Multi-unmanned forklift scheduling control method comprising path planning and vehicle passing strategy

Abstract

This invention provides a multi-unmanned forklift scheduling and control method incorporating path planning and vehicle access strategies, comprising the following steps: S1: Constructing a warehouse map; S2: Obtaining a path set, corresponding path time, and congestion degree using an offline path planning algorithm; S3: Pulling a task set to be updated from a task pool; S4: Establishing and solving a task allocation model and a path planning model with the objective of minimizing the maximum task completion time; S5: Updating the vehicle task sequence and corresponding execution path; S6: Updating the real-time vehicle location, motion state, and unexecuted task paths; S7: Calculating the nearest consecutive shared path for each forklift as a potential deadlock path; S8: Determining the real-time control commands for each forklift based on the permission request strategy. This multi-unmanned forklift scheduling and control method, incorporating path planning and vehicle access strategies, comprehensively utilizes rule-based and model-based methods, and significantly improves system operating efficiency through carefully designed rules.

Description

Technical Field

[0001] This invention relates to the field of unmanned forklift scheduling and control technology, and in particular to a multi-unmanned forklift scheduling and control method that includes path planning and vehicle passage strategies. Background Technology

[0002] Unmanned forklifts are playing a crucial role in current warehousing and logistics. In actual forklift operation, warehouses often present complex conflict relationships due to narrow aisles, two-way traffic, and the large size of the unmanned forklifts. Vehicle speed fluctuations can also cause errors, requiring real-time task scheduling instructions and placing high demands on response time. This complexity of conflict relationships distinguishes unmanned forklift systems from other warehouse AGV systems. Currently, most multi-unmanned forklift scheduling methods are based on heuristic scheduling rules, such as "first-come, first-served" and full-time window reservations. However, when multiple vehicles are planned simultaneously, this can lead to more conflicts and negatively impact overall system efficiency.

[0003] "Unmanned forklifts," also known as "driverless forklifts" or "AGVs" (AGV stands for Automated Guided Vehicle), are intelligent industrial vehicle robots that integrate forklift technology and AGV technology. Compared to ordinary AGVs, they can also perform operations in high-bay warehouses. The application of unmanned forklifts can solve problems such as large material flow and high labor intensity in manual handling during industrial production and warehousing logistics operations. The scheduling of unmanned forklifts mainly refers to assigning different tasks to multiple forklifts, determining the execution paths of these tasks, and allowing them to perform the tasks accordingly. The control of unmanned forklifts mainly refers to deciding on the movement, steering, or pausing of the vehicles during task execution. Scheduling and control need to consider resource competition that exists when forklifts share some paths. For example, two forklifts may collide when traveling straight ahead or crossing intersections, or multiple forklifts may form loops in congested areas, causing deadlocks. Such resource competition can significantly increase the maximum task completion time in actual execution of conventional scheduling schemes, as forklifts need to wait to avoid obstacles.

[0004] Currently, there are no publicly available scheduling technologies for unmanned forklift systems. The most relevant technologies are mainly concentrated in the field of AGV scheduling and control, and existing AGV scheduling and control technologies are mainly divided into two categories: rule-based methods and model-based methods.

[0005] Rule-based scheduling and control methods: The most typical scheduling method is first-come, first-served, which matches the nearest AGV with the task to execute the task, and uses heuristic algorithms such as A* to plan the path for the AGV. Some patents combine scheduling and control, and use rule-based methods to make decisions on both. For example, during scheduling, AGVs are assigned to execute according to task priority, and during control, potential conflicts are calculated using time windows. See [Song Limei. A multi-AGV scheduling method applied to a flexible manufacturing workshop with bidirectional paths: 201910315819.X[P]. 2023-04-07].

[0006] Scheduling and control methods based on mathematical models: Mathematical programming models are used, for example, setting the objective function to minimize task completion time or other indicators of interest, setting decision variables to which AGV performs which task and the corresponding task path, and setting constraints to ensure that the vehicle can return to the starting point or go to the starting point of the next task after completing the previous task, etc. See the reference [Singh N, Dang QV, Akcay A, et al. Amatheuristic for AGV scheduling with battery constraints[J]. European Journal of Operational Research, 2022, 298(3):855-873]. Due to the high flexibility and high complexity of the problem, there are currently no mature model-based solutions for control methods.

[0007] Existing rule-based scheduling and control methods struggle to achieve optimal solutions globally, focusing only on local optimization. For instance, the "first-come, first-served" approach uses a greedy algorithm to match tasks with the nearest available vehicle, but this can lead to subsequent tasks being matched with worse vehicles, resulting in lower overall efficiency. Other rule-based methods, such as using time windows to calculate potential conflicts, struggle to handle errors caused by the uncertainty of vehicle operation, are generally inflexible, require significant computation time, and cannot achieve real-time vehicle response.

[0008] Existing scheduling and control methods based on mathematical models are overly complex, with computation time increasing exponentially with the number of tasks and AGVs. This makes them unsuitable for large-scale scenarios in real-world industrial applications. Furthermore, if task information changes or unexpected situations such as vehicle malfunctions occur, requiring recalculation of the scheduling scheme, a lengthy waiting period is necessary. Therefore, relying solely on online solutions to mathematical models cannot meet the response time requirements for real-time vehicle control. Summary of the Invention

[0009] To address the shortcomings of the existing technologies, this invention provides a multi-unmanned forklift scheduling and control method that incorporates path planning and vehicle access strategies. It considers task allocation, path planning, and vehicle access strategies simultaneously, comprehensively utilizes rule-based and model-based methods, and significantly improves system operating efficiency through carefully designed rules.

[0010] To achieve the above objectives, the present invention provides a multi-unmanned forklift scheduling and control method incorporating path planning and vehicle passage strategies, comprising the following steps:

[0011] S1: Build the warehouse map;

[0012] S2: Obtain the path set, corresponding path time, and congestion level using an offline path planning algorithm;

[0013] S3: Retrieve the set of tasks to be updated from the task pool based on the real-time task status of the forklift;

[0014] S4: Considering the congestion level of each path, establish and solve the task allocation model and path planning model with the goal of minimizing the maximum task completion time;

[0015] S5: Update the vehicle task sequence and corresponding execution path using the solution results;

[0016] S6: Update the vehicle's real-time location, motion status, and unexecuted task paths;

[0017] S7: Calculate the nearest continuous shared path CS for each forklift. i As a potential deadlock path;

[0018] S8: Decision-making on real-time control commands for each forklift based on permission requests and policy settings;

[0019] S9: Determine if all tasks have ended. If so, end the process; otherwise, continue to step S10.

[0020] S10: Determine if the task update interval has been reached. If yes, return to step S3; otherwise, return to step S6.

[0021] Preferably, in step S1:

[0022] The system reads map data and task data from the automated unmanned warehouse. The map data includes the location of the unmanned warehouse's docking area, assembly line unloading area, transit warehouse, storage area, cargo capacity information, unloading nodes, picking nodes, and path nodes accessible to forklifts. The task data includes the task's start node, task's end node, and task arrival time.

[0023] Preferably, in step S2:

[0024] The path time calculation takes into account the extra time that the forklift needs to accelerate and decelerate when turning; the congestion degree is calculated based on the average congestion value of each node on the path, and the node congestion value is defined by the number of times that node may appear in each forklift path.

[0025] Preferably, in step S3:

[0026] Establish the task pool and set a fixed time interval as the task update interval; pull the tasks to be planned from the task pool, match the corresponding forklifts for the tasks to be planned and plan the corresponding paths, and simultaneously allocate and update the warehouse location for the task endpoint nodes for online tasks arriving at the pipeline.

[0027] Preferably, in step S4:

[0028] In the process of establishing the task allocation model and the path planning model, the travel time of each path is multiplied by the congestion level of that path for calculation.

[0029] Preferably, in step S7:

[0030] The determination of the potential deadlock path includes the following steps:

[0031] S71: For a specific forklift i, traverse its nodes to obtain the shared path nodes S between the forklift and other vehicles. i :

[0032] S i ={a|a∈R i , a∈R j , j∈A / {i}}

[0033] Where R i Let S be the unexecuted path of forklift i, A be the set of all forklifts, and S be the unexecuted path of forklift i. i It is the shared path node of forklift i, where 'a' represents the path node, and R... j This is the unexecuted path of forklift j, where j represents any forklift other than forklift i;

[0034] S72: Define the starting point g of the nearest contiguous shared path. i :

[0035]

[0036] a i，k It is path R i Let r(a, b) represent whether path nodes a and b are adjacent, with 1 indicating adjacent and 0 indicating non-adjacent; a i，l It is path R i The l-th node;

[0037] S73: Define the nearest contiguous shared path (CS) i As the potential deadlock path:

[0038]

[0039] Preferably, in step S8:

[0040] The permission request strategy includes two policy conditions:

[0041] First strategy condition: the next node on the forklift path and

[0042] Second strategy condition: the next node ai on the forklift path, 1∈CSi, and

[0043] Where s(i, a) represents whether node a is reserved by forklift i, with 1 for reserved and 0 for unreserved;

[0044] Forklift i can move forward and request permission to the next resource point if it meets either the first or second policy condition; otherwise, the forklift will remain at the current node.

[0045] Because the present invention adopts the above technical solution, it has the following beneficial effects:

[0046] In the field of unmanned forklift scheduling and control, a combined modeling and rule-based approach is used, integrating the advantages of both methods. The scheduling problem is transformed into offline solution of a set of alternative paths, yielding higher-quality alternative paths. Online route selection is achieved through mixed-integer programming, minimizing future conflicts in task scheduling and path planning. The control strategy incorporates potential deadlock path identification and node resource permission request strategies, enabling multiple forklifts to operate efficiently and avoid conflicts as much as possible. Offline computation yields higher-quality alternative paths, and online solution of the mixed-integer programming model quickly (around 10 seconds) provides high-quality task allocation and path planning results. With similar online decision-making times, the scheduling quality is significantly better than heuristic or rule-based scheduling schemes. Potential deadlock path identification allows for more flexible priority management of multiple vehicles. Compared to priority management based on time windows or other rules, it better addresses errors caused by uncertainties during vehicle operation and leads to higher traffic efficiency, significantly reducing both maximum task completion time and maximum forklift waiting time. Attached Figure Description

[0047] Figure 1 This is a flowchart of a multi-unmanned forklift scheduling and control method that includes path planning and vehicle passage strategies, according to an embodiment of the present invention. Detailed Implementation

[0048] The following is based on the attached diagram. Figure 1 The present invention provides preferred embodiments and describes them in detail to enable a better understanding of the functions and features of the present invention.

[0049] Please see Figure 1 An embodiment of the present invention provides a multi-unmanned forklift scheduling and control method including path planning and vehicle passage strategy, comprising the following steps:

[0050] S1: Build the warehouse map;

[0051] When constructing the warehouse map, the map data and task data of the automated unmanned warehouse are read. The map data includes the locations of the unmanned warehouse's docking areas, assembly line unloading areas, transit warehouses, and storage areas, information on the goods that can be accommodated, unloading nodes that forklifts can pass through, picking nodes, and path nodes. The task data includes the task's start node, task's end node, and task arrival time. The docking areas are mainly for unmanned forklifts to park and charge; the assembly line unloading areas receive goods input into the warehouse after production from the upstream assembly line; the transit warehouses are temporary single-layer storage areas for goods; and the storage areas are long-term storage rack areas for goods. Tasks can arrive at time 0 (i.e., they exist from the beginning) or be periodically arriving assembly line tasks.

[0052] S2: Obtain the path set, corresponding path time, and congestion level using an offline path planning algorithm;

[0053] The path time calculation considers the extra time required for the forklift to accelerate and decelerate during turns; the congestion degree is calculated based on the average congestion value of each node on the path, and the node congestion value is defined by the number of times that node may appear in each forklift path. In the offline path planning algorithm, to provide a decision range for solving the subsequent mixed integer programming model, the problem is extended to the problem of finding the first k shortest paths, using the algorithm proposed by Yen in 1971 [Yen J Y. Finding the kshortest loopless paths in a network[J].Management]. [Science, 1971, 17(11): 712-716.] The solution is obtained by using the deviation path algorithm in the recursive method. The algorithm can be divided into two parts. First, Dijkstra's algorithm is used to obtain the shortest path between nodes. Then, based on this, the other k-1 shortest paths are calculated in turn. All nodes in the shortest path except the terminal node are regarded as deviation nodes. The shortest path from each deviation node to the terminal node is calculated. Then, it is concatenated with the path from the starting node to the deviation node in the shortest path to form a new candidate path. The path time and congestion degree are calculated. The first k paths with shorter path times are retained. The size of k can be set according to the actual needs of the scenario. When k is small, the required storage space is small and the calculation time is fast. When k is large, the decision range is large and the solution quality may be better.

[0054] S3: Retrieve the set of tasks to be updated from the task pool based on the real-time task status of the forklift;

[0055] Some tasks exist from the initial moment, while others arrive during the process of a pipeline. Due to this difference in nature, this invention supports online task updates by establishing a task pool and setting a fixed time interval as the interval for task updates; pulling tasks to be planned from the task pool, matching the corresponding forklifts to the tasks to be planned and planning the corresponding paths; and simultaneously allocating and updating storage locations for online tasks arriving from the pipeline.

[0056] S4: Considering the congestion level of each path, establish and solve the task allocation model and path planning model with the goal of minimizing the maximum task completion time;

[0057] The task allocation and route planning scheme updates tasks online and establishes corresponding mathematical models. The goal of the model is to minimize the maximum completion time of the tasks, and in this process, the travel time of each path is multiplied by the congestion level of that path for calculation. The constraints mainly include that each task is executed by exactly one forklift through one path, ensuring that each forklift task sequence has a corresponding empty route for picking up goods and a loaded route for transporting goods, that the currently executing task does not change its corresponding forklift and path, and the definition of congestion level for each path, etc.

[0058] S5: Update the vehicle task sequence and corresponding execution path using the solution results;

[0059] Input map information, forklift task execution information, and task set information. Check if the pallet area corresponding to the forklift for the assembly line task is full. If there are no free nodes, the forklift will not be included in the task allocation object. Check the number of areas in the transfer warehouse where tasks have not been fully completed. To reduce forklift picking conflicts, the number of transfer warehouse task areas should match the number of forklifts. Since the assembly line task has not yet been assigned a termination node, its termination node is set to the first unoccupied pallet node in the corresponding pallet area during the solution process. Add task priority constraints to the task set to restrict the task retrieval order. Use an open-source solver to express and solve the mathematical model, and synchronize the obtained task allocation and path planning results to the corresponding forklifts for local execution.

[0060] S6: Update the vehicle's real-time location, motion status, and unexecuted task paths;

[0061] During vehicle operation, the vehicle's current position is read in real time to update errors caused by uncertainties in the vehicle's operation, thereby determining the vehicle's movement and updating the unexecuted paths.

[0062] S7: Calculate the nearest continuous shared path CS for each forklift. i As a potential deadlock path;

[0063] Identifying potential deadlock paths involves the following steps:

[0064] S71: For a specific forklift i, traverse its nodes to obtain the shared path nodes S between the forklift and other vehicles. i :

[0065] S i ={a|a∈R i , a∈R j , j∈A / {i}}

[0066] Where R i Let S be the unexecuted path of forklift i, A be the set of all forklifts, and S be the unexecuted path of forklift i. i It is the shared path node of forklift i, where 'a' represents the path node, and R... j It is the unexecuted path of forklift j, where j represents a forklift other than forklift i, i.e. a forklift that may conflict with forklift i.

[0067] The shared path node set represents the nodes on the assigned path that forklift i has not yet executed that conflict with other forklifts. The nodes in the potential deadlock path are a subset of this node set. If the shared path node set is empty, it means that forklift i has no nodes that conflict with other forklifts, i.e., there is no potential deadlock path.

[0068] S72: Define the starting point g of the nearest contiguous shared path. i :

[0069]

[0070] a i，k It is path R i Let r(a, b) represent whether path nodes a and b are adjacent, with 1 indicating adjacent and 0 indicating non-adjacent; a i，l It is path R i The l-th node;

[0071] The nearest continuous shared path must satisfy two conditions: first, the path length must be greater than or equal to two nodes, and these nodes must be adjacent, i.e., forming a continuous part of the vehicle path; second, if multiple continuous shared paths exist on the forklift path, the nearest continuous shared path uniquely points to the segment first traversed by the vehicle. Therefore, if the nearest continuous shared path exists, its starting point can be uniquely determined; otherwise, there is no potential deadlock path.

[0072] S73: Define the nearest contiguous shared path (CS) i As a potential deadlock path:

[0073]

[0074] S8: Decision-making on real-time control commands for each forklift based on permission requests and policy settings;

[0075] After uniquely identifying a potential deadlock path, define the corresponding permission request strategy for it. In the real-time control of the system, real-time commands need to be issued to control vehicle movement. The decision space at this point is either for the vehicle to proceed to a node (adjacent to the current node) or for the vehicle to pause at the current node. For a vehicle to proceed to a node, it needs to request permission for that node. Successful request requires the formulation of a corresponding strategy, which includes two policy conditions:

[0076] First strategy condition: the next node on the forklift path and

[0077] Second strategy condition: the next node a on the forklift path i，1 ∈CS i ,and

[0078] Where s(i, a) represents whether node a is reserved by forklift i, with 1 for reserved and 0 for unreserved;

[0079] Forklift i can move forward and request permission to the next resource point if it meets either the first or second policy condition; otherwise, the forklift will remain at the current node.

[0080] The first strategy condition applies when the forklift's next node is not on the most recent continuous shared path. In this case, the next node is not on the shared path, so the forklift will not conflict with other vehicles when it reaches the next node. The second strategy condition applies when the vehicle's next node is on the most recent continuous shared path. If no other vehicle has reserved the segment, it means that forklift i is the first forklift to enter the area and can move freely on the segment while always enjoying the highest priority. If other forklifts have reserved the segment (occupying a node or about to enter the segment), it means that forklift i is not the first forklift to enter the area and must wait for the forklifts on the current segment to leave before it can enter, thus ensuring the avoidance of deadlock on potentially deadlocked paths.

[0081] S9: Determine if all tasks have ended. If so, end the process; otherwise, continue to step S10.

[0082] S10: Determine if the task update interval has been reached. If yes, return to step S3; otherwise, return to step S6.

[0083] An embodiment of the present invention provides a multi-unmanned forklift scheduling and control method that includes path planning and vehicle passage strategies, mainly addressing the following issues:

[0084] Problem 1: Existing unmanned forklift scheduling methods use heuristics for task allocation and path planning, which can quickly obtain feasible solutions. However, the planned paths are usually not the shortest and can easily increase conflicts between forklifts. Therefore, it is necessary to propose an efficient algorithm that can take into account the reduction of future conflicts between vehicles and plan the shortest and optimal path under this condition, thereby improving the efficiency of logistics transportation within the warehouse.

[0085] Question 2: Forklifts encounter path conflicts between multiple vehicles during operation. Due to size limitations, two AGVs may collide or trigger anti-collision lasers if they are too close together. Furthermore, deadlocks may occur (multiple vehicles conflicting paths, causing them to be "locked" in place and unable to move). Conflicts between forklifts need to be resolved through delays or manual handling, which affects system efficiency. Therefore, it is necessary to consider vehicle conflict issues during operation and avoid conflicts by issuing control commands.

[0086] Question 3: Errors during vehicle movement can cause original control commands to fail, collisions to occur, and vehicles to be forced to wait, reducing system efficiency. These errors can be broadly categorized into vehicle speed fluctuations and unexpected events within the warehouse (such as personnel obstruction or vehicle communication failures). In actual control, dynamic adjustments are needed to adapt to these errors, and the solution must be robust enough to handle unforeseen circumstances.

[0087] This invention simultaneously considers task allocation, path planning, and vehicle traffic strategies, comprehensively utilizing rule-based and model-based methods, and significantly improving system efficiency through carefully designed rules. In the task allocation and path planning sections, a model-based approach is employed, combining offline computation and online decision-making to enhance the efficiency and quality of real-time decision-making. In the vehicle traffic strategy section, resource point reservation (to avoid collisions) and potential deadlock path identification (to avoid deadlocks) are used to prevent conflicts between vehicles. This method is compatible with both one-way and two-way traffic paths and effectively handles errors caused by uncertainties during vehicle operation.

[0088] The beneficial effects of this invention are as follows:

[0089] 1. Reduced maximum completion time of forklift system: As shown in Table 1, the method of the present invention significantly reduces the maximum completion time of all tasks compared with the rule-based heuristic method, and the improvement becomes more obvious as the number of tasks increases, thereby improving the system's operating efficiency.

[0090] Table 1: Results of Maximum Completion Time for Multiple Tasks

[0091]

[0092] (Unit: minutes)

[0093] 2. Reduced maximum waiting time for forklifts: As shown in Table 2, the method of this invention significantly reduces the maximum waiting time for forklifts compared to rule-based heuristic methods. It reduces congestion through reasonable path planning and eliminates some invalid waiting based on potential deadlock paths, showing good results with varying task volumes.

[0094] Table 2: Maximum Waiting Time for Forklifts for Various Task Quantities

[0095]

[0096] (Unit: minutes)

[0097] 3. Reduced number of forklift system collisions: As shown in Table 3, the method of the present invention significantly reduces the number of forklift collisions compared to rule-based heuristic methods.

[0098] Table 3 Results of Collision Counts for Multiple Tasks

[0099]

[0100] The present invention has been described in detail above with reference to the accompanying drawings and embodiments. Those skilled in the art can make various modifications to the present invention based on the above description. Therefore, certain details in the embodiments should not be construed as limiting the present invention, and the scope of protection of the present invention shall be defined by the appended claims.

Claims

1. A method for scheduling and controlling multiple unmanned forklifts, including path planning and vehicle access strategies, comprising the following steps: S1: Build the warehouse map; S2: Obtain the path set, corresponding path time, and congestion degree by the offline path planning algorithm. The path time calculation takes into account the extra time generated by the forklift needing to accelerate and decelerate when turning. The congestion level is calculated based on the average congestion value of each node on the path, and the node congestion value is defined by the number of times the node appears in each forklift path; S3: Retrieve the set of tasks to be updated from the task pool based on the real-time task status of the forklift; S4: Considering the congestion level of each path, establish and solve the task allocation model and path planning model with the goal of minimizing the maximum completion time of the task. In the process of establishing the task allocation model and the path planning model, the travel time of each path is multiplied by the congestion level of that path for conversion. S5: Update the vehicle task sequence and corresponding execution path using the solution results; S6: Update the vehicle's real-time location, motion status, and unexecuted task paths; S7: Calculate the nearest continuous shared path for each forklift. As a potential deadlock path; The determination of the potential deadlock path includes the following steps: S71: For a specific forklift i, traverse the nodes on the path to obtain the shared path nodes between the forklift and other vehicles. : ； in: Let A be the unexecuted path of forklift i, and let A be the set of all forklifts. It is the shared path node of forklift i; Indicates a path node. This is the unexecuted path of forklift j, where j represents any forklift other than forklift i; S72: Define the starting point of the nearest contiguous shared path. : ； in: It is a path Upper 1 node Represents path nodes and Whether they are adjacent: 1 for adjacent, 0 for not adjacent; It is a path Upper One node; S73: Define the nearest contiguous shared path As the potential deadlock path: ； S8: Decision on real-time control commands for each forklift based on the permission request strategy; the permission request strategy includes two policy conditions: First strategy condition: the next node on the forklift path ,and , ; Second strategy condition: the next node on the forklift path ,and , , ; in, This indicates whether node a has been reserved by forklift i; 1 indicates reserved and 0 indicates not reserved. Forklift i can move forward and request permission to the next resource point if it meets either the first or second policy condition; otherwise, the forklift will remain at the current node. S9: Determine if all tasks have ended. If so, end the process; otherwise, continue to step S10. S10: Determine whether the task update interval has been reached. If yes, return to step S3; otherwise, return to step S6.

2. The multi-unmanned forklift scheduling and control method including path planning and vehicle passage strategy according to claim 1, characterized in that, In step S1: The system reads map data and task data from the automated unmanned warehouse. The map data includes the location of the unmanned warehouse's docking area, assembly line unloading area, transit warehouse, storage area, cargo capacity information, unloading nodes, picking nodes, and path nodes accessible to forklifts. The task data includes the task's start node, task's end node, and task arrival time.

3. The multi-unmanned forklift scheduling and control method according to claim 1, comprising path planning and vehicle passage strategy, is characterized in that, In step S3: Establish the task pool and set a fixed time interval as the task update interval; pull the tasks to be planned from the task pool, match the corresponding forklifts for the tasks to be planned and plan the corresponding paths, and simultaneously allocate and update the warehouse location for the task endpoint nodes for online tasks arriving at the pipeline.

Images