety.

[0008] 4. Deadlock problem is difficult to solve: When multiple robots meet in a narrow area or intersection, deadlock is likely to occur, causing the system to crash. Existing solutions often lack flexibility and effectiveness, making it difficult to quickly solve the deadlock problem. Summary of the invention

[0009] The purpose of the present invention is to provide a multi-objective collaborative optimization intelligent multi-robot handling scheduling system and method, aiming to solve the problems of low operating efficiency and insufficient safety faced by multi-robot systems in complex handling operation environments. By integrating a variety of advanced algorithms and technologies, the present invention realizes the comprehensive optimization of task assignment, path planning, traffic control and deadlock resolution strategies, optimizes the handling operations of multiple robots in complex environments, improves the overall operating efficiency, safety and flexibility, and thus significantly improves the overall operating efficiency and safety of the multi-robot system.

[0010] The multi-objective collaborative optimization intelligent multi-robot handling and scheduling system of the present invention comprises:

[0011] (1) Multi-robot task assignment module, used to select the optimal task assignment scheme based on the comprehensive evaluation strategy;

[0012] (2) Single robot walking path optimal algorithm module, which uses the A* algorithm to search for the optimal walking path for each robot on the grid map;

[0013] (3) Multi-robot traffic control module, which realizes anti-collision control among multiple robots through the locking mechanism;

[0014] (4) The multi-robot deadlock resolution module adopts random and polling timeout waiting methods combined with path recalculation strategy.

[0015] Preferably, the comprehensive evaluation strategy is an evaluation strategy based on task priority, robot load capacity, handling distance and time window.

[0016] Preferably, the single robot walking path optimal algorithm module considers obstacles, robot size and turning cost on the grid map to achieve accurate path planning and obstacle avoidance.

[0017] Preferably, the multi-robot traffic control module ensures efficient and safe passage of multiple robots at intersections by dynamically allocating intersection use rights and priorities.

[0018] Preferably, when a deadlock situation is detected, the multi-robot deadlock resolution module automatically triggers a path recalculation mechanism by using a random and polling timeout waiting method to replan a conflict-free path for the affected robots to resolve the deadlock state.

[0019] Preferably, it also includes a system monitoring and scheduling optimization module to regularly evaluate the operating efficiency and security of the system and provide performance reports for user reference and optimization.

[0020] The scheduling method of a multi-objective collaboratively optimized intelligent multi-robot handling scheduling system of the present invention comprises:

[0021] (1) Multi-robot task assignment algorithm, used to select the optimal task assignment scheme based on comprehensive evaluation strategy;

[0022] (2) The optimal algorithm for the walking path of a single robot uses the A* algorithm to search for the optimal walking path for each robot on the grid map;

[0023] (3) Multi-robot traffic control algorithm, which realizes anti-collision control among multiple robots through the locking mechanism;

[0024] (4) Multi-robot deadlock resolution algorithm, which adopts random and polling timeout waiting methods combined with path recalculation strategy.

[0025] Preferably, (2) in the application scenario of multi-AGV path planning in a warehousing environment, Manhattan distance is used to match packages with AGVs, and path planning is performed by calculating the total time of moving packages as the objective function.

[0026] That is, the path planning of AGV needs to decide the path of AGV and whether to stop. Its strategic goal is to minimize the total time Cmax of transporting packages.

[0027] Cmax is the sum of the handling time and waiting time of all packages, as follows: Cmax = ∑Jj = 1(wj + tj)

[0028] Where wj is the waiting time of the parcel placed on the sorting table, and tj is the time it takes for the parcel to be transported by the AGV to the designated cargo buffer area. Because the speed v of the AGV is constant, tj = dtranj × v, and dtranj is the Manhattan distance between the starting coordinates pstartj and the end coordinates pendj of the parcel.

[0029] Preferably, (2) considers the application under specific map rules, that is, the grids of the map in the warehousing scenario have special rules, and the cargo loading and unloading time and speed limit factors in different areas need to be considered when the AGV passes by. The calculation of the G value of the actual moving cost from the starting point to the current point is defined according to the special warehousing rules; and the application combined with the warehousing task priority, that is, in the warehousing system, tasks usually have priorities, and the task priority is integrated into the path planning of the A algorithm.

[0030] Preferably, (3) is a two-step locking method and a locking conflict resolution strategy based on priority negotiation, wherein the two-step locking method, namely

[0031] a Each car moves one step and locks the points of the two steps ahead;

[0032] When car b is locking a grid and finds that the point on the map to be locked has been locked, the two cars negotiate to see which one has a higher priority, which car goes first, and the other car stops and waits;

[0033] c The car will unlock after it passes by, and can re-lock the point it is about to travel while waiting, and continue to move forward;

[0034] The d loop keeps locking each step.

[0035] The priority negotiation-based locking conflict resolution strategy is that when a car finds that the point to be locked is already locked when locking the grid, the two cars negotiate to decide which car goes first based on the priority.

[0036] The multi-robot handling scheduling system and method based on multi-objective optimization proposed in the present invention have broad application prospects. In the fields of automated warehouses, production lines, logistics distribution, etc., the system and method can significantly improve production efficiency, reduce costs, and improve safety, bringing significant economic and social benefits to enterprises. With the in-depth development of intelligent manufacturing and Industry 4.0, the market demand for this system will continue to grow, and it has broad market prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of the intelligent multi-robot handling scheduling of multi-objective collaborative optimization described in the present invention.

[0038] Figure 2 It is a relationship diagram of the components of the intelligent multi-robot handling scheduling module for multi-objective collaborative optimization described in the present invention. DETAILED DESCRIPTION

[0039] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] Figure 1 It is a flow chart of the intelligent multi-robot handling scheduling of multi-objective collaborative optimization described in the present invention.

[0041] Figure 2 It is a relationship diagram of the components of the intelligent multi-robot handling scheduling module for multi-objective collaborative optimization described in the present invention.

[0042] The multi-objective collaborative optimization intelligent multi-robot handling and scheduling system of the present invention comprises:

[0043] (1) Multi-robot task assignment module, used to select the optimal task assignment scheme based on the comprehensive evaluation strategy;

[0044] (2) Single robot walking path optimal algorithm module, which uses the A* algorithm to search for the optimal walking path for each robot on the grid map;

[0045] (3) Multi-robot traffic control module, which realizes anti-collision control among multiple robots through the locking mechanism;

[0046] (4) Multi-robot deadlock resolution module, which uses random and polling timeout waiting methods combined with path recalculation strategy;

[0047] Specifically, it includes:

[0048] 1. Multi-robot task assignment

[0049] It is used to select the optimal task assignment scheme according to the comprehensive evaluation strategy, where the "comprehensive evaluation strategy" refers to different methods and rules for determining how to assign tasks to robots. That is, the comprehensive evaluation strategy is a comprehensive evaluation strategy based on task priority, robot load capacity, handling distance and time window.

[0050] System module function: Comprehensively analyze multiple factors such as task priority, robot load capacity, handling distance and time window, and assign tasks to the most suitable robot according to the decision weight defined within the system to optimize the overall scheduling efficiency.

[0051] 2. Optimal planning of single robot walking path

[0052] The single robot walking path optimal algorithm module considers obstacles, robot size and turning cost on the grid map to achieve accurate path planning and obstacle avoidance.

[0053] System module function: With A* algorithm as the core, combined with the grid map environment, comprehensive consideration of the robot's kinematic characteristics, obstacle distribution and environmental constraints, real-time calculation and output of the shortest or optimal path from the starting point to the end point.

[0054] 3. Multi-robot traffic collaborative control

[0055] The multi-robot traffic control module ensures efficient and safe passage of multiple robots at intersections by dynamically allocating intersection use rights and priorities.

[0056] System module functions: Implement strategies such as locking mechanisms to effectively avoid collisions between robots and traffic jams, ensure the smooth operation of the multi-robot system, and have adaptive capabilities to dynamically adjust traffic control strategies according to environmental changes and task requirements, thereby improving system flexibility and response speed.

[0057] 4.Solution to multi-robot deadlock problem / intelligent solution to deadlock problem

[0058] The multi-robot deadlock resolution module automatically triggers a path recalculation mechanism when a deadlock situation is detected, and replans a conflict-free path for the affected robots to resolve the deadlock state.

[0059] System module functions: Integrate strategies such as random waiting, polling timeout and path recalculation to monitor and solve deadlock problems in multi-robot systems in real time. Once a deadlock is detected, it can quickly identify and locate the deadlock area, and quickly restore the normal operation of the system by adjusting the robot path, priority or taking other intervention measures.

[0060] The multi-objective collaborative optimization intelligent multi-robot handling and scheduling system described in the present invention also includes a system monitoring and scheduling optimization module, which regularly evaluates the operating efficiency and safety of the system and provides performance reports for user reference and optimization.

[0061] The multi-objective collaborative optimization intelligent multi-robot handling scheduling method of the present invention comprises:

[0062] (1) Multi-robot task assignment algorithm steps, used to select the optimal task assignment scheme based on multiple strategies;

[0063] Method and steps: Use this module to clarify task attributes and dynamically evaluate task priorities to ensure that key tasks can be scheduled for execution first. Use this module to store tasks to be scheduled, monitor task execution and warehouse status in real time, and provide real-time basis for subsequent scheduling decisions.

[0064] (2) The optimal algorithm steps for a single robot’s walking path are to use the A* algorithm to search for the optimal walking path for each robot on the grid map;

[0065] Method and steps: This module takes into account multiple relevant factors and intelligently assigns tasks to the most suitable robot.

[0066] (3) Multi-robot traffic control algorithm steps, realizing anti-collision control among multiple robots through the locking mechanism;

[0067] This module is used to plan the optimal path for a single robot on a grid map based on the A* algorithm.

[0068] (4) The steps of the multi-robot deadlock resolution algorithm use random and polling timeout waiting methods combined with path recalculation strategies;

[0069] Method and steps: Use this module to monitor and handle deadlock problems in real time, and quickly restore the system to normal operation by adjusting the robot path, priority, etc.

[0070] Embodiment 1:

[0071] The multi-objective collaborative optimization intelligent multi-robot handling and scheduling system of the present invention comprises:

[0072] Multi-robot task assignment module: This module is a key component of the system. As the brain of the system, it is responsible for receiving task requests from upper-level management systems (such as MES, WMS, etc.), and intelligently assigning tasks to the most suitable robots based on a series of complex and sophisticated considerations. Task priority, robot load capacity, and inter-task dependencies are the three major factors in this module when assigning tasks. Task priority ensures that urgent and important tasks are given priority; robot load capacity avoids robot overload or idleness caused by improper task assignment by comprehensively considering the robot's current load, maximum load, and load balance; task dependency analysis ensures that tasks are executed in the correct order, maintaining the logic and consistency of system operation. By comprehensively considering the three factors, this module can accurately match tasks and robots, optimize task allocation, and thus improve the operating efficiency and response speed of the entire system.

[0073] Single Robot Walking Path Optimization Module: This module is responsible for planning the optimal path from the starting point to the end point for a single robot. To achieve this goal, the module uses the A* algorithm as the core of path planning. This algorithm can search for the shortest path on a grid map or other types of maps while taking into account the robot's kinematic characteristics and environmental constraints. When planning the path, the module will also update the map information in real time to reflect changes in the environment (such as new obstacles). In this way, the module can ensure that accurate and reliable path planning results are provided to the robot.

[0074] Multi-robot traffic control module: This module is the core component to ensure the smooth operation of the multi-robot system. This module monitors the interaction between robots in real time and adopts multiple strategies to effectively coordinate the movement of robots, thereby preventing collisions and traffic jams. Among them, the lock grid mechanism sets a virtual lock on the map to ensure that only one robot can enter a specific area at the same time, greatly reducing the risk of collision. At the same time, the speed control strategy dynamically adjusts the driving speed of the robot according to the real-time position and speed of the robot to ensure that the robots maintain a safe distance and avoid congestion. In addition, when a potential conflict is detected, the path adjustment strategy will quickly intervene to re-plan the path for the robot to ensure the overall smooth operation of the system. Through these sophisticated strategies, the multi-robot traffic control module successfully achieves precise control of robot movement, providing a strong guarantee for the efficient and stable operation of the system.

[0075] Multi-robot deadlock problem solving module: This module is responsible for monitoring and solving deadlock problems in multi-robot systems. Deadlock refers to a situation where multiple robots cannot continue to perform tasks because they are waiting for each other. To solve this problem, this module uses multiple strategies to break the deadlock. The random waiting strategy is to let some robots wait randomly for a period of time when a deadlock is detected, and then retry to perform the task. The priority adjustment strategy dynamically adjusts the priority of the robot according to the priority of the task or the status of the robot to break the deadlock. In addition, the path recalculation strategy replans the path for the deadlocked robot to avoid conflicts with other robots. Through these strategies, this module can effectively solve the deadlock problem in the multi-robot system and ensure the stability and reliability of the system.

[0076] The multi-objective collaborative optimization intelligent multi-robot handling scheduling method of the present invention comprises:

[0077] Multi-robot task assignment algorithm: This algorithm supports multiple strategy options to adapt to different business scenarios and actual needs. By comprehensively considering multiple factors such as task priority, robot load capacity, and inter-task dependencies, the algorithm can intelligently assign tasks to the most suitable robot to ensure balanced and efficient task allocation;

[0078] Optimal algorithm for single robot walking path: A* algorithm is used as the core of path planning, which realizes the optimal path search on the grid map. By comprehensively considering the robot's kinematic characteristics, obstacle distribution and environmental constraints, the algorithm can plan the shortest or optimal path from the starting point to the end point in real time, reducing the robot's walking distance and time, and improving the handling efficiency;

[0079] Multi-robot traffic control (anti-collision) algorithm: The traffic control algorithm effectively avoids collisions between robots and traffic congestion through strategies such as the locking mechanism. The algorithm monitors the position and speed of the robots in real time, predicts and avoids potential traffic conflicts, and ensures the safety of vehicles passing at intersections. At the same time, the algorithm also has adaptive capabilities and can dynamically adjust traffic control strategies according to environmental changes and task requirements;

[0080] Multi-robot deadlock solution algorithm: The deadlock solution strategy uses random and polling timeout waiting methods, combined with path recalculation and bypassing congested areas, to effectively solve the deadlock problem in the multi-robot system. When a deadlock is detected, the algorithm can quickly identify and locate the deadlock area, and quickly restore the normal operation of the system by adjusting the robot path or priority.

[0081] Figure 1This is the multi-objective collaborative optimization intelligent multi-robot handling scheduling flow chart of the present invention, which shows the key steps of the task scheduling algorithm in the system, from task initialization to execution, and the processing flow after task completion. The entire flow chart indicates the direction of information flow and control flow through arrows, and clearly depicts the interaction and data exchange process between components.

[0082] Environmental Modeling:

[0083] In the application scenario of multi-AGV path planning in a warehousing environment, the complete strategy combination of using Manhattan distance for matching packages with AGVs and calculating the total time of moving packages as the objective function for path planning is an original invention customized according to the characteristics of the warehousing multi-AGV system. The details are as follows:

[0084] set up and Represent the AGVs, packages and cargo buffers in the warehouse respectively. The grid coordinates pstartj of the automatic sorter corresponding to package j and the corresponding cargo buffer coordinates pendj are known, and the coordinates p∈R2 are determined by the Cartesian coordinate system. The loading and unloading of AGVs are represented by ω, that is, let ωi represent the loading state of the i-th AGV:

[0085] When the automatic sorting machine distributes a new package, the task will be assigned based on the Manhattan distance between the package and the surrounding empty AGVs. The Manhattan distance between AGVi and package j is given as: dij = |xi-xj|+|yi-yj|

[0086] Where, there are coordinates p = (x, y), x represents the horizontal coordinate of the grid; y represents the vertical coordinate of the grid. The Manhattan distance between package j and all empty AGVs is given by: mini∈I(dij=|xi-xj|+|yi-yj|),ifωi=0

[0087] AGV path planning requires deciding the path of the AGV and whether to stop. Its strategic goal is to minimize the total time Cmax for transporting packages. Cmax is the sum of the transport time and waiting time of all packages, as follows: Cmax = ∑Jj = 1(wj + tj)

[0088] Where wj is the waiting time of the parcel placed on the sorting table, and tj is the time it takes for the parcel to be transported by the AGV to the designated cargo buffer area. Because the speed v of the AGV is constant, tj = dtranj × v, and dtranj is the Manhattan distance between the starting coordinates pstartj and the end coordinates pendj of the parcel.

[0089] Scheduling algorithm initialization:

[0090] In the embodiment, the scheduling algorithm is first initialized, including setting the task structure, initializing the task queue, the robot equipment list, and the library storage location. The definition of the task structure includes task attributes such as task name, priority, and estimated execution time, the robot equipment attributes include robot movement speed, rotation speed, loading / unloading / processing time, etc., and the storage location structure definition includes parameters such as shelf number, storage location code, storage product ID, maximum storage capacity, initial storage capacity, safety stock, and replenishment quantity.

[0091] Data collection module:

[0092] Next, start the data collection module, which is responsible for collecting order information, storage location information, warehouse floor plan configuration, AGV robot configuration and other information from the blueprint system.

[0093] Task generation and queuing:

[0094] The system automatically generates tasks based on the collected information and adds them to the dynamic task queue according to the preset task structure. This queue is a container for storing all tasks to be scheduled, and each element contains detailed parameters and current status information of the task. When a task is added to the queue, the system will assign it an initial priority according to a specific formula. The priority Ptask calculation formula is:

[0095] Ptask=w1·Utask+w2·Itask+w3·Dtask

[0096] Among them, w1, w2, w3 are weight coefficients; Utask is the task urgency index; Itask is the task importance index; and Dtask is the delivery deadline of the task.

[0097] Weight coefficient setting: Analyze business priorities. If you pay more attention to on-time delivery, you can increase w3; if you pay more attention to emergency task processing, increase w1; if the focus is on completing key tasks, increase w2. You can also combine historical data simulation to adjust the weight according to the performance of task execution.

[0098] Task urgency index (Utask) setting: Consider the task time window. Tasks approaching the processing time have high urgency. Combined with inventory and order status, replenishment tasks with tight inventory have high urgency. Tasks involving production or distribution bottlenecks also have high urgency.

[0099] Task importance index (Itask) setting: according to the value of the goods, high-value goods tasks are of high importance. Based on customer priority, VIP customer tasks are of high importance. Tasks that have a significant impact on subsequent processes are also of high importance.

[0100] Task screening and execution:

[0101] The system first carefully screens the task queue based on the task start conditions and product information to determine which tasks meet the conditions for immediate execution. Then, the system selects the one with the highest priority from the screened tasks for execution. The priority calculation here partly depends on the task urgency indicator Utask.

[0102] The task urgency indicator Utask is defined as follows: it is equal to the inverse of the difference between the task delivery deadline and the current time. The specific formula is:

[0103] Utask=1 / (Dtask-Tcurrent)

[0104] Among them, if the task delivery deadline Dtask is earlier than the current time Tcurrent, Dtask will be assigned a very large value to clearly indicate that the task is overdue.

[0105] Real-time monitoring and feedback:

[0106] During the entire process of task execution, the system implements a strict monitoring strategy to track and record the execution status and feedback information of the task in real time. This information is crucial for accurately evaluating the progress of the task and resource consumption, and can be intuitively displayed on the corresponding interface after the user configures the visualization panel.

[0107] Regarding the monitoring details of task execution, the system mainly focuses on the following core parameters:

[0108] Actual start time (S_start): This parameter marks the specific time point when the task is officially started and executed.

[0109] Actual end time (S_end): This parameter records the specific moment when the task is successfully completed.

[0110] Resource consumption (R_consume): This indicator reflects in detail the total amount of various resources consumed during task execution.

[0111] Task completion judgment:

[0112] When the task execution reaches the end point, the system will immediately start the completion check mechanism to determine whether the task has successfully achieved the expected goal. If the task is judged to be completed, the system will quickly remove the task from the task queue and automatically advance it to the subsequent work process. The calculation formula for the task completion indicator Ctask is:

[0113] Ctask=Atotal / Acompleted

[0114] Among them, Acompleted is the amount of work that has been completed, and Atotal is the total amount of work set for the task.

[0115] Priority adjustment and rotation:

[0116] If the task is not completed or there is a new task, the system will dynamically adjust the priority of the task based on the current execution status and feedback information. After that, the system returns to the task screening stage and continues to execute the next executable task. For example, when executing the previous order, if the selected shelf goods do not meet the task order requirements, it is necessary to send a robot to screen the shelves again to execute the task. At this time, the priority of the task will be temporarily adjusted to adapt to the newly added task.

[0117] New task processing:

[0118] During the cycle, if a new order or task appears, the system will add the new task to the queue and re-prioritize it according to the current task execution status and resource status.

[0119] AGV Scheduling Component: This algorithm uses advanced scheduling algorithms and path planning technology to achieve efficient scheduling and safe operation of AGVs. The specific functions of the scheduler module include task reception, status monitoring, scheduling algorithm execution, path planning, and task allocation. The following is a detailed technical solution:

[0120] Task receiving module:

[0121] Function description: Receive and process task requests from clients and store task information in the database. Task information includes task type, priority, starting location, and target location.

[0122] Specific implementation: Use API interface or message queue to receive and process task requests to ensure timely and accurate storage of task information.

[0123] Status monitoring module:

[0124] Function description: Real-time monitoring of all AGV status information, including current location, speed, power, task completion status, etc., and update this information to the database.

[0125] Specific implementation: Use timed polling or event-driven methods to ensure real-time update of status information. Status information is stored and managed through memory data, and the operation results are stored locally after the operation is completed.

[0126] Scheduling algorithm module:

[0127] Function description: According to the task information and AGV status, the optimization scheduling algorithm is executed to generate the optimal scheduling plan.

[0128] Steps: Generate an initial scheduling plan, evaluate the pros and cons of the plan according to the fitness function, select the best plan for crossover and mutation operations, and generate a new generation of plans. After multiple generations of iterations, the optimal scheduling plan is obtained.

[0129] Fitness function: Considers factors such as task priority, AGV driving distance, and task completion time to ensure the efficiency of the scheduling plan.

[0130] Let the fitness function be F(S), where S represents the scheduling scheme.

[0131] F(S)=w1*P(S)+w2*D(S)+w3*T(S), where w1, w2, and w3 are weight coefficients used to balance the importance of different factors.

[0132] P(S) is the part related to task priority. It is the weighted average of all task priorities or some quantitative value after sorting by task priority.

[0133] D(S) is the part related to the AGV travel distance. The sum of the travel distances of all AGVs when executing the scheduling plan S can be calculated.

[0134] T(S) is the part related to task completion time. It is the maximum value of all task completion times or the average value of task completion times.

[0135] Path planning module:

[0136] Function description: According to the scheduling plan, use the A* algorithm and dynamic window method (DWA) to plan the optimal path for each AGV to ensure that the AGV completes the task efficiently and safely.

[0137] A* Algorithm:

[0138] Steps: Based on heuristic search, calculate the shortest path from the starting point to the end point, consider factors such as path length and road conditions, and generate a conflict-free optimal path.

[0139] Heuristic function: Combined with Manhattan distance or Euclidean distance, it estimates the cost from the node to the target node and guides the search process. The A* algorithm is similar to finding the optimal tree in graph theory. It usually considers the path cost of selecting a certain path comprehensively. Among all the passable paths, there is always a path that has the least cost compared to any other passable path. When searching for the best path in graph theory, the cost of each path is known and fixed, but when using the A* algorithm to solve the best path, moving along different paths, although the cost of the same node may be different, this is the essence of heuristic pathfinding.

[0140] When using this method, we first abstract the actual problem and use a matrix to represent the elements in the problem, including the starting point, the target point, and the obstacles. We will gradually find that in pathfinding, the actual problem is abstractly represented by a matrix, and then the pathfinding process is realized by simulating the operation of the matrix.

[0141] The basic idea is to take the starting point as the center, and point the four adjacent points around it through pointers. Then select the best path point among the surrounding points and use it as the center point to point the surrounding points that have not yet established pointer connections (it is feasible, which will be explained later) through pointers. Then select the best path point and use this point as the center to find the path until the target point is found among the surrounding points of the point. In this way, the path being searched is connected one by one through pointers, and finally these points are output as the path being searched.

[0142] The following mainly analyzes the path-finding process of the A* algorithm step by step through the following aspects:

[0143] The entire pathfinding process is summarized as follows:

[0144] The starting cell (row 3, column 3 in the table) is added to the open list;

[0145] Repeat the following steps:

[0146] a. Find the point in the open list with the lowest F value, where the F value is the priority of the current node in the path search, which we call the current point (row 4, column 3 in the table);

[0147] b. Add it to the closed list;

[0148] c. For each point in the 4 adjacent grids

[0149] If it is not passable or already in the closed list, skip it. Otherwise:

[0150] If it is not in the open list, add it. Set the current point as the parent node of this grid. At the same time, record the F, G, and H values ​​of the point. The calculation formula is F = G + H, where the F value is the priority of the current node in the path search, the G value is the actual moving cost from the starting point to the current point, and the H value is the heuristic estimated cost from the current point to the target point;

[0151] If it is already in the open list, use the G value as a reference to check whether the new path is better. A lower G value means a better path. If so, change the parent node of this point to the current point and recalculate the G and F values ​​of this point. After the change, you need to re-sort the open list by F value. If not, there is no need to do the subsequent work of changing the pointer and recalculating the G and F values;

[0152] Stop searching in two cases:

[0153] When the target point is added to the closed list, the path is found and the search ends normally;

[0154] The target point is not found, but the open list is empty. No suitable path is found at this time, and the search ends;

[0155] Save the path. Start from the target point and move along the pointer of each point until you return to the starting point and output the path.

[0156]

[0157] The present invention is based on the core framework of the A* algorithm, and the original inventions are as follows:

[0158] (1) Application under specific map rules

[0159] If in a warehousing scenario, the grids of the map (representing warehouse areas) have special rules, such as some grids representing cargo storage areas, the AGV (automatic guided vehicle) needs to consider factors such as cargo loading and unloading time and speed limits in different areas when passing through. In this case, the calculation of the G value (the actual moving cost from the starting point to the current point) can be defined according to these special warehousing rules. For example, when the AGV passes through a grid that needs to load and unload goods, the calculation of the G value must add the time cost of loading and unloading goods (assuming that time can be converted into distance cost, etc.), which is an original invention for the application of the A* algorithm in warehousing scenarios.

[0160] (2) Application combined with warehouse task priority

[0161] In a warehousing system, tasks are usually prioritized. Task priorities can be incorporated into the path planning of the A algorithm. For example, when the task associated with the target point (which may be a cargo storage location or a cargo delivery location) has a higher priority, when calculating the H value (the heuristic estimated cost from the current point to the target point), the path exploration of these high-priority targets can be prioritized through certain weight adjustments. Or when selecting the point with the lowest F value in the opening list, a certain preference is given to nodes related to high-priority tasks, so that the A* algorithm can better serve task scheduling in warehousing scenarios.

[0162] 4.2 Traffic control based on locking mechanism

[0163] After the vehicle road planning is completed, multiple vehicles start to move at the same time. It is inevitable that vehicles will meet on multiple roads. When meeting, the vehicles will mainly follow, go in opposite directions, and meet at crossroads or T-junctions. When following, there will be sensors in front of the vehicle to avoid following collisions. In order to avoid collisions at crossroads and T-junctions, the present invention has created a two-step locking method and a locking conflict resolution strategy based on priority negotiation. That is:

[0164] (1) Each time the car moves one step, it locks the points of the two steps ahead;

[0165] (2) When a car is locking a grid and finds that the point on the map it is about to lock has already been locked, the two cars negotiate to see which one has a higher priority and which one goes first while the other stops and waits;

[0166] (3) The car will unlock after it passes by, and can relock the point it is about to travel while waiting, and continue to move forward;

[0167] (4) The loop continues to lock the grid at each step.

[0168] The two-step locking method is to more effectively avoid collisions at complex intersections and T-shaped roads, and is the originality of the present invention in terms of the locking mechanism.

[0169] Locking grid conflict resolution strategy based on priority negotiation: When a car finds that the point it is about to lock has been locked when locking the grid, the two cars negotiate to decide which car goes first based on priority. In general traffic control scenarios, it may rely more on fixed traffic rules (such as first come first served, priority in a specific direction, etc.), and the present invention can better adapt to the flexible operation requirements of intelligent vehicles (such as AGV) in complex environments through priority negotiation, thereby improving overall traffic efficiency.

[0170] Dynamic Window Approach (DWA):

[0171] Steps: According to the AGV's motion model and real-time environmental information, dynamically adjust the driving path to avoid obstacles.

[0172] Motion model: Consider the speed, acceleration and steering angle of the AGV, calculate the safe path area, and select the optimal path within the area. When planning the current vehicle path, the shortest path should be planned taking into account road congestion to achieve the best overall planning result. Use the A* algorithm and use the G value as a reference to check whether the new path is better. Increase the G value of the points on the paths planned by other vehicles in the map to avoid searching for the same path as much as possible. Similarly, when the vehicle needs to turn, increase the G value of the next point of the turn, so as to plan the path to avoid turning as much as possible, so as to achieve the highest overall efficiency and the best global path.

[0173] In addition, since the path planning is static, each vehicle needs to be locked during the movement to ensure that the vehicles do not collide.

[0174] Path optimization combined with A* algorithm: In the dynamic window method, the strategy of avoiding vehicles from searching the same path and reducing turning situations by using the A algorithm and adjusting the G value according to the actual situation is the original point of the present invention. In the traditional application of the dynamic window method, there may be few ways to combine the G value of the A algorithm with the avoidance of congestion and reduction of turning situations in vehicle path planning. By increasing the G value of other car planning path points in the map, the vehicle can actively avoid occupied or congested paths during path planning, thereby achieving the best overall planning result. This is an innovative application of the dynamic window method in the intelligent vehicle path planning scenario.

[0175] Collaborative application with the lock grid mechanism: Emphasis is placed on combining dynamic path planning with lock grid traffic control to ensure that vehicles do not collide. This close combination of dynamic path planning and traffic control is the originality of the present invention. General path planning methods may only focus on the optimality of the path, while ignoring the impact of traffic control on path implementation during actual operation. The present invention, through this collaborative application, enables vehicles to travel more safely and efficiently in complex environments.

[0176] Advantages and features of the present invention:

[0177] 1. Multi-objective optimization: Comprehensively consider multiple optimization goals such as task completion time, cost, and safety to achieve comprehensive performance improvement of the multi-robot system.

[0178] 2. Intelligent scheduling: By integrating a variety of advanced algorithms and technologies, intelligent scheduling of task assignment, path planning, traffic control and deadlock resolution strategies can be achieved.

[0179] 3. Efficient operation: reduce the robot's walking distance and time, improve handling efficiency, and reduce energy consumption costs.

[0180] 4. Safety assurance: effectively avoid collisions between robots and traffic jams, and ensure the safety of handling operations.

[0181] 5. Flexible adaptation: supports a variety of business scenarios and changes in demand, with good scalability and customizability.

[0182] Therefore, the present invention solves the following problems:

[0183] 1. Support multiple task assignment strategies to adapt to different business scenarios and changing needs;

[0184] 2. Use advanced path planning algorithms to achieve the optimal path search for robots in complex environments;

[0185] 3. Establish an effective traffic control mechanism to avoid collisions between robots and traffic congestion;

[0186] 4. Provide flexible deadlock resolution strategies to ensure stable operation of the system under high load and complex environments.

Claims

1. An intelligent multi-robot handling and scheduling system with multi-objective collaborative optimization, characterized in that: include: (1) Multi-robot task assignment module, used to select the optimal task assignment scheme based on the comprehensive evaluation strategy; (2) Single robot walking path optimal algorithm module, which uses the A* algorithm to search for the optimal walking path for each robot on the grid map; (3) Multi-robot traffic control module, which realizes anti-collision control among multiple robots through the locking mechanism; (4) The multi-robot deadlock resolution module adopts random and polling timeout waiting methods combined with path recalculation strategy.

2. The intelligent multi-robot handling and dispatching system according to claim 1 is characterized in that: The comprehensive evaluation strategy is an evaluation strategy based on task priority, robot load capacity, handling distance and time window.

3. The intelligent multi-robot handling and dispatching system according to claim 1 is characterized in that: The single robot walking path optimal algorithm module considers obstacles, robot size and turning cost on the grid map to achieve accurate path planning and obstacle avoidance.

4. The intelligent multi-robot handling and dispatching system according to claim 1, characterized in that: The multi-robot traffic control module ensures efficient and safe passage of multiple robots at intersections by dynamically allocating intersection use rights and priorities.

5. The intelligent multi-robot handling and dispatching system according to claim 1, characterized in that: The multi-robot deadlock resolution module, when a deadlock situation is detected, automatically triggers a path recalculation mechanism by using a random and polling timeout waiting method, and replans a conflict-free path for the affected robots to resolve the deadlock state.

6. The intelligent multi-robot handling and dispatching system according to claim 1, characterized in that: It also includes a system monitoring and scheduling optimization module, which regularly evaluates the system's operating efficiency and security, and provides performance reports for users' reference and optimization.

7. The scheduling method of a multi-objective collaborative optimization intelligent multi-robot handling scheduling system according to any one of claims 1 to 6, characterized in that: include: (1) Multi-robot task assignment algorithm, used to select the optimal task assignment scheme based on comprehensive evaluation strategy; (2) The optimal algorithm for the walking path of a single robot uses the A* algorithm to search for the optimal walking path for each robot on the grid map; (3) Multi-robot traffic control algorithm, which realizes anti-collision control among multiple robots through the locking mechanism; (4) Multi-robot deadlock resolution algorithm, which adopts random and polling timeout waiting methods combined with path recalculation strategy.

8. The method of a multi-objective collaborative optimization intelligent multi-robot handling scheduling system according to claim 7, characterized in that: Among them, (2) in the multi-AGV path planning application scenario in the warehouse environment, Manhattan distance is used to match packages with AGVs, and the path planning is performed by calculating the total time of moving packages as the objective function. That is, the path planning of AGV needs to decide the path of AGV and whether to stop. Its strategic goal is to minimize the total time Cmax of transporting packages. Cmax is the sum of the handling time and waiting time of all packages, as follows: Cmax = ∑Jj = 1(wj + tj) Where wj is the waiting time of the parcel placed on the sorting table, and tj is the time it takes for the parcel to be transported by the AGV to the designated cargo buffer area. Because the speed v of the AGV is constant, tj = dtranj × v, and dtranj is the Manhattan distance between the starting coordinates pstartj and the end coordinates pendj of the parcel.

9. The method of a multi-objective collaborative optimization intelligent multi-robot handling scheduling system according to claim 7, characterized in that: Among them, (2) considers the application under specific map rules, that is, the grids of the map in the warehousing scenario have special rules. When the AGV passes, it is necessary to consider the cargo loading and unloading time and the speed limit factors in different areas. The calculation of the G value of the actual moving cost from the starting point to the current point is defined according to the special warehousing rules; and the application combined with the warehousing task priority, that is, in the warehousing system, tasks usually have priorities, and the task priority is integrated into the path planning of the A algorithm.

10. The method of a multi-objective collaborative optimization intelligent multi-robot handling scheduling system according to claim 7, characterized in that: Among them, (3) is a two-step locking method and a locking conflict resolution strategy based on priority negotiation. The two-step locking method, namely, a Each car moves one step and locks the points of the two steps ahead; When car b is locking a grid and finds that the point on the map to be locked has been locked, the two cars negotiate to see which one has a higher priority, which car goes first, and the other car stops and waits; c The car will unlock after it passes by, and can re-lock the point it is about to travel while waiting, and continue to move forward; The d loop keeps locking each step. The priority negotiation-based locking conflict resolution strategy is that when a car finds that the point to be locked is already locked when locking the grid, the two cars negotiate to decide which car goes first based on the priority.

Images