Whole-process order fulfillment optimization method for workbin type robot warehousing system

Abstract

The invention provides a full-process order fulfillment optimization method for a workbin type robot warehousing system, and the method comprises the steps: S1, aiming at the full-process characteristics that an order arrives in the workbin type robot warehousing system, a workbin needs to be matched with the order for warehouse-out, and a robot is responsible for carrying; establishing a three-layer service link flexible coupling decision framework of order allocation, storage hit and robot scheduling; s2, a model decomposition algorithm based on mathematical programming is adopted, and efficient solving of a joint decision problem is achieved through hierarchical decomposition and iterative optimization; s3, according to the logic of the step S2, designing a mathematical heuristic algorithm based on a rotary fixed variable and suitable for medium and large scale problems so as to construct an order, workbin and robot optimal assignment and sorting result meeting the actual requirements of the multi-workbin robot warehouse; and S4, performing operation according to the order, workbin and robot optimal assignment and sorting result constructed in the step S3, and realizing the full-process order fulfillment joint optimization of the workbin type robot system.

Description

Technical Field

[0001] The present invention relates to the field of intelligent warehousing technology, and specifically to the full-process characteristics of a bin-type robot warehousing system, in which orders arrive, bins need to be matched with orders for shipment, and robots are responsible for handling. A full-process order fulfillment optimization method has been established. Background Art

[0002] Warehousing systems constitute half of the logistics network and represent a crucial breakthrough in reducing overall logistics costs. As a new form of productivity, robotic warehousing systems strongly support industrial production and logistics services, and are a key path to reducing costs and increasing efficiency in the warehousing and logistics industry. With robots replacing humans, warehousing systems operate continuously and uninterrupted. However, robotic warehousing systems present significant challenges in fulfilling large orders in a timely and flexible manner. In warehousing operations, order picking and storage costs account for over 60% of total warehouse operating costs, and this process is closely tied to the operation of orders, bins, and robots within the warehousing system. Once a new order arrives at the system, it must efficiently output high-quality results for multiple decisions, including ordering, storage, and handling. This places even more stringent demands on the solution algorithm, and different warehouse systems correspond to varying scales of solution requirements. The challenges of fast and flexible order fulfillment necessitate high-quality, innovative decision-making optimization technology. Therefore, it is particularly important to integrate the operation decision-making processes related to order arrival, bin hitting and robot assignment in the bin-type robot system and to design large-scale joint optimization algorithms. By optimizing the overall operation decision-making process and proposing an algorithm for solving large-scale mixed integer programming models, it is possible to meet the needs of modern warehousing systems for efficient and accurate order fulfillment and provide better support for its promotion and development.

[0003] Currently, existing algorithms are difficult to adapt to the complex decision-making environment of bin-type robotic warehousing systems. Bin-type robotic warehousing systems are highly dynamic and have complex variable coupling. They need to consider multi-level decision-making issues such as order distribution, bin assignment, and robot scheduling. However, when dealing with such complex systems, existing technologies are often unable to effectively integrate decisions at all levels to achieve global optimization. There are obvious limitations when dealing with large-scale mixed integer programming models. Such models are often used to describe complex decision-making problems in order fulfillment processes, but their solution process is complex and computational efficiency is low, making it difficult to quickly obtain effective solutions in large-scale data scenarios. They cannot meet the needs of actual warehousing systems for efficient and stable solution algorithms.

[0004] The existing technology urgently needs a full-process optimization method that can integrate order allocation, storage hits and robot scheduling to achieve efficient and accurate order processing in the warehousing system. Summary of the Invention

[0005] The main purpose of the present invention is to overcome the shortcomings of the existing technology and propose a full-process order fulfillment optimization method for a bin-type robot warehousing system. By constructing a flexible coupling decision framework and an innovative solution algorithm, it solves the local optimality and decision-making myopia caused by the separate optimization of orders or bins or robot operation decisions in the existing multi-bin robot system, and it is difficult to fully consider the technical problems of the mutual influence between orders, bins and robot operation decisions.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions: a full-process order fulfillment optimization method for a bin-type robot warehousing system, comprising: S1, in view of the full-process characteristics of the bin-type robot warehousing system, in which orders arrive, bins need to be matched with orders for delivery, and robots are responsible for handling, a three-layer business link flexible coupling decision framework of "order allocation-storage hit-robot scheduling" is established; S2, a model decomposition algorithm based on mathematical programming (MP-MD) is adopted to decompose the large-scale mixed integer programming model into a main problem and sub-problems, and the key decision variables, namely order allocation, bin assignment and robot task allocation, are fixed through hierarchical iteration to reduce the complexity of solution and achieve efficient solution of joint decision problems; for To solve the problem of low efficiency in solving large-scale mixed integer programming models for order fulfillment, the model decomposition algorithm based on mathematical programming is used to achieve efficient solution of joint decision-making problems through hierarchical decomposition and iterative optimization; S3, a mathematical heuristic algorithm based on rotating fixed variables is designed to adapt to medium and large-scale problems, and sub-problems are solved by cyclically fixing two layers of variables and optimizing the remaining layer variables, combined with the variable neighborhood search (VNS) framework to generate the optimal assignment and sorting results of orders, bins and robots that meet the actual needs of multi-bin robot warehouses; S4, operations are performed according to the optimal assignment and sorting results of orders, bins and robots obtained in step S3 to achieve full-process order fulfillment joint optimization of bin-type robot systems.

[0007] Furthermore, in step S1, the three-stage model for decision-making orders, storage units, and handling units is integrated. This includes the following: The order fulfillment integration model aims to minimize total order fulfillment time and integrates the following three sub-models: The order layer optimizes order allocation to picking stations and execution sequence, with the goal of minimizing order completion time differences; the storage unit layer solves the problem of matching bins with orders and sorting outbound shipments, with the goal of minimizing bin picking time and conveyor waiting time; and the handling unit layer implements task allocation and robot path planning, with the goal of minimizing robot travel distance and load balancing. Temporal logic constraints connect the decisions at each layer, forming a cross-stage joint optimization system. Let o represent the order number to be picked, where o = 1, 2, …, O, and O represents the number of orders to be picked. Let i represent the bin number to be stored and unstored, where i = 1, 2, …, I, and I represents the number of bins to be stored and unstored. Let r represent the robot number, where r = 1, 2, …, R, and R represents the number of bins to be stored and unstored. Use p to represent the number of the picking station, where p = 1, 2, ..., P, and P represents the number of incoming and outgoing bins. op ,y ip ,x ir They represent the assignment of orders to picking stations, the assignment of bins to picking stations, and the assignment of inbound and outbound tasks to robots.

[0008] Furthermore, in step S1, the decision framework for the entire process of order arrival, box matching with order, and robot handling in the bin-type robot warehousing system includes: S11, variable hierarchical design: the key decision variables include the order allocation matrix z o，p , bin assignment matrix y i,p , robot task matrix x i,k , directly determines the core decision of each stage; common decision variables: order sorting α, bin sequence β, robot path γ, are derived and generated through key variables. Through the flexible combination and configuration of two types of variables, it supports multi-link tightly coupled joint optimization and loosely coupled configurable decision-making, and independently optimizes the transportation path. S12. Objective function integration: The three phased sub-models of the entire order fulfillment process each define an independent optimization goal, reflecting the core efficiency indicators of different links. The integrated decision model takes the maximum value of the three-stage objective function T=max(T1, T2, T3) to ensure global optimization rather than local optimization. S13. The model associates cross-stage variables through time logic constraints, integrates the phased constraints of the order layer, storage unit layer (bin), and transportation unit layer (robot) into a unified system, and ensures the feasibility and coordination of the whole process decision. The order picking start time must be later than the time when the corresponding bin arrives at the picking station. Through the mathematical expression Only when order o is assigned to picking station p and bin i is assigned to p, the order start time is constrained by the bin, avoiding invalid calculation. Constraints at the storage unit level and the handling unit level: The time when the bin is picked and leaves the conveyor must be earlier than the start time of the robot handling task, through Ensure that the robot can only perform the return task after the material box reaches the conveyor exit to avoid path conflicts. Indirect constraints between the order layer and the handling unit layer: Through the indirect association of the storage unit layer, the order allocation z o,p Determine bin assignment y i,p , and then determine the robot task assignment x i,k S14. Finally, each stage model added to the model must meet its own business rules to ensure the feasibility of local decision-making. Picking station capacity limit: The number of orders processed simultaneously by each picking station does not exceed the number of slots ∑ o∈O z op ≤b; order execution order is mutually exclusive: through variable α o1,o2 Ensure that orders are not executed at the same picking station at the same time Storage unit level: Bin picking time constraint: The processing time of the bin at the picking station is only valid when assigned Bin outbound order constraints.

[0009] Furthermore, in step S2, the improved operation decision process of orders, bins and robots in the multi-bin robot system includes: S11. When O orders to be picked enter the multi-bin robot system, it is necessary to assign a picking station to each order and plan the execution order of the orders, that is, the order in which the orders enter the slots on the picking station. Because the number of slots on the picking station is limited, only when there are free slots on the picking station can the orders enter the free slots and start picking. The so-called execution order of the orders is the order in which the orders are bound to the picking station slots. Until all orders to be picked are assigned to picking stations, and the execution order of all orders to be picked is planned. S12. After the operation decision of the order is completed, it is necessary to make an assignment and sorting operation decision for the bins to be shipped obtained by the positioning operation for the order. When N bins waiting to be shipped enter the multi-bin robot system, they must be assigned to the corresponding picking stations based on the order assignment results. Based on the order execution sorting results at the picking stations, the execution order of all bins waiting to be shipped is planned. The so-called bin execution order is the order in which the bins are shipped. This continues until all shipping bins are assigned to picking stations and the execution order of all shipping bins is planned. S13: After the bin operation decision is completed, each bin waiting to be shipped is considered a shipping task. A robot must be assigned to these shipping tasks and the order in which the robots execute the tasks must be planned. When R robots enter the multi-bin robot system, they need to be assigned to robots in order based on the sorting results of the bins to be shipped. After all the tasks to be shipped are assigned, the order in which each robot will execute the tasks needs to be planned, and these tasks to be shipped need to be transported to the entrance of the ring conveyor. The so-called task execution order is the order in which the robots access these tasks. This process continues until all the shipping tasks are assigned to robots and the order in which each robot executes the tasks is planned. S14. When the bins enter the ring conveyor and are picked at the corresponding picking station, the bins will move to the exit of the ring conveyor. At this time, these bins are considered shipping tasks. Similarly, a robot needs to be assigned to these incoming tasks and the order in which the robots execute the tasks needs to be planned. When R robots enter the multi-bin robot system, they need to be assigned to the robots in order according to the order in which the bins arrive at the ring conveyor exit. After completing the assignment of all pending warehousing tasks, it is necessary to plan the order in which each robot executes the tasks, and transport these pending warehousing tasks to the corresponding shelf storage locations. The so-called task execution order is the order in which the robots access these tasks, until all warehousing tasks are assigned to robots and the order in which each robot executes the tasks is planned.

[0010] Furthermore, step S2 includes: for the large-scale mixed integer programming (MIP) model of order fulfillment in robotic warehousing system, the model decomposition algorithm (MP-MD) decomposes the original problem into a main problem and multiple sub-problems through hierarchical decomposition and iterative optimization, and gradually fixes the key decision variables to transform the complex full variable coupling problem into a single variable layer optimization problem, thereby reducing the solution complexity. Its core logic is: key variable identification: assign order z o,p , Material box assignment i,p , robot task allocation x i,k Defined as key variables, two of these variables are fixed in each iteration, and only one variable is optimized. Master-subproblem framework: The master problem logic controls the selection and iteration of subproblems. Subproblems are solved independently while other key variables are fixed. Ultimately, the optimal solutions to the subproblems are integrated to obtain the solution to the original problem.

[0011] Furthermore, in the full-process order fulfillment optimization method for the bin-type robotic warehousing system, in step S2, the main problem is logical decomposition and sub-problem selection. The current optimal sub-problem is dynamically selected through logical constraints to balance the optimization order of each layer of decision-making. Objective function: Among them, S=3 corresponds to the three sub-problems of order layer, storage layer, and transportation layer, θ s Select variables for subproblems (0-1 variables, only one subproblem is selected at a time), is the objective function value of subproblem s. Constraints: θ s ≤1-φ s ,θ s ∈{0,1},∑θ s = 1 to ensure that only one subproblem is optimized at a time; φ s The subproblem selected in the previous iteration is avoided from continuously optimizing the same subproblem, thus improving search diversity.

[0012] Furthermore, the three key decision variables in the original problem are assigned to order z o,p , Material box assignment i,p , robot task allocation x i,k It is divided into three layers. In each iteration, two layers of variables are fixed and only the remaining layer of variables is optimized. This transforms the original multivariable coupling problem into a relaxed optimization problem at the single variable layer, thus significantly reducing the complexity of the solution. Sub-problem 1 focuses on solving the problem of "how to allocate orders to picking stations and sort them". By fixing the decision-making of the bins and robots, multivariable interference is avoided and the optimal order allocation strategy is quickly found. Sub-problem 2 is the key variable of the storage unit layer sub-problem: y i,p , order allocation z o,p Fixed to z' o,p , robot task allocation x i,kFixed to x' i,k The goal is to optimize the strategy y for assigning bins to picking stations i,p and the order of the bins leaving the warehouse, minimize the maximum time T2 for the bins to complete the picking and arrive at the conveyor exit. Sub-problem 3 is the handling unit level sub-problem (key variable x i,k ) Fixed variable: Order allocation z o,p Fixed to z' o,p , Material box assignment i,p Fixed to y' i,p The goal is to optimize the robot task allocation strategy x i,k and path planning to minimize the maximum time it takes for the robot to complete all tasks, namely the total completion time T3.

[0013] Furthermore, step S3 includes four parts: the input of the main problem framework, the sub-problem algorithm, the main problem mathematical heuristic framework, and the main problem framework output, wherein the sub-problem algorithm includes: the sub-problem iterative search framework and the sub-problem general dynamic programming algorithm, and the sub-problem dynamic programming algorithm includes three dynamic programming algorithms, corresponding to the order layer, storage layer, and transportation layer of the model respectively. Specifically, this article divides the variables into multiple categories, and temporarily and iteratively fixes some variables in a rotating manner. Among them, the mathematical heuristic framework of the main problem is to control the iteration and end of the algorithm, and to determine the key variables to be fixed and relaxed. The sub-algorithm of the sub-problem is to solve several sub-problems, that is, the relaxed sub-problem model. The input and output of the main problem framework are tools for initializing the algorithm and outputting the optimal objective function value, respectively.

[0014] Furthermore, the initial solution of the joint decision model is constructed using a heuristic strategy. (1) Construction of the initial solution for order assignment and sorting: By calculating the similarity between any two orders To assign and sort orders, where Indicates whether order i contains the kth SKU. Assign orders with high similarity to the same picking station and sort the orders according to the size of the similarity. (2) Initial solution for bin assignment and sorting: Derived from the assignment results of orders to picking stations, the assignment of bins required by the order to picking stations is deduced, and the urgency of any bin i is calculated. To determine the order of the bins to be shipped out, the more urgent the bin is, the earlier it will be shipped out. onIndicates the orders currently at each picking station slot. (3) Initial solution for robot assignment and sorting: An adaptive nearest neighbor strategy is used to assign tasks to robots and determine the order in which robots execute tasks. Starting from the robot's starting point, the adaptive nearest neighbor strategy assigns the robot the nearest unvisited task as its next task until all tasks are assigned to the robot. The order in which tasks are assigned to the robot is the order in which the robot executes these tasks.

[0015] Furthermore, the function of the mathematical heuristic framework of the main problem is to control the iterative process, dynamically select the key variable layer (order layer, storage layer, and transport layer) that needs to be optimized, and update the upper and lower bounds of the problem. The core logic is (1) variable rotation strategy: according to "order allocation z o,p → Material box assignment y i,p →Robot task assignment x i,k "The variables are fixed in the order of loop. Each iteration only optimizes one type of key variables, and the other two types are fixed as the optimal solutions of the previous round. (2) Termination condition: When m consecutive max The target value of iterations has not been improved, or the maximum number of iterations id has been reached max Terminates when.

[0016] Furthermore, the relaxation models of its three sub-problems are solved based on the framework of variable neighborhood search: the framework consists of four parts: initial solution generation, neighborhood perturbation, local search, and new solution acceptance strategy. The dynamic programming algorithm is combined to solve the remaining variables to form a closed-loop optimization process. The generation of its initial solution lies in: the current key variable values ​​​​transmitted by the main problem and the other two layers of fixed variable values ​​​​transmitted by the main problem directly use the current solution transmitted by the main problem as the initial solution; if the initial solution is not feasible and violates the picking station capacity constraint, a feasible solution is generated through heuristic repair. For the key decision variables of each sub-problem, they all belong to the assignment problem. Therefore, the variable neighborhood search (VNS) framework based on the assignment problem will be initialized according to the results obtained in the previous stage to provide a starting point for the subsequent neighborhood search.

[0017] Furthermore, the purpose of perturbation in the neighborhood is to escape from local optimal solutions by perturbing the current solution, thereby increasing the diversity of solutions. For example, in the initial solution algorithm for order allocation and sorting, a greedy algorithm is used to sort the orders assigned to each picking station. However, to prevent falling into local optimality, the execution order of some orders is randomly swapped during the iteration process to generate new solutions.

[0018] Furthermore, its local search involves searching for a better solution within the neighborhood of the solution generated by Shaking. For example, during the initialization of storage unit assignment and sorting, the order of dispatching is determined based on the urgency of the current dispatch needs of all storage units in the warehouse. However, after each dispatch order is determined, a simulation is performed to simulate the dispatch of the storage unit, and the order status of the picking station is updated. Based on this new status, the dispatch urgency of the storage unit is recalculated to find a better dispatch order within the local area.

[0019] Furthermore, the new solution acceptance and algorithm termination characteristics are as follows: for the probability of new solution acceptance, we use simulated annealing as the standard, and the new solution S obtained in the kth iteration is new The probability of being accepted is Among them, T k Indicates the current temperature T k =T0c k , T0 represents the initial temperature, and c represents the cooling rate. The algorithm terminates when the number of iterations reaches a set value or when the number of iterations with a continuous solution reaches a set value. The output is: after multiple iterations of shaking and local search, the final optimal objective function value and decision variable values ​​are output. This is the neighborhood of the key decision variable value for the corresponding subproblem and the values ​​of the other two fixed decision variables.

[0020] Furthermore, the dynamic programming algorithm of its sub-problem is characterized by: (1) setting the starting point of the transport unit as the end point at the same time, converting it into a traveling salesman problem with access order constraints and resource limit constraints. Starting from the starting node, through dynamic programming, the optimal path and minimum travel distance to each node are calculated step by step, while considering the load limit and access order constraints of the transport unit. Specific steps: Initialize the starting node as the current node and calculate the next node; calculate the current load of the transport equipment based on the current node and the visited nodes; if the load is within the limit and all nodes have not been visited, continue to select the next node and update the path and objective function value; if the load is exceeded or all nodes have been visited, backtrack and calculate the path, and update the optimal path and objective function value. (2) After knowing the situation of each transport unit visiting each task point, simulate the storage unit's outbound arrival at the conveyor entrance, as well as the storage unit's access and picking at each picking station. The time it takes for the transport unit to arrive at each task point is calculated by the order in which the transport unit visits each task point, thereby deducing the time it takes for the storage unit to arrive at each task point, the conveyor entrance, and each picking station. After obtaining these times, the various flow conditions of the storage unit on the conveyor can be simulated to obtain the values ​​of the corresponding decision variables. Specific steps: According to the known decision variable values, the storage units are sorted and the outbound order is obtained; the flow conditions of each storage unit on the conveyor are calculated in sequence according to the outbound order; according to the flow conditions, the values ​​of the corresponding decision variables are calculated, such as the time it takes for the storage unit to arrive at the conveyor entrance, the picking time at each picking station, etc. (3) For each picking station, the picking conditions of the order at each picking station are calculated based on the known flow conditions of the storage unit on the conveyor. Using forward dynamic programming, starting from the distribution slot of the picking station, the picking order and completion time of each order are calculated in sequence. Specific steps: Calculate the picking order of each order in turn according to the order of the distribution slots of the picking station; calculate the completion time of the order at each picking station based on the picking order and the flow time of the conveyor; update the value of the decision variable, the completion time of the order, etc.

[0021] Furthermore, step S4 includes: S41. When a to-be-picked order enters the bin-type robot system, optimization of the flexible coupled decision framework for the three-layer business process of "order allocation-storage matching-robot scheduling" is initiated; S42. Initial solutions for order assignment and sorting, bin assignment and sorting, and robot assignment and sorting are obtained. S43. Dynamic programming simulations are performed for transport unit path planning, storage unit flow simulation, and order picking situation calculation, respectively, to calculate fitness values. S44. Fixed variables are iteratively rotated to obtain the optimal solution for the three-layer decision using a mathematical heuristic algorithm based on the fixed variable rotation. S45. According to the optimal result of order assignment and sorting, the order is assigned to the corresponding picking station, and according to the order sorting result, the order is placed in the free slot for picking in sequence; S46. According to the optimal result of material box assignment and sorting, the material box is assigned to the corresponding picking station, and according to the material box sorting result, the material box is shipped out in sequence; S47. According to the optimal result of outbound task assignment and sorting, the outbound task is assigned to the corresponding robot, and according to the outbound task sorting result, the outbound task is executed in sequence; S48. When the material box enters the ring conveyor and is picked at the corresponding picking station, the material box will go to the exit of the ring conveyor; S49. According to the optimal result of inbound task assignment and sorting, the inbound task is assigned to the corresponding robot, and according to the outbound task sorting result, the inbound task is executed in sequence; S50. Stop, all pending picking orders entering the multi-material box robot system are completed and leave the system.

[0022] The beneficial effects of the technical solution of the present invention are as follows: First, by constructing a flexible coupling decision framework for the three-layer business links of "order allocation-storage hit-robot scheduling", the present invention achieves systematization and coordination of the entire order fulfillment process, which can effectively improve the efficiency and quality of order processing. The close connection between each link reduces waiting time and resource waste in the order processing process, ensuring that orders can be completed quickly and accurately. Then, based on the decision framework, a model decomposition algorithm based on mathematical programming is designed to hierarchically decompose and iteratively optimize the complex large-scale mixed integer programming model for order fulfillment, breaking the original problem into a main problem and multiple sub-problems. By gradually fixing the key decision variables, the solution complexity is reduced. This can more accurately optimize order allocation, bin assignment, and robot task allocation, improve resource utilization, and reduce redundancy and waste. Then, a mathematical heuristic algorithm based on rotating fixed variables is used for medium and large-scale problems, which can construct the optimal assignment and sorting results of orders, bins, and robots that meet the actual needs of multi-bin robot warehouses. The algorithm transforms the complex multivariable coupling problem into a relaxed optimization problem at the single variable layer by rotating fixed variables, thereby greatly reducing the solution complexity, improving the solution efficiency, and providing support for the efficient operation of large-scale warehousing systems; finally, based on the constructed optimal assignment and sorting results of orders, bins and robots, the joint optimization results of orders, bins and robot operation decisions in the improved multi-bin robot system are obtained, so that the method of the present invention can obtain the joint optimization results of orders, bins and robot operation decisions at a faster computing speed, thereby improving the operating efficiency of bin-type robot warehouses, reducing the operating costs of warehouses, and providing better support for their promotion and development. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of the full-process order fulfillment optimization method of the material box type robot warehousing system in an embodiment of the present invention.

[0024] Figure 2 It is a decision framework diagram of the entire order fulfillment process in the material box type robot system in an embodiment of the present invention.

[0025] Figure 3 4 is a flow chart of a mathematical heuristic algorithm based on rotating fixed variables in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0027] like Figure 1 As shown, an embodiment of the present invention provides a full-process order fulfillment optimization method for a bin-type robotic warehousing system, comprising the following steps S1 to S4:

[0028] S1. In view of the whole process characteristics of the box-type robot warehousing system, where orders arrive, boxes need to be matched with orders and shipped out, and robots are responsible for handling, a flexible coupling decision-making framework for the three-layer business links of "order allocation-storage hit-robot scheduling" is established.

[0029] The order fulfillment integration model aims to minimize the total order fulfillment time and integrates the following three-stage sub-models: Order layer: Optimizes the allocation of orders to picking stations and the execution sequence, with the goal of minimizing the time difference between order completion; Storage unit layer: Solve the matching of bins and orders and sorting outbound shipments, with the goal of minimizing bin picking time and conveyor waiting time; Handling unit layer: Implements task allocation and robot path planning, with the goal of minimizing robot travel distance and load balancing. Through time logic constraints, the decisions of each layer are connected to form a cross-stage joint optimization system. Specific improvement steps include S11 to S14:

[0030] S11. Variable hierarchical design: key decision variables include the order allocation matrix z o，p , bin assignment matrix y i，p , robot task matrix x i，k , directly determining the core decisions at each stage; common decision variables: order sorting α, bin sequence β, and robot path γ, are derived from key variables. Through the flexible combination and configuration of these two types of variables, both tightly coupled joint optimization and loosely coupled configurable decision-making are supported for independent optimization of transport paths.

[0031] S12. Objective Function Integration: The three sub-models of the order fulfillment process each define independent optimization objectives, reflecting the core efficiency indicators of different links. The integrated decision model takes the maximum value of the three-stage objective function, T = max(T1, T2, T3), to ensure global optimization rather than local optimization.

[0032] S13. The model associates cross-stage variables through temporal logic constraints, integrating the staged constraints of the order layer, storage unit layer (bin), and handling unit layer (robot) into a unified system to ensure the feasibility and coordination of the whole process decision. The order picking start time must be later than the time when the corresponding bin arrives at the picking station. The mathematical expression Only when order o is assigned to picking station p and bin i is assigned to p, the order start time is constrained by the bin, avoiding invalid calculation. Constraints at the storage unit level and the handling unit level: The time when the bin is picked and leaves the conveyor must be earlier than the start time of the robot handling task, through Ensure that the robot can only perform the return task after the material box reaches the conveyor exit to avoid path conflicts. Indirect constraints between the order layer and the handling unit layer: Through the indirect association of the storage unit layer, the order allocation z o,p Determine bin assignment y i,p , and then determine the robot task assignment xi,k ;

[0033] S14. Finally, each stage model added to the model must meet its own business rules to ensure the feasibility of local decisions. Picking station capacity limit: Each picking station can process no more than the number of slots ∑ o∈O z op ≤b; order execution order is mutually exclusive: through variable α o1,o2 Ensure that orders are not executed at the same picking station at the same time Storage unit level: Bin picking time constraint: The processing time of the bin at the picking station is only valid when assigned Bin outbound order constraints.

[0034] The existing method for optimizing order fulfillment throughout the entire process of a bin-based robotic warehousing system was optimized individually. However, the improvements in step S1 above have enabled a systematized and coordinated approach to the entire order fulfillment process, effectively improving both the efficiency and quality of order processing. This seamless integration of all steps reduces waiting time and resource waste during order processing, ensuring that orders are completed quickly and accurately. This provides a model foundation for subsequent optimization efforts.

[0035] S2. Based on the joint order, bin and robot operation decision process, the model decomposition algorithm (MP-MD) is introduced. Through hierarchical decomposition and iterative optimization, the original problem is decomposed into a main problem and multiple sub-problems. By gradually fixing the key decision variables, the complex full-variable coupling problem is transformed into a single-variable layer optimization problem, thereby reducing the solution complexity. Its core logic is: key variable identification: order allocation z o,p , Material box assignment i,p , robot task allocation x i,k Defined as key variables, two of these variables are fixed in each iteration, and only one variable is optimized. Master-subproblem framework: The master problem logic controls the selection and iteration of subproblems. Subproblems are solved independently while other key variables are fixed. Ultimately, the optimal solutions to the subproblems are integrated to obtain the solution to the original problem.

[0036] like Figure 2 As shown in the figure, in an example warehouse, three orders are input, and these three orders are bound to five boxes that need to be shipped out. These five boxes will be shipped out by two robots and arrive at the entrance of the ring conveyor. The boxes entering the ring conveyor will go to one of the three picking stations for order picking. After the picking is completed, the boxes will go to the exit of the ring conveyor and finally be put into storage by two robots. Therefore, the operation decision of orders, boxes and robots can be made using z op ,y ip ,x ir and The optimal result of joint optimization of orders, bins and robot operation decisions can be obtained by simply solving the optimal values ​​of these variables.

[0037] S3. Using a mathematical heuristic algorithm based on rotating fixed variables, with the "rotating fixed variables + variable neighborhood search (VNS)" mechanism, dynamically optimize the key decision variables of orders, bins, and robots to construct the optimal assignment and sorting results of orders, bins, and robots that meet the actual needs of bin-type robot warehouses. Solving the operations research problem of assignment and sorting using a mathematical heuristic algorithm based on rotating fixed variables is achieved through the following steps. Figure 3 The figure below is a flowchart of an exemplary mathematical heuristic algorithm based on rotating fixed variables. First, an initial solution (initial encoding) for assignment and sorting is constructed to initialize the input parameters. Next, the key decision variables are determined. The VNS framework is then used to construct subproblems. Next, a dynamic programming algorithm is called to obtain the values ​​of other variables and update the upper and lower bounds. Finally, the conditions for accepting the new solution and terminating the algorithm are designed.

[0038] Specifically, the initial solution for order, bin and robot assignment and sorting, i.e., initial coding, is constructed using a heuristic strategy. (1) Initial coding for order assignment and sorting: By calculating the similarity between any two orders To assign and sort orders, where Indicates whether order i contains the kth SKU. Assign orders with high similarity to the same picking station, and sort the orders according to the size of the similarity. As shown in Table 1, based on the similarity between any two orders, orders o1 and o3 will be assigned to the same picking station, orders o2 and o4 will be assigned to the same picking station, and orders o5 and o6 will be assigned to the same picking station. (2) Initial coding of bin assignment and sorting: Through the assignment results of orders to picking stations, deduce the assignment of bins required by the order to the picking station, and calculate the urgency of any bin i To determine the order of the bins to be shipped out, the more urgent the bin is, the earlier it will be shipped out. onIndicates the orders currently at the slots of each picking station. As shown in Table 2, the order of bin delivery is 1, 4, 3, 2, and 5. (3) Initial coding of robot assignment and sorting: The adaptive nearest neighbor strategy is used to assign tasks to robots and determine the order in which robots perform tasks. Starting from the robot's starting point, the adaptive nearest neighbor strategy assigns the robot the nearest unvisited task as the next task until all tasks are assigned to the robot. At the same time, the order in which tasks are assigned to the robot is the order in which the robot performs these tasks. As shown in Table 3, tasks i1, i2, i3, i4, and i5 are assigned to robot r1, and tasks i6, i7, i8, i9, and i10 are assigned to robot r2. Table 1 Initial coding of order assignment and sorting Order number o1 o2 o3 o4 o5 o6 o1 1 0.43 0.64 0.22 0.30 0.10 o2 0.43 1 0.22 0.83 0.22 0.11 o3 0.64 0.22 1 0.16 0.38 0.15 o4 0.22 0.83 0.16 1 0.41 0.19 o5 0.30 0.22 0.38 0.41 1 0.88 o6 0.10 0.11 0.15 0.19 0.88 1 Table 2 Initial coding of bin assignment and sorting Bin\picking station p1 p2 p3 i1 0 2 6(1) i2 3(4) 0 0 i3 2 1 4(3) i4 5(2) 4 2 i5 2(5) 2 1 Table 3 Initial coding of task assignment and sorting

[0039] Specifically, the function of the mathematical heuristic framework for the main problem is to: control the iterative process, dynamically select the key variable layer (order layer, storage layer, and transport layer) that needs to be optimized, and update the upper and lower bounds of the problem. The core logic is (1) variable rotation strategy: according to "order allocation z o,p → Material box assignment y i,p →Robot task assignment x i,k "The variables are fixed in the order of loop. Each iteration only optimizes one type of key variables, and the other two types are fixed as the optimal solutions of the previous round. (2) Termination condition: When m consecutive max The target value of iterations has not been improved, or the maximum number of iterations id has been reached max Terminates when.

[0040] Specifically, the relaxation models of the three sub-problems are solved based on the framework of variable neighborhood search: the framework consists of four parts: initial solution generation, neighborhood perturbation, local search, and new solution acceptance strategy. The dynamic programming algorithm is combined to solve the remaining variables to form a closed-loop optimization process. The generation of its initial solution lies in: the current key variable values ​​​​delivered by the main problem and the other two layers of fixed variable values ​​​​delivered, directly using the current solution delivered by the main problem as the initial solution; if the initial solution is not feasible and violates the picking station capacity constraint, a feasible solution is generated through heuristic repair. For the key decision variables of each sub-problem, they all belong to the assignment problem, so the variable neighborhood search framework based on the assignment problem will be initialized according to the results obtained in the previous stage to provide a starting point for the subsequent neighborhood search.

[0041] Specifically, neighborhood perturbation aims to increase solution diversity by perturbing the current solution to escape from local optimal solutions. For example, in the initial solution algorithm for order allocation and sorting, a greedy algorithm is used to sort the orders assigned to each picking station. However, to prevent falling into local optimality, the execution order of some orders is randomly swapped during the iteration process to generate new solutions.

[0042] Specifically, local search involves searching for a better solution within the neighborhood of the perturbed solution. For example, during the initialization of storage unit assignment and sorting, the order of dispatching all storage units in the warehouse is determined based on the urgency of their current dispatch needs. However, after each dispatch order is determined, a simulation is performed to simulate the dispatch of the storage unit, updating the order status of the picking station. Based on this new status, the dispatch urgency of the storage unit is recalculated to find a better dispatch order within the local area.

[0043] Specifically, we design the characteristic conditions for accepting new solutions and terminating the algorithm. For the probability of accepting new solutions, we use simulated annealing as the standard. The new solution S obtained in the kth iteration is new The probability of being accepted is Among them, T k Indicates the current temperature T k =T0c k , T0 represents the initial temperature, and c represents the cooling rate. The algorithm terminates when the number of iterations reaches a set value or when the number of iterations during which the solution remains constant reaches a set value. Design output: After multiple iterations of dithering and neighborhood search, the final optimal objective function value and decision variable values ​​are output. This is the neighborhood of the key decision variable value for the corresponding subproblem and the values ​​of the other two fixed decision variables.

[0044] Specifically, a dynamic programming algorithm is designed to solve three sub-problems. (1) The starting point of the transport unit is set as the end point at the same time, and it is converted into a traveling salesman problem with access order constraints and resource limitation constraints. Starting from the starting node, the optimal path and minimum driving distance to each node are calculated step by step through dynamic programming, while considering the load limit and access order constraints of the transport unit. Specific steps: Initialize the starting node as the current node and calculate the next node; calculate the current load of the transport equipment based on the current node and the visited nodes; if the load is within the limit and all nodes have not been visited, continue to select the next node and update the path and objective function value; if the load is within the limit or all nodes have been visited, backtrack and calculate the path, and update the optimal path and objective function value. (2) After knowing the situation of each transport unit visiting each task point, simulate the storage unit's outbound arrival at the conveyor entrance, as well as the storage unit's access and picking at each picking station. The time it takes for the transport unit to arrive at each task point is calculated by the order in which the transport unit visits each task point, thereby deducing the time it takes for the storage unit to arrive at each task point, the conveyor entrance, and each picking station. After obtaining these times, the various flow conditions of the storage unit on the conveyor can be simulated to obtain the values ​​of the corresponding decision variables. Specific steps: According to the known decision variable values, the storage units are sorted and the outbound order is obtained; the flow conditions of each storage unit on the conveyor are calculated in sequence according to the outbound order; according to the flow conditions, the values ​​of the corresponding decision variables are calculated, such as the time it takes for the storage unit to arrive at the conveyor entrance, the picking time at each picking station, etc. (3) For each picking station, the picking conditions of the order at each picking station are calculated based on the known flow conditions of the storage unit on the conveyor. Using forward dynamic programming, starting from the distribution slot of the picking station, the picking order and completion time of each order are calculated in sequence. Specific steps: Calculate the picking order of each order in sequence according to the sorting slot order of the picking station; Calculate the completion time of the order at each picking station based on the picking order and the flow time of the conveyor; Update the value of the decision variable, the completion time of the order, etc.

[0045] Finally, follow the Figure 3 The mathematical heuristic algorithm flowchart based on rotating fixed variables and the above steps are performed as shown, and the final result of full-process order fulfillment optimization is obtained.

[0046] S4. Based on the optimal assignment and sorting results of the orders, bins, and robots constructed in step S3, obtain the joint optimization results of the orders, bins, and robot operation decisions in the improved multi-bin robot system. The specific process is as follows:

[0047] S41. When an order to be picked enters the multi-bin robot system, the joint optimization of the order, bins, and robot operation decisions begins.

[0048] S42, obtaining a joint optimization result of orders, bins, and robot operation decisions in the improved multi-bin robot system;

[0049] S43. Assign the order to the corresponding picking station according to the optimal order assignment and sorting result, and place the order in the free slot for picking in order according to the order sorting result;

[0050] S44. Assign the bins to the corresponding picking stations based on the optimal bin assignment and sorting results, and remove the bins in order based on the bin sorting results.

[0051] S45. Assign the outbound tasks to the corresponding robots according to the optimal results of outbound task assignment and sorting, and execute the outbound tasks in order according to the outbound task sorting results;

[0052] S46. When the material box enters the ring conveyor and is picked at the corresponding picking station, the material box will go to the exit of the ring conveyor;

[0053] S47. Assign the incoming tasks to the corresponding robots based on the optimal results of the incoming task assignment and sorting, and execute the incoming tasks in order based on the sorting results of the outgoing tasks;

[0054] S48. Stop. All orders waiting to be picked that have entered the multi-bin robot system are completed and leave the system.

[0055] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. Those skilled in the art will recognize that several equivalent substitutions or obvious variations can be made without departing from the scope of the present invention, and that any equivalent performance or application should be considered to fall within the scope of protection of the present invention.

Claims

1. A full-process order fulfillment optimization method for a bin-type robotic warehousing system, characterized by: S1. In view of the full process characteristics of the bin-type robot warehousing system, where orders arrive, bins need to be matched with orders and shipped out, and robots are responsible for handling, a flexible coupling decision-making framework for the three business links of "order allocation - storage matching - robot scheduling" is established; S2. Using a mathematical programming-based model decomposition algorithm (MP-MD), the large-scale mixed integer programming model is decomposed into a main problem and subproblems. Through hierarchical iteration, the key decision variables, namely order allocation, bin assignment, and robot task allocation, are fixed, reducing the solution complexity and achieving efficient solution of the joint decision problem. S3. Design a mathematical heuristic algorithm based on rotating fixed variables. By cyclically fixing two layers of variables and optimizing the remaining layers of variables, combined with the variable neighborhood search (VNS) framework to solve subproblems, it generates the optimal assignment and sorting results of orders, bins, and robots that meet the actual needs of a multi-bin robotic warehouse. S4. Perform operations based on the optimal assignment and sorting results of the order, material box and robot obtained in step S3 to achieve full-process order fulfillment joint optimization of the material box type robot system.

2. The full-process order fulfillment optimization method for a bin-type robotic warehousing system according to claim 1, characterized in that: In step S1, the three-stage models of decision order, storage unit, and handling unit are integrated, specifically including: The order fulfillment integration model aims to minimize the total order fulfillment time and integrates the following three-stage sub-models: order layer: optimizes the allocation of orders to picking stations and the execution sequence, with the goal of minimizing the order completion time difference; storage unit layer: solves the matching of bins and orders and the outbound sorting, with the goal of minimizing the bin picking time and conveyor waiting time; handling unit layer: realizes task allocation and robot path planning, with the goal of minimizing the robot's travel distance and load balancing; connects the decisions of each layer through time logic constraints to form a cross-stage joint optimization system; use o to represent the number of the order to be picked, where o = 1, 2, ..., O, O represents the number of orders to be picked; use i to represent the number of the bins to be in and out of the warehouse, where i = 1, 2, ..., I, I represents the number of bins in and out of the warehouse; use r to represent the number of the robot, where r = 1, 2, ..., R, R represents the number of bins in and out of the warehouse; use p to represent the number of the picking station, where p = 1, 2, ..., P, P represents the number of bins in and out of the warehouse; use z op ,y ip , x ir They represent the assignment of orders to picking stations, the assignment of bins to picking stations, and the assignment of inbound and outbound tasks to robots.

3. The full-process order fulfillment optimization method for a bin-type robotic warehousing system according to claim 2, characterized in that: In step S1, the three-stage models of decision order, storage unit, and handling unit are integrated, specifically including: S11. Variable hierarchical design: key decision variables include the order allocation matrix z o,p , bin assignment matrix y i,p , robot task matrix x i,k , directly determines the core decisions of each stage; common decision variables: order sorting α, bin sequence β, robot path γ, are derived and generated through key variables; through the flexible combination and configuration of the two types of variables, it supports multi-link tightly coupled joint optimization and loosely coupled configurable decision-making, and independently optimizes the handling path; S12. Objective Function Integration: The three sub-models of the entire order fulfillment process each define independent optimization objectives, reflecting the core efficiency indicators of different links. The integrated decision model takes the maximum value of the three-stage objective function T = max(T1, T2, T3) to ensure global optimization rather than local optimization. S13. The model associates cross-stage variables through temporal logic constraints, integrating the phased constraints of the order layer, storage unit layer (i.e., bin), and handling unit layer (i.e., robot) into a unified system to ensure the feasibility and coordination of the entire process decision-making process. The order picking start time is later than the time when the corresponding bin arrives at the picking station. Through the mathematical expression Only when order o is assigned to picking station p and bin i is assigned to p, the order start time is constrained by the bin to avoid invalid calculation; the constraints of storage unit layer and handling unit layer: the time when the bin is picked and leaves the conveyor must be earlier than the start time of the robot handling task, through Ensure that the robot performs the return task only after the material box reaches the conveyor exit to avoid path conflicts; indirect constraints between the order layer and the handling unit layer: through indirect association at the storage unit layer, order allocation z o,p Determine bin assignment y i,p , and then determine the robot task assignment x i,k ; S14. Each stage model added to the model must meet its own business rules to ensure the feasibility of local decision-making; Picking station capacity limit: Each picking station can process no more than the number of slots ∑ o∈O z op ≤b; order execution order is mutually exclusive: through variable α o1,o2 Ensure that orders are not executed at the same picking station at the same time Storage unit level: Bin picking time constraint: The processing time of the bin at the picking station is only valid when assigned Bin outbound order constraints.

4. The full-process order fulfillment optimization method for a bin-type robotic warehousing system according to claim 3, characterized in that: Step S2 specifically includes: Aiming at the large-scale mixed integer programming (MIP) model of order fulfillment in robotic warehousing system, the model decomposition algorithm (MP-MD) decomposes the original problem into a main problem and multiple sub-problems through hierarchical decomposition and iterative optimization. By gradually fixing the key decision variables, the complex full-variable coupling problem is transformed into a single-variable layer optimization problem, thereby reducing the solution complexity. Among them, the order allocation z o,p , Material box assignment i,p , robot task allocation x i,k Defined as key variables, two of these variables are fixed in each iteration, and only one variable is optimized. The selection and iteration of subproblems are controlled by the logic of the main problem. Subproblems are solved independently on the basis of fixing other key variables, and the optimal solutions of the subproblems are finally integrated to obtain the solution to the original problem. Preferably, in step S2, the main problem is logical decomposition and sub-problem selection, and the current optimal sub-problem is dynamically selected through logical constraints to balance the optimization order of each layer of decision-making; the objective function is: Among them, S=3 corresponds to the three sub-problems of order layer, storage layer, and transportation layer, θ s Select variables for subproblems, that is, 0-1 variables, and only select one subproblem at a time. is the objective function value of subproblem s; constraints: θ s ≤1-φ s ,θ s ∈{0,1},∑θ s = 1 to ensure that only one subproblem is optimized at a time; φ s The subproblem selected for the previous iteration.

5. The full-process order fulfillment optimization method for a bin-type robotic warehousing system according to claim 4, characterized in that: In step S2, the three key decision variables in the original problem are assigned to order z o,p , Material box assignment i,p , robot task allocation x i,k The algorithm is divided into three layers. In each iteration, two layers of variables are fixed, and only the remaining layer is optimized. This transforms the original multivariable coupling problem into a relaxed optimization problem with a single variable layer. Subproblem 1 focuses on how to allocate and sort orders to picking stations. By fixing the bins and robot decisions, multivariable interference is avoided and the optimal order allocation strategy is quickly found. Sub-problem 2 is the key variable of the storage unit layer sub-problem: y i,p , order allocation z o,p Fixed to z' o,p , robot task allocation x i,k Fixed to x' i,k ; The goal is to optimize the strategy y for assigning bins to picking stations i,p and the order of the bins leaving the warehouse, minimize the maximum time T2 for the bins to complete picking and reach the conveyor exit; Sub-problem 3 is the key variable x of the handling unit layer sub-problem i,k , fixed variable: order allocation z o,p Fixed to z' o,p , Material box assignment i,p Fixed to y' i,p ; The goal is to optimize the robot task allocation strategy x i,k and path planning to minimize the maximum time it takes for the robot to complete all tasks, namely the total completion time T3.

6. The full-process order fulfillment optimization method for a bin-type robotic warehousing system according to claim 5, characterized in that: Step S3 includes: main problem framework input, sub-problem algorithm, main problem mathematical heuristic framework, and main problem framework output, wherein the sub-problem algorithm includes: sub-problem iterative search framework and sub-problem general dynamic programming algorithm, and the sub-problem dynamic programming algorithm includes three dynamic programming algorithms, corresponding to the order layer, storage layer and transportation layer of the model respectively; the variables are divided into multiple categories, and some variables are temporarily and iteratively fixed in a rotating manner; wherein, the mathematical heuristic framework of the main problem controls the iteration and end of the algorithm, and determines the key variables to be fixed and relaxed, and solves several sub-problems through the sub-algorithm of the sub-problem, that is, the relaxed sub-problem model, and the input and output of the main problem framework are tools for algorithm initialization and output of the optimal objective function value respectively.

7. The full-process order fulfillment optimization method for a bin-type robotic warehousing system according to claim 6, characterized in that: In step S3, the initial solution of the joint decision model is constructed using a heuristic strategy; (1) Construction of the initial solution for order assignment and sorting: by calculating the similarity between any two orders To assign and sort orders, where Indicates whether order i contains the kth SKU; assigns orders with high similarity to the same picking station, and sorts the orders according to the size of the similarity; (2) Initial solution for bin assignment and sorting: deduce the assignment of bins required for the order to the picking station through the assignment results of orders to picking stations, and calculate the urgency of any bin i To determine the order of the bins to be shipped out, the more urgent the bin is, the earlier it will be shipped out. on Indicates the orders currently at the slots of each picking station; (3) Initial solution for robot assignment and sorting: The adaptive nearest neighbor strategy is used to assign tasks to robots and determine the order in which robots execute tasks; starting from the robot's starting point, the adaptive nearest neighbor strategy assigns the nearest unvisited task to the robot as the next task until all tasks are assigned to the robot. At the same time, the order in which tasks are assigned to the robot is the order in which the robot executes these tasks; Preferably, the mathematical heuristic framework of the main problem is configured to: control the iterative process, dynamically select the key variable layer to be optimized, namely the order layer, storage layer, and transportation layer, and update the upper and lower bounds of the problem; the core logic is (1) variable rotation strategy: according to "order allocation z o,p → Material box assignment y i,p →Robot task assignment x i,k "The variables are fixed in the order of loop. Each iteration only optimizes one type of key variables, and the other two types are fixed as the optimal solutions of the previous round. (2) Termination condition: When m consecutive max The target value of iterations has not been improved, or the maximum number of iterations id has been reached max Terminates when.

8. The full-process order fulfillment optimization method for a bin-type robotic warehousing system according to claim 7, characterized in that: The relaxed models of the three subproblems are solved based on the framework of variable neighborhood search: The framework consists of four parts: initial solution generation, neighborhood perturbation, local search, and new solution acceptance strategy. It is combined with a dynamic programming algorithm to solve the remaining variables, forming a closed-loop optimization process. The initial solution generation includes: directly using the current key variable values ​​and other two-layer fixed variable values ​​transmitted by the main problem as the initial solution; if the initial solution is infeasible, violating the picking station capacity constraint, generating a feasible solution through heuristic repair; for the key decision variables of each subproblem, the variable neighborhood search (VNS) framework based on the assignment problem is initialized based on the results obtained in the previous stage, providing a starting point for subsequent neighborhood searches; The neighborhood perturbation includes: perturbing the current solution to escape from the local optimal solution and increase the diversity of solutions; optionally, in the initial solution algorithm for order allocation and sorting, a greedy algorithm is used to sort the orders assigned to each picking station, and the execution order of some orders is randomly exchanged during the iteration process to generate new solutions; The local search includes: performing a local search within the neighborhood of the solution generated by Shaking to find a better solution; for example, in the initialization of storage unit assignment and sorting, determining the order of outbound delivery based on the urgency of the current delivery needs of all storage units in the warehouse; after each determination of the delivery order, simulating the situation where the storage unit has already been delivered, updating the order status of the picking station, and then recalculating the delivery urgency of the storage unit based on the new status, thereby finding a better delivery order within the local range; The new solution acceptance includes: using simulated annealing as a standard for the probability of new solution acceptance, and the new solution S obtained in the kth iteration new The probability of being accepted is Among them, T k Indicates the current temperature T k =T0c k , T0 represents the initial temperature, c represents the cooling rate; the algorithm’s termination strategy is to use the number of iterations to reach the set value or the number of iterations with the solution continuously unchanged to reach the set value; After multiple iterations of jittering and local search, the final optimal objective function value and decision variable value are output, that is, the neighborhood of the key decision variable value of the corresponding subproblem and the values ​​of the other two fixed decision variables are obtained.

9. The full-process order fulfillment optimization method for a bin-type robotic warehousing system according to claim 8, characterized in that: The dynamic programming algorithms for the subproblems include: (1) The starting point of the transport unit is set as the end point at the same time, and it is converted into a traveling salesman problem with access order constraints and resource limit constraints; starting from the starting node, the optimal path and minimum travel distance to each node are calculated step by step through dynamic programming, while considering the load limit and access order constraints of the transport unit; the specific steps include: initializing the starting node as the current node and calculating the next node; calculating the current load of the transport equipment based on the current node and the visited nodes; if the load is within the limit and all nodes have not been visited, continue to select the next node and update the path and objective function value; if the load is exceeded or all nodes have been visited, backtrack and calculate the path, and update the optimal path and objective function value; (2) After the access of each transport unit to each task point is known, simulate the storage unit's outbound arrival at the conveyor entrance, as well as the storage unit's access and picking at each picking station; calculate the time when the transport unit arrives at each task point by the order in which the transport unit visits each task point, thereby deriving the time when the storage unit arrives at each task point, the conveyor entrance, and each picking station; based on these times, simulate various flow conditions of the storage unit on the conveyor, thereby obtaining the values ​​of the corresponding decision variables; the specific steps include: sorting the storage units and obtaining the outbound order according to the known decision variable values; calculating the flow of each storage unit on the conveyor in sequence according to the outbound order; and calculating the values ​​of the corresponding decision variables based on the flow conditions, such as the time when the storage unit arrives at the conveyor entrance, the picking time at each picking station, etc.; (3) For each picking station, the picking status of the order at each picking station is calculated based on the known flow of storage units on the conveyor; starting from the distribution slot of the picking station, the picking sequence and completion time of each order are calculated in sequence through forward dynamic programming; the specific steps include: calculating the picking sequence of each order in sequence according to the distribution slot sequence of the picking station; calculating the completion time of the order at each picking station according to the picking sequence and the flow time of the conveyor; updating the value of the decision variable, the completion time of the order, etc.

10. The full-process order fulfillment optimization method for a bin-type robotic warehousing system according to claim 9, characterized in that: Step S4 specifically includes: S41: When an order to be picked enters the bin-type robot system, the flexible coupling decision framework optimization of the three-layer business links of "order allocation - storage hit - robot scheduling" begins; S42, obtaining an initial solution structure for order assignment and sequencing, bin assignment and sequencing, and robot assignment and sequencing; S43, performing dynamic programming simulations for transport unit path planning, storage unit flow simulation, and order picking situation calculation, respectively, and calculating fitness values; S44, iteratively rotating the fixed variables to obtain a three-layer decision-making optimal solution based on a mathematical heuristic algorithm of the rotating fixed variables; S45. Based on the optimal result of order assignment and sorting, the order is assigned to the corresponding picking station, and the order is placed in the free slot for picking in order according to the order sorting result; S46. Assign the bins to the corresponding picking stations based on the optimal bin assignment and sorting results, and remove the bins in order based on the bin sorting results. S47. Assign the outbound tasks to the corresponding robots according to the optimal outbound task assignment and sorting results, and execute the outbound tasks in order according to the outbound task sorting results; S48, when the material box enters the ring conveyor and is picked at the corresponding picking station, the material box moves to the exit of the ring conveyor; S49. Assign the incoming tasks to the corresponding robots based on the optimal results of the incoming task assignment and sorting, and execute the incoming tasks in order based on the sorting results of the outgoing tasks; S50: Stop. All orders waiting to be picked that have entered the multi-bin robot system are completed and leave the system.

Images