ASRS Task Scheduling and Storage Location Allocation Method and System under Classified Storage

Abstract

The present invention discloses a method and system for ASRS task scheduling and storage location allocation under classified storage. The shelves are partitioned according to the characteristics of goods. The model for inbound and outbound task scheduling and storage location allocation considers the dynamic allocation of storage locations. This model is a dynamic programming model that includes the allocation of stacker cranes for inbound and outbound tasks, the sorting of inbound tasks, the sorting of outbound tasks, and the state transition of the storage location set. An integer programming model is established with the goal of minimizing the Hamming distance of the inbound and outbound task matching to sort the inbound tasks. The inbound task sequence optimization model is embedded in the memetic algorithm to achieve the overall optimization of the inbound and outbound task sorting, allocation, and storage location selection problems. The local search of the memetic algorithm includes two storage location exchange operators; the present invention optimizes the sequence of outbound and inbound tasks. The empty storage locations generated by the outbound tasks can be used for subsequent inbound operations. At the same time, the optimal solution is obtained through a method combining global and local searches, reducing the task completion time and delay time.

Description

Technical Field

[0001] The present invention relates to a method and system for warehouse task scheduling and allocation, especially an ASRS task scheduling and storage location allocation method under classified storage. Background Art

[0002] Automated Storage / Retrieval System (ASRS) has the advantages of high space utilization rate, low labor cost, fast goods in-and-out speed, etc., and is widely used in distribution centers and other fields. The classified storage strategy is one of the commonly used storage strategies in ASRS. It considers the in-and-out frequency or attribute characteristics of goods, divides the goods into different zones for placement, and adopts the random storage strategy within each zone. The integrated optimization of ASRS task sequencing and storage location allocation means that given an outbound task list and a set of empty storage locations, the outbound tasks are sequenced and appropriate storage locations are selected for the inbound tasks.

[0003] Currently, the inbound task sequence adopts the First-Come-First-Served (FCFS) strategy, without considering the impact of the inbound task sequencing under zone constraints on the operation efficiency of the stacker crane, especially the running distance of the stacker crane and the delay of outbound tasks. The existing technologies are studied based on a single aisle, while in reality, ASRS contains multiple aisles. There are few literatures considering the stacker crane allocation problem from the global perspective of multiple aisles. And in the existing technologies, the optimization goal is to improve the operation efficiency of ASRS, and less consideration is given to the situation where the outbound tasks have deadlines. Blindly pursuing operation efficiency may lead to delays in customer order delivery, a decrease in customer satisfaction, and a reduction in the competitiveness of enterprises. Summary of the Invention

[0004] Object of the Invention: The object of the present invention is to provide an ASRS task scheduling and storage location allocation method under classified storage that reduces the task completion time and delay time; the second object of the present invention is to provide an ASRS task scheduling and storage location allocation system under classified storage that reduces the task completion time and delay time.

[0005] Technical Solution: In the ASRS task scheduling and storage location allocation method under classified storage according to the present invention, the shelves are partitioned according to the characteristics of the goods, and the goods characteristics include the in-and-out frequency of the goods and the attributes of the goods; the model of the in-and-out task scheduling and storage location allocation is a dynamic programming model including the allocation of stacker cranes for in-and-out tasks, the sequencing of inbound tasks, the sequencing of outbound tasks, and the state transition of the set of storage locations; the empty storage locations generated by outbound tasks can be used by subsequent inbound tasks; each trip of the stacker crane is an in-and-out trip, including inbound tasks and outbound tasks; the method for in-and-out task scheduling and storage location allocation is as follows:

[0006] Aiming at minimizing the completion time of the inbound and outbound trips and the delay time of the outbound tasks, a memetic algorithm is used to solve the dynamic programming model. When the termination condition is met, the outbound and inbound tasks and their processing sequences to be executed by each stacker are output, as well as the optimal solutions of the storage locations corresponding to each inbound and outbound task.

[0007] In the global optimization of the memetic algorithm, according to the initial outbound task sequence and stacker allocation, by solving the assignment problem model aiming at minimizing the matching metric index of the inbound and outbound tasks, the inbound task sequence and stacker allocation are obtained.

[0008] In the memetic algorithm, local search is used to further optimize the inbound task sequence and the outbound task sequence. The local search includes two storage location exchange operators, which respectively represent exchanging the inbound storage location and the outbound storage location of two inbound and outbound trips on a stacker. In each local search, one of them is selected for local search according to the historical performance of the two storage location exchange operators to form a new inbound and outbound trip.

[0009] Furthermore, it is characterized in that the matching metric index of the outbound task and the inbound task is the new Hamming distance, and the formula of the new Hamming distance is:

[0010]

[0011] Where is the outbound task cargo area sequence, is the inbound task cargo area sequence, a i ∈Ω, b i ∈Ω, Ω represents the cargo area sequence. The cargo areas are divided into p categories of cargo areas in turn from near to far from the I / O port, Ω=(Ω1, Ω2...Ω p ). By using the new Hamming distance as the matching metric index of the outbound task and the inbound task, the inbound task sorting problem is transformed into an assignment problem for solution.

[0012] Furthermore, the dynamic programming model divides stages according to the number of outbound tasks. Each stage, based on the current storage location set state, solves an integer programming model aiming at minimizing the completion time of the inbound and outbound trips and the delay time of the outbound tasks to select storage locations for the inbound and outbound tasks. After a single stage ends, the storage location set state is updated and transferred to the next stage until all inbound and outbound tasks are executed.

[0013] Furthermore, in the memetic algorithm, a genetic algorithm is used for global search, including: performing chromosome coding on the outbound task sequence and stacker allocation to generate the initial outbound task sequence and stacker allocation;

[0014] Select storage locations for inbound and outbound tasks according to chromosome coding, inbound and outbound task sequences, and stacker allocation; select the outbound storage location closest to the I / O port and containing the outbound goods within the storage area to which the outbound task belongs, and select an empty storage location within the storage area to which the inbound task belongs that can minimize the completion time of the inbound and outbound journey; after allocating storage locations for an inbound and outbound journey, update the status of the storage locations until all storage locations are allocated.

[0015] Further, the method for further optimizing the inbound task sequence and outbound task sequence using local search is as follows:

[0016] Perform local search on the initial population in the memetic algorithm, as well as the individuals after crossover and mutation. During each local search, use P(N) = c N / ∑ N∈{1，2} c N Select a storage location swapping operator for local search, where N ∈ {1, 2}, representing two storage location swapping operators, P(N) is the probability of the storage location swapping operator being selected, and c N is a counter. The counter is used to record the historical performance of the storage location swapping operator. When the solution after local search using a certain storage location swapping operator is better than the current solution, the counter is incremented by one.

[0017] Further, in the memetic algorithm, the reciprocal of the objective function that minimizes the completion time of the inbound and outbound journey and the delay time of the outbound task is used as the fitness function. The optimal solution of the memetic algorithm is the solution that maximizes the fitness function value when the termination condition is met.

[0018] The ASRS task scheduling and storage location allocation system under classified storage according to the present invention includes:

[0019] A model establishment unit for establishing a dynamic programming model for stacker allocation of inbound and outbound tasks, inbound task sorting, outbound task sorting, and state transition of the storage location set. The empty storage locations generated by the outbound tasks can be used for subsequent inbound operations; each journey of the stacker is an inbound and outbound journey, including inbound tasks and outbound tasks; the shelves are partitioned according to the characteristics of the goods, and the characteristics of the goods include the frequency of inbound and outbound of the goods and the attributes of the goods;

[0020] A model solving unit for solving the dynamic programming model using a memetic algorithm with the goal of minimizing the completion time of the inbound and outbound journey and the delay time of the outbound task, and outputting the optimal solution of the outbound and inbound tasks and the processing sequence to be executed by each stacker, as well as the storage location corresponding to each inbound and outbound task when the termination condition is met;

[0021] The memetic algorithm determines the initial outbound task sequence and stacker allocation through population initialization in global optimization, and obtains the inbound task sequence and stacker allocation by solving an assignment problem model with the goal of minimizing the matching metric of inbound and outbound tasks;

[0022] In the memetic algorithm, local search is used to further optimize the inbound task sequence and outbound task sequence; the local search includes two location exchange operators, which respectively represent exchanging the inbound location and outbound location of two inbound and outbound trips on a stacker. During each local search, one of them is selected for local search according to the historical performance of the two location exchange operators to form a new inbound and outbound trip.

[0023] The electronic device of the present invention includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is loaded into the processor, the ASRS task scheduling and location allocation method under classified storage described above is implemented.

[0024] The computer-readable storage medium of the present invention stores a computer program, characterized in that when the computer program is executed by a processor, the ASRS task scheduling and location allocation method under classified storage described in any one of the above is implemented.

[0025] Beneficial effects: Compared with the prior art, the advantages of the present invention are as follows: (1) Considering the dynamic allocation of locations and classified storage, a dynamic programming model combined with an integer programming model is proposed to describe the problem, and the empty locations generated by outbound tasks can be used for subsequent inbound operations; (2) By combining global search with a local search strategy based on an inbound and outbound location exchange operator, the algorithm can search in the neighborhood space where better solutions are generated, reducing the completion time of all tasks and the delay time of outbound tasks as a whole; (3) According to the characteristics of shelf zoning, a new Hamming distance is used as the matching metric for inbound and outbound tasks, and the inbound task sorting sub-problem is transformed into a classical assignment problem for solution, reducing the complexity of solving the task sorting problem, avoiding the formation of poor trips, and achieving fast calculation; (4) Integrating the inbound and outbound task scheduling problem and the location allocation problem into a dynamic programming model to achieve the integrated optimization of the two problems. Description of the Drawings

[0026] Figure 1 It is a top view of an automated storage and retrieval system in an embodiment of the present invention.

[0027] Figure 2 It is a front view of a shelf in an embodiment of the present invention.

[0028] Figure 3 It is a flowchart of the inbound and outbound task scheduling and location allocation method of the present invention.

[0029] Figure 4 Schematic diagram of the storage location exchange operator of the present invention.

[0030] Figure 5 Schematic diagram of the outbound location exchange operator of the present invention.

[0031] Figure 6 Comparison chart of the solution quality of the method of the present invention with discrete ICA and FCFS in the embodiment of the present invention.

[0032] Figure 7 Comparison chart of the average calculation time of the method of the present invention with discrete ICA in the embodiment of the present invention. Detailed implementation manners

[0033] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0034] As Figure 1 shown, the research object of the ASRS task scheduling and location allocation method and system under classified storage of the present invention is an ASRS adopting a classified storage strategy. There is a roadway between every two rows of shelves in the ARRS, and a stacker crane runs on each roadway. The stacker crane can only carry one unit of goods at a time, and its inlet and outlet (I / O) are located at the front end of each roadway. In this embodiment, each shelf is divided into four areas, and the areas can also be divided according to actual needs, such as Figure 2 shown. The stacker crane executes the inbound and outbound tasks in a cross-loading mode, that is, two operations of inbound and outbound are completed in one trip, and each trip is an inbound and outbound trip. For the inbound operation, first unload the incoming inbound goods and temporarily store them in the inbound buffer area. The forklift transports the goods to the assigned roadway conveyor belt station, and the handling robot responsible for this roadway places them on the automatic conveyor belt and transports them to the I / O port, and then the corresponding stacker crane executes the internal inbound operation. On the contrary, for the outbound operation, the stacker crane takes out the outbound goods from the corresponding storage location, places them on the conveyor belt at the I / O port, and transports them to the general shipping port to complete the outbound. The outbound task is driven by the customer order and usually must be completed before a given deadline to ensure that the truck can depart within the specified time window in subsequent transportation and distribution links. The problem of the present invention can be represented by the triple symbol: [F, para|IO 2 , zone|∑C i , ∑T i , where para represents multiple parallel stacker cranes.

[0035] In an ASRS adopting a classified storage strategy, (1) due to the constraints of the storage area for inbound and outbound tasks, the sorting of inbound and outbound tasks plays a crucial role in improving the operating efficiency of the stacker crane and reducing the task delay time. When the execution order of inbound and outbound tasks is unreasonable, the operating time of the stacker crane will become longer, resulting in an increase in the task completion time and delay time. To improve the inbound and outbound efficiency and customer satisfaction, the present invention not only considers sorting the outbound tasks but also optimizes the order of inbound tasks. (2) The present invention considers an ASRS with multiple aisles. There are goods to be shipped out and empty storage locations on each shelf, that is, inbound and outbound tasks can be assigned to any stacker crane for processing. Therefore, the optimal stacker crane allocation for inbound and outbound tasks needs to be considered. (3) There are multiple optional storage locations for inbound and outbound goods within their respective areas. Therefore, it is necessary to allocate suitable storage locations for inbound and outbound tasks. To make better use of the storage space, the present invention considers the dynamic allocation of storage locations and allows the reuse of the empty locations generated by outbound operations.

[0036] In summary, the present invention decomposes the problem into the following four sub-problems: the inbound task sorting problem, the outbound task sorting problem, the stacker crane allocation problem for inbound and outbound tasks, and the storage location selection problem for inbound and outbound tasks. The first two sub-problems achieve the pairing of inbound and outbound tasks. These four sub-problems are interrelated and interact with each other, making the problem-solving more complex. To clarify the problem, the present invention makes the following assumptions:

[0037] ● There is no inventory shortage in the warehouse, and to ensure the formation of inbound and outbound trips, the number of inbound and outbound tasks is equal.

[0038] ● There are no goods of the same type in the inbound and outbound task queues. If there are, they can be directly taken out from the inbound queue.

[0039] ● The partition of the shelf and the initial state of the storage location are known in advance, and the storage area to which the inbound and outbound tasks belong and the deadline of the outbound tasks are known in advance.

[0040] ● The stacker crane can move vertically and horizontally simultaneously at a constant speed. The time for it to execute the inbound and outbound trips is calculated using the Chebyshev metric method commonly used in the literature, and the time for it to perform the access operation at the storage location does not affect the optimization result.

[0041] It is negligible.

[0042] As Figure 3 shown, the method for ASRS task scheduling and storage location allocation under classified storage includes the following steps:

[0043] (1) Establish a mathematical model for ASRS task scheduling and storage location allocation

[0044] The present invention considers the dynamic allocation of storage locations, that is, the empty storage locations generated by the outbound tasks can be used for subsequent inbound operations. This storage location allocation method will cause the set of empty storage locations and the set of storage locations containing outbound goods, which are used as input parameters of the model, to change with each storage location selection decision. It is difficult for a single integer programming model to completely describe the problem of the present invention. The present invention introduces a dynamic programming model combined with integer programming to model the problem. Among them, the dynamic programming model depicts the allocation, sorting of stacker for outbound tasks and the state transition of the storage location set, and the integer programming model depicts the inbound and outbound storage location selection decisions in the static state. The dynamic programming model divides stages according to the number of outbound tasks, that is, each execution of an inbound and outbound travel operation can be regarded as a stage. In each stage, based on the current state of the storage location set, by solving an integer programming model with the goal of minimizing the current task completion and delay time, storage locations are selected for the inbound and outbound tasks, and the completion time of the inbound and outbound travel and the delay time of the outbound tasks are calculated. When a single stage ends, the state of the storage location set is updated, and then it transfers to the next stage until all tasks are completed.

[0045] The stages in the dynamic programming model are divided according to the number of outbound tasks. n outbound tasks represent n stages. The state variables of the dynamic programming model are C mb , and the decision variables are The state transition equation is:

[0046]

[0047]

[0048]

[0049] The recurrence formula of the optimal value function of the dynamic programming model is:

[0050]

[0051]

[0052]

[0053] Among them, f b (r, S) represents the minimum objective value when the r-th outbound task is arranged after the set S containing b tasks; r = 1, 2,..., n; b = 1, 2,..., n - 1; is the boundary condition, representing the minimum objective value when the r-th outbound task is arranged after the empty set ; r = 1, 2,..., n.

[0054] In formula (4), represents that in the state of C mbThe completion time and delay time of the r-th inbound and outbound trip, the values of which are calculated by the following integer programming model, where C mb is an input parameter of the integer programming model.

[0055] The integer programming model is:

[0056] Minimize Obj r = ω·C m(b+1) +(1 - ω)·TT r (6)

[0057]

[0058]

[0059]

[0060]

[0061]

[0062]

[0063] Formula (6) is the objective function, that is, to minimize the completion time and delay time of the r-th inbound and outbound trip; Constraint (7) ensures that only one inbound and outbound storage location is selected to perform the inbound and outbound trip operation; Constraints (8) and (9) ensure that the goods to be stored are placed in an empty location within the area to which the task belongs, and the outbound storage location must be a storage location within the area to which the task belongs and containing the outbound goods; Constraint (10) determines the inbound and outbound trip time, the value of which is calculated by the Chebyshev formula; Constraint (11) determines the completion time of the inbound and outbound trip; Constraint (12) represents the delay time of the outbound task.

[0064] (2) Memetic algorithm solution model

[0065] For the sub-problem of optimizing the inbound task sequence, on the premise of knowing the outbound task sequence, a new inbound and outbound matching index based on the Hamming distance is constructed, and this sub-problem is transformed into an assignment problem. The inbound task sequence is optimized by solving the assignment problem. For the sub-problems of outbound task sorting, allocation and storage location selection, the memetic algorithm is used for optimization and solution.

[0066] (2.1) Inbound task sorting

[0067] On the premise of a given outbound task sequence, the present invention optimizes the inbound task sequence by considering the matching degree of inbound and outbound tasks to reduce the traveling distance of the stacker. The present invention determines the initial outbound task sequence and stacker allocation through population initialization, considers the characteristics of the storage areas with classified storage, and introduces a new matching metric for inbound and outbound tasks based on the Hamming distance, which can effectively improve the matching efficiency.

[0068] The Hamming Distance is a way of distance measurement. That is, for two strings of equal length, the number of different characters at the same positions is the Hamming distance. In an ASRS adopting a classified storage strategy, the stacker takes less time to perform the inbound and outbound trips for the same or similar storage areas. Therefore, the matching degree of tasks for the same or similar storage areas is high, while the matching degree of tasks for storage areas far apart is low. The traditional Hamming distance uses the exclusive OR operation and cannot reflect the matching degree between storage areas under classified storage. Therefore, the present invention adaptively modifies the Hamming distance calculation method, defines the degree of proximity between the storage area and the I / O port by the rank of the storage area, and thus converts the storage area type into ordinal data that can reflect the degree of difference. Accordingly, the new Hamming distance for matching inbound and outbound tasks based on the rank of the storage area is defined as follows:

[0069] Definition (new Hamming distance H for matching inbound and outbound tasks based on the rank of the storage area under classified storage) The storage areas of the shelf are divided into p categories of storage areas in order from near to far from the I / O port. Ω represents the storage area sequence, Ω = (Ω1, Ω2...Ω p ), and the order of proximity of the storage area to the I / O port is defined as the rank of the storage area, that is, Rank(Ω p ) = p, and the greater the rank of the storage area is, the farther it is from the I / O port. Let the storage area sequence of the outbound task be Let the storage area sequence of the inbound task be Then the Hamming distance H for matching the two groups of inbound and outbound tasks is calculated as in formula (13):

[0070]

[0071] The Hamming distance metric under classified storage can quickly evaluate the matching degree of inbound and outbound tasks, and thus the sub-problem of inbound task sorting can be converted into an assignment problem. Among them, the weight degree of assignment between tasks is measured based on the H index, so as to construct an assignment problem model with the goal of minimizing The assignment problem model is an integer programming model for inbound sorting:

[0072]

[0073]

[0074]

[0075] x ro ∈ {0, 1}^r, o = 1, 2, ..., n (17)

[0076] Equation (14) is the objective function, Constraint (15) indicates that only one outbound task can be matched to one inbound task, Constraint (16) indicates that only one inbound task can be matched to one outbound task, and Constraint (17) is a binary decision variable, x ro being 1 indicates that the r-th outbound task is paired with the o-th inbound task to form an inbound-outbound journey, otherwise it is 0. The solution result of the model can be converted into a sequence of inbound tasks with the optimal inbound-outbound journey matching degree.

[0077] Taking 6 outbound and 6 inbound tasks and 4 storage areas as an example, the storage area sequence Ω = (A, B, C, D), then Rank(Ω) = (1, 2, 3, 4). The sequence of storage areas to which the outbound tasks belong The initial sequence of storage areas to which the inbound tasks belong At this time, the Hamming distance of the inbound-outbound journey matching of the inbound-outbound tasks is:[[]] By solving the above integer programming model for inbound sorting to optimize the order of inbound tasks, the sequence of storage areas of inbound tasks with the optimal inbound-outbound journey matching is obtained as Its Hamming distance is:[[]]

[0078] (2.2) Solving the model by combining global search and local search

[0079] The genetic algorithm is used as the global search strategy, and a local search strategy based on the exchange of inbound and outbound storage locations is proposed according to the problem characteristics.

[0080] The global search strategy includes:

[0081] ① Chromosome coding and population initialization

[0082] A complete coding scheme includes the allocation of stacker cranes for outbound tasks and their processing order. In the present invention, the outbound tasks are numbered in the original order as 1, 2...n, and the chromosome coding is set as an integer sequence composed of non-repeating numbers from 1 to n, and then z - 1 numbers greater than n are evenly inserted to represent the allocation of stacker cranes (z is the number of stacker cranes). This coding strategy ensures that the number of tasks processed by each stacker crane is balanced, improving the algorithm efficiency. For example, the coding of a solution for 6 outbound tasks and 2 stacker cranes can be expressed as: [2 4 5 7 1 6 3]. This scheme means that the outbound tasks numbered 2, 4, and 5 are allocated to the first stacker crane for processing in this order, and the outbound tasks numbered 1, 6, and 3 are allocated to the second stacker crane for processing in this order. To ensure the diversity of the population, the initial population of the algorithm is generated randomly. At the same time, to give the algorithm a good initial solution, a solution generated by the EDD (Early Due Date, EDD) rule is inserted into the population.

[0083] ② Dynamic storage location allocation under classified storage

[0084] The present invention proposes an effective inbound and outbound storage location allocation strategy, which selects storage locations for inbound and outbound tasks according to the chromosome coding and the inbound task sequence. First, the outbound storage location closest to the I / O port and containing the outbound goods is selected within the storage area to which the outbound task belongs, and then an empty storage location that can minimize the inbound and outbound travel time is selected within the storage area to which the inbound task belongs. Once the inbound and outbound storage locations are determined, the inbound and outbound travel time and the delay of the outbound task can be calculated. After the storage locations are allocated for a group of inbound and outbound trips, the status of the storage locations is updated, and the storage locations for subsequent inbound and outbound trips are selected according to this strategy.

[0085] ③ Fitness evaluation and population management strategy

[0086] The objective function of the present invention is to minimize the completion time of inbound and outbound tasks and the delay time of outbound tasks. Therefore, the reciprocal of the objective function is used as the fitness function, as defined in formula (18), where pop is the population size, C m is the completion time of all tasks on stacker crane m, and TT r is the delay time of the rth outbound task. The present invention adopts the elite retention strategy, retaining some excellent chromosomes in the parent population and merging them with some excellent chromosomes in the offspring population to form a new population.

[0087]

[0088] ④ Crossover and mutation operations

[0089] The present invention adopts the roulette wheel strategy to select suitable individuals from the population for crossover and mutation operations. The probability Prob of an individual being selected pCalculated by formula (19), and the crossover and mutation operations adopt the partially mapped crossover operator and the insertion operator respectively.

[0090]

[0091] The local search strategy includes:

[0092] The present invention proposes two storage location exchange operators to perform local search on the initial population, the individuals after crossover and mutation respectively to further optimize the solution. In each local search, an adaptive selection mechanism is used to select one from the two operators LO N , N ∈ {1, 2} for local search. Let the counter c N be used to record the historical performance of the two operators. The initial values of all counters are set to 1. When the solution after local search using a certain operator is better than the current solution, the counter of this operator is incremented by one. The present invention uses the roulette wheel selection strategy to select the operator, that is, the better the historical performance of the operator, the greater the probability of being selected. The operator selection formula is: P(N) = c N / ∑ N∈{1，2} c N .

[0093] The specific descriptions of the two operators are as follows:

[0094] (1) Inbound storage location exchange operator LO1: Exchange the inbound storage locations of two inbound and outbound trips on a stacker, as Figure 4 shown. Such an exchange can change the inbound task order, and the outbound task order remains unchanged, thus forming a new inbound and outbound trip, which can further optimize the inbound task sequence.

[0095] (2) Outbound storage location exchange operator LO2: Exchange the outbound storage locations of two inbound and outbound trips on a stacker, as Figure 5 shown. Such an exchange can change the outbound task order, and the inbound task order remains unchanged, thus forming a new inbound and outbound trip, which can further optimize the outbound task sequence.

[0096] Since the present invention adopts dynamic storage location allocation, there may be a precedence relationship between two storage locations in the solution, that is, the outbound storage location of an inbound and outbound trip may be the inbound storage location of a subsequent inbound and outbound trip. If the order of these two storage locations is changed, an infeasible solution will be caused. To improve the algorithm efficiency, the local search of the present invention is only executed within the feasible region of the solution.

[0097] (3) After the termination condition is satisfied, output the outbound and inbound tasks and their processing sequences that each stacker is responsible for executing, as well as the optimal solution of the storage location corresponding to each inbound and outbound task. The termination condition is reaching the maximum number of iterations.

[0098] The symbol descriptions are shown in the following table:

[0099]

[0100]

[0101] The method of the present invention is verified by experiments below.

[0102] In this embodiment, the size of the shelf is set to 10×10, and the shelf partition is as Figure 2 shown. The number of aisles in the warehouse is set to 3, and there are a total of 100×6 = 600 storage locations; at the same time, the utilization rate of each storage area on each shelf is 80%, each storage area contains 5 types of goods, and the status of each storage area is randomly generated; the total number N of inbound and outbound tasks is set to 40, 80, 120, where 40% of the outbound tasks are goods in area A, 30% are goods in area B, 20% are goods in area C, and 10% are goods in area D. The deadline of the outbound task is uniformly generated within the interval

[0103] [min{p r |r∈R}, (2·(1 - γ)·∑ r∈R p r + min{p r |r∈R}) / z], where p r is the inbound and outbound travel time of the task, and its value is calculated using the storage location selection strategy according to the initial inbound and outbound sorting. γ describes the tightness of the deadline, and the smaller it is, the looser the deadline. In this embodiment, γ is set to {0.6, 0.7, 0.8}.

[0104] There are two termination conditions: ① the maximum number of iterations, which is set to 100 in this experiment; ② the maximum number of iterations without improvement of the algorithm, which is set to 50 in this experiment.

[0105] In this embodiment, the commonly used FCFS algorithm in enterprises and the discrete ICA algorithm (Discrete Imperialist Competitive Algorithm) for solving the task scheduling problem of automated stereoscopic warehouses in the prior art are used as comparison schemes. The FCFS strategy evenly distributes tasks to the stacker according to the order of task arrival, that is, it does not consider the sorting of inbound and outbound tasks and the allocation of stackers; the comparison results of the algorithm solution quality under each example combination (γ_N) are as Figure 6 shown, where CT is the task completion time, Tar is the task delay time, Obj is the target value,

[0106] GAP1 = 100%*(Alg_CT - Matheuristic_CT) / Matheuristic_CT

[0107] GAP2 = 100% * (Alg_Tar - Matheuristic_Tar) / Matheuristic_Tar

[0108] GAP3 = 100% * (Alg_Obj - Matheuristic_Obj) / Matheuristic_Obj

[0109] Matheuristic refers to the method described in the present invention, and Alg represents the discrete ICA or FCFS algorithm.

[0110] The average calculation time of the method described in the present invention and the discrete ICA under each example combination (γ_N) is as Figure 7 shown.

[0111] From Figure 6 and Figure 7 it can be obtained that: (1) The present invention is superior to the discrete ICA in terms of solution quality, and the improvement amount increases with the increase in the number of tasks, with an average improvement amount of 16.63%. It is superior to the discrete ICA and FCFS in terms of task completion time and task delay. The reason is that the present invention optimizes the inbound task sequence by introducing a Hamming distance metric for inbound and outbound task matching based on the storage area rank, and combines a global and local search method to pair the inbound and outbound tasks. Compared with the "hitchhiking" strategy for inbound task sequence optimization in ICA, the present invention combines the characteristics of classified storage, which can avoid the formation of poor inbound and outbound trips, thereby reducing the task execution time and delay time. At the same time, the global and local search method in the present invention can search within the neighborhood space of a better solution by combining a global search strategy based on a genetic algorithm and a local search strategy based on an inbound and outbound storage location exchange operator.

[0112] (2) The calculation time-consuming of the present invention is significantly lower than that of the discrete ICA. The reason is that, compared with the discrete ICA using the distance between specific storage locations as a matching index, the present invention can achieve fast calculation by sorting the inbound tasks by considering the inbound and outbound task matching index based on the storage area rank. At the same time, although Gurobi is used to solve the integer programming model in this embodiment, during the implementation of its algorithm, pre-calculation of the matching index is adopted to avoid multiple loop calculations of the integer programming model, reducing the overall time complexity and calculation time-consuming of the algorithm.

[0113] Based on the same inventive concept, the ASRS task scheduling and storage location allocation system under classified storage described in the present invention includes:

[0114] A model establishment unit is configured to establish a dynamic programming model for the allocation of stacker cranes for inbound and outbound tasks, the sorting of inbound tasks, the sorting of outbound tasks, and the state transition of the storage location set. The empty storage locations generated by outbound tasks can be used for subsequent inbound operations. Each trip of the stacker crane is an inbound and outbound trip, including inbound tasks and outbound tasks. The storage rack is partitioned according to the characteristics of the goods, and the characteristics of the goods include the frequency of inbound and outbound of the goods and the attributes of the goods.

[0115] A model solving unit is configured to use a memetic algorithm to solve the dynamic programming model with the goal of minimizing the completion time of inbound and outbound trips and the delay time of outbound tasks. When the termination condition is met, it outputs the optimal solutions for the outbound and inbound tasks and the processing order that each stacker crane is responsible for executing, as well as the storage locations corresponding to each inbound and outbound task.

[0116] In the memetic algorithm, during global optimization, the initial outbound task order and stacker crane allocation are determined through population initialization. By solving an assignment problem model with the goal of minimizing the matching metric index of inbound and outbound tasks, the inbound task order and stacker crane allocation are obtained.

[0117] In the memetic algorithm, local search is used to further optimize the inbound task order and the outbound task order. The local search includes two storage location exchange operators, which respectively represent exchanging the inbound storage location and the outbound storage location of two inbound and outbound trips on a stacker crane. During each local search, one of them is selected for local search according to the historical performance of the two storage location exchange operators to form a new inbound and outbound trip.

[0118] Based on the same inventive concept, the electronic device of the present invention includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is loaded into the processor, it implements the ASRS task scheduling and storage location allocation method under classified storage described above.

[0119] Based on the same inventive concept, the computer-readable storage medium of the present invention stores a computer program, characterized in that when the computer program is executed by a processor, it implements the ASRS task scheduling and storage location allocation method under classified storage described above.

[0120] The computer-readable storage medium may include RAM, ROM, EEPROM, CD-ROM, or other optical disc storage devices, magnetic disk storage devices, or other magnetic storage devices, flash memory, or any other medium that can store the required program code in the form of instructions or data structures and can be accessed by a computer.

[0121] The processor is used to execute the computer program stored in the memory to implement each step in the methods described in the above embodiments.

Claims

1. An ASRS task scheduling and storage location allocation method under classified storage, characterized in that The shelves are partitioned according to the characteristics of goods, where the characteristics of goods include the frequency of goods in and out of storage and the attributes of goods; the model of in and out of storage task scheduling and storage location allocation is a dynamic programming model that includes the allocation of stacker cranes for in and out of storage tasks, the sorting of inbound tasks, the sorting of outbound tasks, and the state transition of the set of storage locations; the empty storage locations generated by outbound tasks can be used by subsequent inbound tasks; each trip of the stacker crane is an in and out of storage trip, including inbound tasks and outbound tasks; the method for in and out of storage task scheduling and storage location allocation is as follows: With the goal of minimizing the completion time of in and out of storage trips and the delay time of outbound tasks, a memetic algorithm is used to solve the dynamic programming model. When the termination condition is met, the optimal solution of the outbound and inbound tasks and the processing sequence responsible for each stacker crane, as well as the storage location corresponding to each in and out of storage task, is output; In global optimization, the memetic algorithm obtains the inbound task sequence and stacker crane allocation by solving an assignment problem model with the goal of minimizing the matching metric index of in and out of storage tasks based on the initial outbound task sequence and stacker crane allocation; In the memetic algorithm, local search is used to further optimize the inbound task sequence and the outbound task sequence; the local search includes two storage location exchange operators, which respectively represent exchanging the inbound storage location and the outbound storage location of two in and out of storage trips on a stacker crane. In each local search, one of them is selected for local search according to the historical performance of the two storage location exchange operators to form a new in and out of storage trip; The matching metric index for outbound tasks and inbound tasks is the new Hamming distance, and the formula for the new Hamming distance is: Among them is the outbound task storage area sequence is the inbound task storage area sequence, a i ∈Ω, b i ∈Ω, Ω represents the storage area sequence. The storage areas are divided into p categories of storage areas in order from near to far from the I / O port. Ω = (Ω1, Ω2…Ω p ); The method for using local search to further optimize the inbound task sequence and the outbound task sequence is as follows: Perform local search on the initial population in the meme algorithm, as well as the individuals after crossover and mutation. When performing local search each time, use P(N) = c N / ∑ N∈{1,2} c N Select a slot exchange operator to perform local search, where N ∈ {1, 2}, representing two slot exchange operators, P(N) is the probability that the slot exchange operator is selected, and c N is a counter. The counter is used to record the historical performance of the slot exchange operator. When the solution after local search using a certain slot exchange operator is better than the current solution, the counter is incremented by one.

2. The ASRS task scheduling and storage location allocation method under classified storage according to claim 1, characterized in that The dynamic programming model divides stages according to the number of outbound tasks. Each stage, based on the current state of the set of storage locations, selects storage locations for in and out of storage tasks by solving an integer programming model with the goal of minimizing the completion time of in and out of storage trips and the delay time of outbound tasks; after a single stage ends, the state of the set of storage locations is updated and transferred to the next stage until all in and out of storage tasks are completed.

3. The ASRS task scheduling and storage location allocation method under classified storage according to claim 1, characterized in that, In the memetic algorithm, a genetic algorithm is used for global search, including: generating the initial outbound task sequence and stacker crane allocation by chromosome encoding for the outbound task sequence and stacker crane allocation; Selecting storage locations for in and out of storage tasks according to the chromosome encoding, the inbound task sequence, and the stacker crane allocation; selecting the outbound storage location closest to the I / O port and containing the outbound goods in the storage area where the outbound task belongs, and selecting an empty storage location in the storage area where the inbound task belongs that can minimize the completion time of in and out of storage trips; after allocating storage locations for an in and out of storage trip, update the storage location state until all storage locations are allocated.

4. The ASRS task scheduling and storage location allocation method under classified storage according to claim 3, characterized in that In the memetic algorithm, the reciprocal of the objective function with the goal of minimizing the completion time of in and out of storage trips and the delay time of outbound tasks is used as the fitness function, and the optimal solution of the memetic algorithm is the solution that maximizes the fitness function value when the termination condition is met.

5. An ASRS task scheduling and storage location allocation system under classified storage, characterized in that, Including: A model establishment unit for establishing a dynamic programming model for the allocation of stackers for inbound and outbound tasks, the sorting of inbound tasks, the sorting of outbound tasks, and the state transition of the set of storage locations. The empty storage locations generated by outbound tasks can be used for subsequent inbound operations. Each trip of the stacker is an inbound and outbound trip, including inbound tasks and outbound tasks. The shelves are partitioned according to the characteristics of the goods, and the characteristics of the goods include the frequency of inbound and outbound of the goods and the attributes of the goods. A model solving unit for solving the dynamic programming model by using a memetic algorithm with the goal of minimizing the completion time of inbound and outbound trips and the delay time of outbound tasks. When the termination condition is met, it outputs the optimal solution of the outbound and inbound tasks and the processing order responsible for each stacker, as well as the storage location corresponding to each inbound and outbound task. In the memetic algorithm, the initial outbound task order and stacker allocation are determined through population initialization in global optimization. By solving an assignment problem model with the goal of minimizing the matching metric index of inbound and outbound tasks, the inbound task order and stacker allocation are obtained. In the memetic algorithm, local search is used to further optimize the inbound task order and the outbound task order. The local search includes two storage location exchange operators, which respectively represent exchanging the inbound storage location and the outbound storage location of two inbound and outbound trips on a stacker. In each local search, one of them is selected for local search according to the historical performance of the two storage location exchange operators to form a new inbound and outbound trip. The matching metric index for outbound tasks and inbound tasks is the new Hamming distance, and the formula for the new Hamming distance is: Among them is the outbound task cargo area sequence is the inbound task cargo area sequence, a i ∈Ω, b i ∈Ω, Ω represents the cargo area sequence. The cargo areas are divided into p categories of cargo areas in order from near to far from the I / O port. Ω = (Ω1, Ω2…Ω p ); The method for further optimizing the inbound task order and the outbound task order by using local search is: Perform local search on the initial population in the memetic algorithm, as well as the individuals after crossover and mutation. Use P(N) = c during each local search N / ∑ N∈{1,2} c N Select a slot exchange operator for local search, where N ∈ {1, 2}, representing two slot exchange operators, P(N) is the probability that the slot exchange operator is selected, and c N is a counter. The counter is used to record the historical performance of the slot exchange operator. When the solution after local search using a certain slot exchange operator is better than the current solution, the counter is incremented by one.

6. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the computer program is loaded into the processor, it implements the ASRS task scheduling and storage location allocation method under classified storage according to any one of claims 1-4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the ASRS task scheduling and storage location allocation method under classified storage according to any one of claims 1-4.

Images