A multi-carrier mobile robot path planning method and system
Patent Information
- Application Number
- CN202610547550.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-23
- Publication Date
- 2026-08-04
AI Technical Summary
[0002]在仓储场景中,多载位移动机器人(ML-AMR)集群的多智能体路径规划面临诸多挑战,现有单载位路径规划方法难以适配多载位回库与拣选耦合任务,易造成巷道占用、拥堵及无效路径增多;传统CBS类方法在大规模场景下会出现搜索空间膨胀的问题,导致路径求解效率低下,无法满足仓储作业的高效、稳定需求
本发明面向仓储场景下多载位移动机器人(ML-AMR)集群的多智能体路径规划问题,针对现有单载位路径规划方法难以适配多载位回库与拣选耦合任务、以及传统 CBS 类方法在大规模场景下搜索空间膨胀导致求解效率低的问题,提出一种基于改进大邻域搜索的多载位移动机器人路径规划方法及系统,包括获取机器人任务数据;基于机器人任务数据建立多载位移动机器人作业模型,基于多载位移动机器人作业模型进行路径规划;所述多载位移动机器人作业模型建立过程包括混合任务建模与任务接续、基于任务排序的初始解生成、基于优先级规则的邻域选择、改进SIPPS的局部重规划、LNS接受准则与迭代更新和终止条件与结果输出,能够在多载位、多任务、狭窄巷道的仓储环境中有效降低冲突数量、减少无效路径与重复路径,并提升大规模路径规划的成功率、效率与稳定性。
Smart Images

Figure CN122505255A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of path planning technology, specifically relating to a path planning method and system for multi-position mobile robots. Background Technology
[0002] In warehousing scenarios, multi-agent path planning for multi-position mobile robot (ML-AMR) clusters faces many challenges. Existing single-position path planning methods are difficult to adapt to the coupled tasks of multi-position return to the warehouse and picking, which can easily lead to aisle occupation, congestion and an increase in invalid paths. Traditional CBS-type methods will suffer from search space expansion in large-scale scenarios, resulting in low path solving efficiency and failing to meet the high-efficiency and stable requirements of warehousing operations. Summary of the Invention
[0003] The purpose of this invention is to overcome the shortcomings of the prior art. In a first aspect, this invention provides a path planning method for a multi-position mobile robot, comprising: Acquire robot task data; A multi-position mobile robot operation model is established based on robot task data, and path planning is performed based on the multi-position mobile robot operation model. The process of establishing the multi-position mobile robot operation model includes hybrid task modeling and task succession, initial solution generation based on task sorting, neighborhood selection based on priority rules, local replanning of improved SIPPS, LNS acceptance criteria and iterative update, and termination conditions and result output.
[0004] Furthermore, the hybrid task modeling and task succession steps include: after each robot completes delivery in the warehouse map, it does not enter a waiting state, but continues to the next round of tasks. The current return target and the next round of picking target are combined into a hybrid task set, and the sequence is optimized under the constraints of carrying capacity, task type and accessibility to form the candidate task sequence of the robot.
[0005] Furthermore, to address the issue that each multi-position robot needs to execute multiple target points, task sequence optimization is performed first, followed by calling a fast path planner to generate an initial path solution; dynamic programming is used for task sorting; let the set of task points to be executed by the robot be... The state cost is defined as:
[0006] in: A subset of visited task points; This is the current task point; To start from the mission point arrive The cost; To complete the set And the endpoint is Minimum cumulative cost at that time; After obtaining the task sequence, the fast MAPF planner is used to generate an initial set of paths for all robots:
[0007] This initial solution is allowed to contain conflicts, which can be used as the starting point for subsequent LNS2 iterations to fix them.
[0008] Furthermore, the neighborhood selection correction based on priority rules includes: in the t-th iteration, let the current global path solution be... The system collects conflict information and selects a subset of robots to be replanned based on priority rules. ; Define the neighborhood selection priority score for the i-th robot as:
[0009] in: Let i be the number of collisions in the current solution; This refers to the quantity of cargo carried. This represents the remaining task distance indicator; Weighting coefficients; according to The top K robots, selected from highest to lowest, form the replanning neighborhood:
[0010] This design allows neighborhood selection to no longer rely solely on conflicting locations, but rather incorporates multi-site task characteristics, thereby improving the effectiveness of replanning.
[0011] Furthermore, the improved local replanning steps of SIPPS include: fixing non-neighborhood robot paths and only replanning neighborhood paths. The robot in the process invokes an improved SIPPS for local replanning to generate new path solutions. ; The soft conflict tightening mechanism, in order to balance the feasibility in the early stage and the non-conflict in the later stage, sets the soft conflict penalty coefficient to increase with iteration, so that the algorithm allows a small number of soft conflicts in the early stage to quickly obtain a feasible path, and gradually strengthens the conflict penalty in the later stage until it approaches the hard conflict constraint. Define the soft conflict penalty weight as follows:
[0012] Where: t is the current iteration number; T is the maximum iteration number; These are the initial and maximum soft conflict penalty coefficients, respectively; As t increases As the conflict increases, the cost of soft conflicts gradually rises, achieving an iterative tightening effect. A replanning cost function with a turning penalty is introduced. This incorporates turning and waiting penalties into the individual path search cost of SIPPS to improve path execution quality. For candidate paths of robot i... Its cost is defined as:
[0013] in: Cost is the path length or travel time. The cost of waiting; Cost of turning (number of turns or turning time); The cost of soft conflict; Penalty coefficients for waiting and turning; Through this cost function, the replanner can resolve conflicts while suppressing unnecessary turns and waits, thereby improving path smoothness and execution efficiency.
[0014] Furthermore, the LNS acceptance criterion and iterative update steps include: defining the number of conflicting pairs in the current solution as... If the new solution after local reprogramming If the new solution reduces the global logarithm of conflicts, then accept the new solution; otherwise, keep the original solution unchanged. The acceptance criterion is written as:
[0015] This step ensures that the iteration process moves in the direction of conflict reduction, while maintaining the efficient local repair characteristics of LNS2.
[0016] Furthermore, the termination conditions and result output are as follows: The iteration terminates when any of the following conditions are met: 1) Global path solutions are conflict-free; 2) Reaching the maximum number of iterations T; 3) The time limit has been reached; Output the final set of conflict-free paths:
[0017] It is used to drive multi-position mobile robots to perform mixed picking, warehousing, and delivery tasks.
[0018] Secondly, the present invention provides a path planning system for a multi-position mobile robot, comprising: The data acquisition module is used to acquire robot task data; The path planning module establishes a multi-position mobile robot operation model based on robot task data, and performs path planning based on the multi-position mobile robot operation model. The process of establishing the multi-position mobile robot operation model includes hybrid task modeling and task succession, initial solution generation based on task sorting, neighborhood selection based on priority rules, local replanning of improved SIPPS, LNS acceptance criteria and iterative update, and termination conditions and result output.
[0019] The beneficial effects of this invention include: This invention addresses the multi-agent path planning problem of multi-position mobile robot (ML-AMR) clusters in warehousing scenarios. It tackles the challenges of existing single-position path planning methods being ill-suited for coupled tasks of multi-position back-to-warehouse and picking, and the low efficiency of traditional CBS-type methods due to search space expansion in large-scale scenarios. The invention proposes a multi-position mobile robot path planning method and system based on improved large neighborhood search. The method includes acquiring robot task data; establishing a multi-position mobile robot operation model based on the robot task data; and performing path planning based on the multi-position mobile robot operation model. The multi-position mobile robot operation model establishment process includes hybrid task modeling and task continuation, initial solution generation based on task ranking, neighborhood selection based on priority rules, local replanning using improved SIPPS, LNS acceptance criteria and iterative updates, termination conditions, and result output. This method effectively reduces the number of conflicts, invalid and duplicate paths, and improves the success rate, efficiency, and stability of large-scale path planning in warehousing environments with multiple positions, multiple tasks, and narrow aisles. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0021] Figure 1 This is a flowchart illustrating a path planning method for a multi-position mobile robot according to an embodiment of the present invention. Detailed Implementation
[0022] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.
[0023] like Figure 1As shown, this invention discloses a path planning method and system for a multi-position mobile robot, comprising: This invention addresses the multi-agent path planning problem of multi-position mobile robot (ML-AMR) swarms in warehousing scenarios. It tackles the challenges of existing single-position path planning methods being ill-suited for the coupled tasks of multi-position back-to-warehouse and picking, and the low solution efficiency of traditional CBS-type methods due to search space expansion in large-scale scenarios. Therefore, it proposes a multi-position mobile robot path planning method based on improved large neighborhood search. This invention constructs an integrated solution scheme combining a "return-to-stock picking hybrid strategy + improved MAPF-LNS2 path planning" without altering the overall framework of "large neighborhood search iterative repair". This scheme includes: 1) To minimize operation time and the proportion of inefficient paths, a multi-position robot hybrid picking-return-to-warehouse-delivery model is established; 2) Introduce priority rules based on the characteristics of multi-bit tasks during the neighborhood selection stage to improve the efficiency of conflict subset selection; 3) Introduce multi-task sequence-oriented sorting optimization in the initial planning stage and couple it with the fast planner to improve the quality of the initial path; 4) Improve SIPPS during the replanning phase by using a "soft conflict tightening" mechanism and a turning penalty cost design to enhance conflict resolution capabilities and path execution quality in complex warehousing environments.
[0024] Through the above solution, the present invention can effectively reduce the number of conflicts, reduce invalid and duplicate paths, and improve the success rate, efficiency and stability of large-scale path planning in warehouse environments with multiple locations, multiple tasks and narrow aisles.
[0025] The innovation of this invention lies not in simply replacing the existing MAPF solver, but in systematically modifying the LNS2 framework to suit the characteristics of multi-location warehousing operations, forming a complete technical solution, specifically including: (1) Hybrid task strategy of returning to warehouse and picking To address the problem of lane occupancy and congestion caused by the traditional "waiting for the next task after returning to the warehouse" method, this invention performs sequential joint optimization of the warehouse return task and the next picking task under constraints. This allows the robot to continue executing a mixed task sequence after completing delivery, reducing idle waiting and repetitive paths, and lowering the number of potential conflicts.
[0026] (2) Neighborhood selection mechanism based on priority rules To address the issue that the standard LNS2 neighborhood selection method does not fully utilize multi-cargo task information, this invention introduces task-related priorities (such as cargo quantity, distance, conflict intensity, etc.) into the neighborhood selection process, thereby improving the targeting of the replanned subset and thus enhancing the conflict resolution efficiency of each iteration.
[0027] (3) Initial planner for integrating task sorting In response to the characteristic of multi-position robots performing multiple return-to-warehouse / pickup target points in a single operation, this invention adds a task sequence optimization step before the initial path generation, making the initial solution not only "quickly obtainable" but also having a better task execution order, thereby reducing the burden of subsequent LNS replanning from the source.
[0028] (4) Improve the SIPPS replanner To address the issues of concentrated conflicts and frequent turning in high-density warehousing scenarios, this invention designs a mechanism in SIPPS where the "soft conflict tolerance gradually tightens with iteration" and adds a turning penalty term to the cost function, enabling the algorithm to have a stronger ability to obtain feasible solutions in the early stages and a higher path quality optimization ability in the later stages.
[0029] The mathematical objective and basic definitions, in order to balance operational efficiency and path quality, this embodiment uses "total operation time + inefficient path ratio" as the optimization objective. Assuming the robots jointly perform a mixed picking-return-to-delivery task, the overall objective function can be written as:
[0030] in: This represents the total operation time for all robots to complete the mixed task; This represents the ratio of inefficient path time to efficient path time. This is a weighting coefficient used to balance efficiency and path quality.
[0031] The inefficient path ratio can be expressed as:
[0032] in: This represents the total time for straight travel; Total turning time; Total waiting time; These are the turning penalty factor and the waiting penalty factor, respectively.
[0033] Algorithm steps: S1: Hybrid Task Modeling and Task Continuation A multi-position mobile robot operation model is established in the warehouse map. After each robot completes a delivery, it does not enter a waiting state, but continues to the next round of tasks. The current return target and the next round of picking target are combined into a mixed task set. The sequence is optimized under the constraints of position capacity, task type and accessibility to form the candidate task sequence of the robot.
[0034] The core of this step is to change the "return to warehouse" and "picking" processes from segmented execution to a hybrid sorting execution under constraints, in order to reduce empty runs and waiting in traffic.
[0035] S2: Initial solution generation based on task sequencing. For the problem that each multi-position robot needs to execute multiple target points, the task sequence is optimized first, and then a fast path planner (such as PP) is called to generate the initial path solution.
[0036] Dynamic programming can be used for task sorting. Let the set of tasks to be performed by the robot be... The state cost is defined as:
[0037] in: A subset of visited task points; This is the current task point; To start from the mission point arrive The cost (which can take into account factors such as distance, steering, and load conditions); To complete the set And the endpoint is The minimum cumulative cost at that time.
[0038] After obtaining the task sequence, the fast MAPF planner is used to generate an initial set of paths for all robots:
[0039] This initial solution is allowed to contain conflicts, which can be used as the starting point for subsequent LNS2 iterations to fix them.
[0040] S3: Neighborhood selection based on priority rules In the t-th iteration, let the current global path solution be... The system collects conflict information and selects a subset of robots to be replanned based on priority rules. .
[0041] Define the neighborhood selection priority score for the i-th robot as:
[0042] in: Let i be the number of collisions in the current solution; This refers to the quantity of cargo carried. This represents the remaining task distance indicator; These are the weighting coefficients.
[0043] according to The top K robots, selected from highest to lowest, form the replanning neighborhood:
[0044] This design allows neighborhood selection to no longer rely solely on conflicting locations, but rather incorporates multi-site task characteristics, thereby improving the effectiveness of replanning.
[0045] S4: Improve SIPPS's local replanning, fix non-neighborhood robot paths, and only replan neighborhood paths. The robot in the process invokes an improved SIPPS for local replanning to generate new path solutions. .
[0046] S4.1 Soft conflict gradually tightening mechanism: In order to balance the feasibility in the early stage and the non-conflict in the later stage, this invention sets the soft conflict penalty coefficient to increase with iteration, so that the algorithm allows a small number of soft conflicts in the early stage to quickly obtain a feasible path, and gradually strengthens the conflict penalty in the later stage until it approaches the hard conflict constraint.
[0047] The soft conflict penalty weight can be defined as follows:
[0048] in: t is the current iteration number; T is the maximum number of iterations; These are the initial and maximum soft conflict penalty coefficients, respectively.
[0049] As t increases As the conflict intensifies, the cost of soft conflict gradually increases, achieving an iterative tightening effect of "soft conflict gradually hardening".
[0050] S4.2 Introduces a reprogramming cost function with a turning penalty; To improve path execution quality, turning penalties and waiting penalties are added to the individual path search cost of SIPPS. For candidate paths of robot i... Its cost is defined as:
[0051] in: Cost is the path length or travel time. The cost of waiting; Cost of turning (number of turns or turning time); The cost of soft conflict; This refers to the penalty coefficient for waiting and turning.
[0052] Through this cost function, the replanner can resolve conflicts while suppressing unnecessary turns and waits, thereby improving path smoothness and execution efficiency.
[0053] S5: LNS Acceptance Criterion and Iterative Update, defining the number of conflicting pairs in the current solution as... If the new solution after local reprogramming... If a new solution reduces the global logarithm of conflicts, then the new solution is accepted; otherwise, the original solution is retained. The acceptance criterion is written as:
[0054] This step ensures that the iteration process moves in the direction of conflict reduction, while maintaining the efficient local repair characteristics of LNS2.
[0055] S6: Termination Condition and Output. The iteration terminates when any of the following conditions are met: 1) Global path solutions are conflict-free; 2) Reaching the maximum number of iterations T; 3) The time limit has been reached.
[0056] Output the final set of conflict-free paths:
[0057] It is used to drive multi-position mobile robots to perform mixed picking-back-to-delivery tasks.
[0058] Example 2 This embodiment provides a path planning system for a multi-position mobile robot, characterized by comprising: The data acquisition module is used to acquire robot task data; The path planning module establishes a multi-position mobile robot operation model based on robot task data, and performs path planning based on the multi-position mobile robot operation model. The process of establishing the multi-position mobile robot operation model includes hybrid task modeling and task succession, initial solution generation based on task sorting, neighborhood selection based on priority rules, local replanning of improved SIPPS, LNS acceptance criteria and iterative update, and termination conditions and result output.
[0059] The specific steps of the data acquisition module and the path planning module correspond one-to-one with the method in Example 1, and will not be repeated here.
[0060] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.
Claims
1. A path planning method for a multi-position mobile robot, characterized in that, include: Acquire robot task data; A multi-position mobile robot operation model is established based on robot task data, and path planning is performed based on the multi-position mobile robot operation model. The process of establishing the multi-position mobile robot operation model includes hybrid task modeling and task succession, initial solution generation based on task sorting, neighborhood selection based on priority rules, local replanning of improved SIPPS, LNS acceptance criteria and iterative update, and termination conditions and result output.
2. The path planning method for a multi-position mobile robot as described in claim 1, characterized in that, The steps of hybrid task modeling and task succession include: after each robot completes delivery in the warehouse map, it does not enter the waiting state, but continues to the next round of tasks. The current return target and the next round of picking target are combined into a hybrid task set. The sequence is optimized under the constraints of carrying capacity, task type and accessibility to form the candidate task sequence of the robot.
3. The path planning method for a multi-position mobile robot as described in claim 1, characterized in that, To address the issue of each multi-position robot needing to execute multiple target points, we first optimize the task sequence, then call a fast path planner to generate an initial path solution; dynamic programming is used for task sorting; let the set of task points to be executed by the robot be... The state cost is defined as: in: A subset of visited task points; This is the current task point; To start from the mission point arrive The cost; To complete the set And the endpoint is Minimum cumulative cost at that time; After obtaining the task sequence, the fast MAPF planner is used to generate an initial set of paths for all robots: This initial solution is allowed to contain conflicts, which can be used as the starting point for subsequent LNS2 iterations to fix them.
4. The path planning method for a multi-position mobile robot as described in claim 1, characterized in that, Neighborhood selection correction based on priority rules includes: In the t-th iteration, let the current global path solution be... The system collects conflict information and selects a subset of robots to be replanned based on priority rules. ; Define the neighborhood selection priority score for the i-th robot as: in: Let i be the number of collisions in the current solution; This refers to the quantity of cargo carried. This represents the remaining task distance indicator; Weighting coefficients; according to The top K robots, selected from highest to lowest, form the replanning neighborhood: This design allows neighborhood selection to no longer rely solely on conflicting locations, but rather incorporates multi-site task characteristics, thereby improving the effectiveness of replanning.
5. The path planning method for a multi-position mobile robot as described in claim 1, characterized in that, The improved local replanning steps of SIPPS include: fixing non-neighborhood robot paths and only replanning neighborhood paths. The robot in the process invokes an improved SIPPS for local replanning to generate new path solutions. ; The soft conflict tightening mechanism, in order to balance the feasibility in the early stage and the non-conflict in the later stage, sets the soft conflict penalty coefficient to increase with iteration, so that the algorithm allows a small number of soft conflicts in the early stage to quickly obtain a feasible path, and gradually strengthens the conflict penalty in the later stage until it approaches the hard conflict constraint. Define the soft conflict penalty weight as follows: Where: t is the current iteration number; T is the maximum iteration number; These are the initial and maximum soft conflict penalty coefficients, respectively; As t increases As the conflict increases, the cost of soft conflicts gradually rises, achieving an iterative tightening effect. A replanning cost function with a turning penalty is introduced. This incorporates turning and waiting penalties into the individual path search cost of SIPPS to improve path execution quality. For candidate paths of robot i... Its cost is defined as: in: Cost is the path length or travel time. The cost of waiting; Cost of turning (number of turns or turning time); The cost of soft conflict; Penalty coefficients for waiting and turning; Through this cost function, the replanner can resolve conflicts while suppressing unnecessary turns and waits, thereby improving path smoothness and execution efficiency.
6. The path planning method for a multi-position mobile robot as described in claim 1, characterized in that, The LNS acceptance criterion and iterative update steps include: defining the number of conflicting pairs in the current solution as... If the new solution after local reprogramming If the new solution reduces the global logarithm of conflicts, then accept the new solution; otherwise, keep the original solution unchanged. The acceptance criterion is written as: This step ensures that the iteration process moves in the direction of conflict reduction, while maintaining the efficient local repair characteristics of LNS2.
7. The path planning method for a multi-position mobile robot as described in claim 1, characterized in that, Termination conditions and output results: The iteration terminates when any of the following conditions are met: 1) Global path solutions are conflict-free; 2) Reaching the maximum number of iterations T; 3) The time limit has been reached; Output the final set of conflict-free paths: It is used to drive multi-position mobile robots to perform mixed picking, warehousing, and delivery tasks.
8. A path planning system for a multi-position mobile robot, characterized in that, include: The data acquisition module is used to acquire robot task data; The path planning module establishes a multi-position mobile robot operation model based on robot task data, and performs path planning based on the multi-position mobile robot operation model. The process of establishing the multi-position mobile robot operation model includes hybrid task modeling and task succession, initial solution generation based on task sorting, neighborhood selection based on priority rules, local replanning of improved SIPPS, LNS acceptance criteria and iterative update, and termination conditions and result output.
