Port three-dimensional warehousing system pulsation access toughness scheduling method

Abstract

The invention discloses a pulsation access toughness scheduling method for a port three-dimensional warehousing system. The pulsation access toughness scheduling method is used for the warehousing system with a double-depth storage shelf, a shuttle vehicle, an elevator and a control system. Comprising the steps that a control system sends out a warehouse task instruction, and a shuttle vehicle receives the instruction according to the first-arrival first-service principle; judging whether the shuttle vehicle is idle or not, if so, driving to the column where the target goods are located, otherwise, waiting; judging whether to pour goods after arriving at the target column, and if goods need to be poured, firstly pouring goods, then taking target goods and requesting for an elevator, otherwise, directly taking goods and requesting for the elevator; whether the elevator is idle or not is judged, if yes, target goods are loaded, and if not, waiting is conducted; and the hoister loads and returns to the I / O point of the system for unloading, and the ex-warehouse task is And the goods pouring operation comprises the steps of firstly moving the blocked goods to the unoccupied goods allocation, then taking the target goods to the hoister, and finally transporting the blocked goods back to the original goods allocation. The warehouse-out efficiency of the warehousing system can be improved, the waiting time is shortened, and the system toughness and the operation stability are improved.

Description

Technical Field

[0001] The present invention relates to a scheduling method, and more specifically to a pulsating access resilience scheduling method for a port three-dimensional warehousing system. Background Art

[0002] As a new type of intelligent three-dimensional warehouse integrating warehousing management and sorting and distribution functions, the double-depth multi-layer shuttle car warehousing system has high automation and high space utilization rate, and can realize the rapid access of small-piece, multi-variety and small-batch goods. Therefore, it has received extensive attention from logistics enterprises such as ports and has been applied to the access of bulk goods. However, the terminal loading and unloading operations have pulsating characteristics. The ship docking process is complex and delicate, and it takes a long time. The window period for loading and unloading operations after the ship docks is very precious. Therefore, it is necessary to carry out goods transfer operations during the non-loading period to improve the loading efficiency.

[0003] In the warehousing system of double-depth storage racks, the goods transfer operation will increase the operation distance and time, resulting in an increase in the complexity of operation task scheduling. In addition, in the research on the operation task scheduling of double-depth storage racks, although the goods transfer operation problem is considered, only a rough estimate of the goods transfer operation time is made, and the goods transfer operation process is not analyzed in depth. There is a lack of a corresponding mathematical model for the goods transfer operation, which has a certain deviation from the actual automated three-dimensional warehousing system scheduling problem. Therefore, in-depth research on the goods transfer operation process of the double-depth multi-layer shuttle car warehousing system and the establishment of an outbound operation model under different goods transfer strategies to optimize the task scheduling of the double-depth multi-layer shuttle car warehousing system have important theoretical significance and practical value.

[0004] Therefore, there is currently a patent for invention with the application number 202411646109.2 that discloses an intelligent warehousing management method. For a multi-layer shuttle car intensive three-dimensional warehouse, it predicts order information based on historical order data and calculates the storage locations of goods in the inbound orders and the corresponding handling equipment paths. It uses an optimization algorithm to calculate the globally optimal storage locations of goods, the goods handling allocation plan, and the handling equipment paths, and performs inbound scheduling on the goods in the inbound orders. However, it fails to carry out goods transfer operations in advance for large-scale tasks to cope with the pulsating impact similar to the ship loading operation tasks.

[0005] The patent for invention with the application number 202410969032.6 discloses a method, system, and equipment for configuring logistics equipment in a four-way shuttle car warehousing system. It establishes a completion time model for all instructions and a unit usage cost model for logistics equipment according to the warehousing system parameter information. It uses the NSGA-II algorithm that combines simulated annealing differential evolution search and problem domain knowledge to solve and configure the four-way shuttle cars and elevators. However, it can only achieve the basic inbound and outbound scheduling of logistics equipment. Summary of the Invention

[0006] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide a pulsating access resilience scheduling method for a port three-dimensional warehousing system. Based on the actual operation situation, three goods dumping operation strategies, namely, the random point goods dumping strategy, the nearest point goods dumping strategy and the fixed point goods dumping strategy, are proposed. The goods dumping operation process is analyzed and an accurate goods dumping operation time model is established. On the basis of considering the goods dumping operation, the outbound operation is scheduled and optimized. A double-population genetic algorithm is designed to solve the scheduling model, and a double-population recombination and cooperation optimization strategy is proposed to enable the two populations to exchange individuals in the population while evolving differently, breaking the population state and jumping out of the local optimum. In addition, a variable neighborhood search strategy is adopted to increase the diversity of the solution space and improve the search efficiency of the algorithm.

[0007] To achieve the above object, the present invention provides the following technical solutions: A pulsating access resilience scheduling method for a port three-dimensional warehousing system, which is used for a warehousing system with double-depth storage racks, shuttle cars, elevators and a control system, and is characterized in that it includes the following steps: Step 1: The control system issues an outbound task instruction, and the shuttle car receives the outbound task instruction according to the first-come-first-served principle. Step 2: Determine whether the shuttle car is in an idle state. If the shuttle car is in an idle state, the shuttle car travels to the column where the target goods are located; otherwise, wait for the shuttle car until it is in an idle state. Step 3: The shuttle car travels to the column where the target goods are located. Determine whether goods dumping operation is required. If so, first perform the goods dumping operation. After the goods dumping operation is completed, the shuttle car loads the target goods and requests the elevator to complete the remaining operation process. If goods dumping operation is not required, the shuttle car directly loads the target goods and requests the elevator to complete the remaining operation process. Step 4: Determine whether the elevator is in an idle state. If the elevator is in an idle state, the elevator loads the target goods; otherwise, wait for the elevator until it is in an idle state. Step 5: The elevator loads the target goods and returns to the system I / O point, unloads the target goods, and the outbound task ends. Among them, the process of the shuttle car performing the goods dumping operation is as follows: First, the shuttle car takes out the blocked goods and places them in the remaining empty storage locations, and this empty storage location is also called the goods dumping location. Secondly, the shuttle car then takes out the target goods and hands them over to the elevator to complete the vertical transportation of the target goods. Finally, the shuttle car transports the blocked goods from the goods dumping location to the original storage location.

[0008] As a further improvement of the present invention, the shuttle car goods unloading operation strategy in step three is a random goods unloading strategy, and the specific content of this random goods unloading strategy is as follows: The control system randomly selects an idle storage location in the lane where the target goods are located as the goods unloading location. The shuttle car first places the blocked goods at the goods unloading location, then continues to complete the horizontal operation of this outbound task. Finally, the shuttle car returns to this lane again to place the blocked goods at their original positions. Thus, this outbound operation task is completed. The goods unloading operation time of this random goods unloading strategy is the running time of the shuttle car between the goods unloading location and the target storage location, and it is calculated through the following formula: Wherein, \(t_{u}\) is the goods unloading operation time, \(d_{u}\) is the goods unloading operation distance, \(v\) is the average speed of the shuttle car, \(t_{l / u}\) is the loading / unloading time of the shuttle car, \(i\) is the shuttle car number, \(n\) is the number of storage columns in each row of the shelves, \(l\) is the length of a single storage location.

[0009] As a further improvement of the present invention, the shuttle car goods unloading operation strategy in step three is the nearest point goods unloading strategy, and the specific content of this nearest point goods unloading strategy is as follows: The control system selects an idle storage location closest to the target storage location in the lane where the target goods are located as the goods unloading location. The shuttle car first places the blocked goods at the goods unloading location, then continues to complete the horizontal operation of this outbound task. Finally, the shuttle car returns to this lane again to place the blocked goods at their original positions. Thus, this outbound operation task is completed; wherein, if there are two idle storage locations with the same distance to the target storage location at the same time, then select the idle storage location closer to the buffer area as the goods unloading location. The goods unloading operation time of this nearest point goods unloading strategy is calculated through the following formula: Wherein, \(t_{u}\) is the goods unloading operation time, \(d_{u}\) is the goods unloading operation distance, \(v\) is the average speed of the shuttle car, \(t_{l / u}\) is the loading / unloading time of the shuttle car, \(l\) is the length of a single storage location, \(\rho\) is the storage location occupancy rate.

[0010] As a further improvement of the present invention, the shuttle unloading operation strategy in step 3 is a fixed-point unloading strategy. The fixed-point unloading strategy is specifically as follows: the control system selects a fixed free cargo position as the unloading cargo position according to the lane where the target cargo is located, the shuttle car first places the blocked cargo at the unloading cargo position, and then continues to complete the horizontal operation of the outbound task, and finally the shuttle car returns to the lane to place the blocked cargo at the original position. At this point, the outbound operation task is completed. The unloading operation time of the fixed-point unloading strategy is calculated by the following formula: in, For the unloading operation time, is the unloading operation distance, is the average speed of the shuttle, is the shuttle loading / unloading time, Number the lane. for The fixed cargo unloading location coordinates of the aisle, The coordinates of the target location, The length of a single cargo space.

[0011] As a further improvement of the present invention, the total time of the outbound operation is also included. The calculation steps are as follows: outbound operation tasks, the total outbound operation time of the system It can be expressed as: in, is the number of outbound operation tasks, and Respectively indicate execution of The running time of the shuttle and elevator for each outbound task, Shuttle loading / unloading time, It is the loading / unloading time of the elevator.

[0012] As a further improvement of the present invention, the shuttle vehicle performs the The running time of the outbound task is calculated as follows: The coordinates of the outbound task are , indicating that the outbound task is located in Layer, Lane, The shuttle's running time is It consists of three parts: shuttle handling time, loading time and unloading operation time, which are specifically expressed as follows: Among them, is the decision factor for the goods transfer operation. When , it indicates that there is a goods transfer operation during the outbound task process; when , it indicates that there is no goods transfer operation during the outbound task process. is the goods transfer operation time, is the average speed of the shuttle vehicle, is the loading / unloading time of the shuttle vehicle, is the number of aisle columns, is the length of a single storage location.

[0013] As a further improvement of the present invention, the running time of the elevator when executing the -th outbound task is obtained by the following method: For the -th outbound task, the elevator operation time consists of the elevator transportation time and the loading time, as shown in the following formula: Among them, is the number of storage racks, is the height of a single-layer storage rack, is the average speed of the elevator, is the loading / unloading time of the elevator.

[0014] As a further improvement of the present invention, the steps one to five are executed by constructing a mathematical model. The construction process is as follows: Assign outbound tasks to s four-way shuttle vehicles and t elevators for execution. Set the outbound task set as , the four-way shuttle vehicle set as , the elevator set as , the task set of the z -th shuttle vehicle as , and the outbound operation time as . Establish the following mathematical model: The constraint conditions are as follows: Equation (8) indicates that the shuttle vehicle executes all outbound tasks; Equation (9) indicates that an outbound task can only be executed by one shuttle vehicle; Equation (10) indicates that each shuttle vehicle executes at least one outbound task.

[0015] As a further improvement of the present invention, the constructed mathematical model is solved using a double-population genetic algorithm. The specific process is as follows: Step 1: Adopt integer coding based on the outbound operation, design a three-segment chromosome coding method, where each chromosome represents a scheduling plan; the first segment is represented by the outbound task number, and the position where the outbound task number appears indicates the execution order of the outbound task; the second segment is represented by the number of outbound tasks executed by the shuttle car; the third segment is represented by the number of outbound tasks executed by the elevator. Step 2: Introduce two populations, denoted as population 1 and population 2, and perform initialization operations on the two populations using different initialization methods; for population 1, according to the constraint conditions, use the random generation method for initialization; for population 2, first randomly arrange the outbound task numbers to obtain segment 1, then arrange segment 2 according to the principle that the shuttle car and the goods are on the same layer to balance part of the operation time, and finally repeat the generation operation to obtain the initial population. Step 3: Define the fitness function, and then select suitable individuals to enter the next generation according to the fitness values obtained by calculating the fitness function. Step 4: For both population 1 and population 2, adopt the tournament selection strategy, randomly select 2 parent generations for fitness value comparison, and select the one with the higher fitness value to enter the next generation. Repeat the operation to obtain a population with a quantity of N. Step 5: Perform adaptive crossover and mutation. Step 6: Given the optimization factor , generate a random number. If the random number is less than the given optimization factor , then perform variable neighborhood search on the individual.

[0016] The beneficial effects of the present invention are as follows. In the double-depth multi-layer shuttle car storage system, there is generally a goods transfer operation, which leads to an increase in the inbound and outbound operation time and a decrease in the system operation efficiency. To address this problem, by analyzing the goods transfer operation process, three goods transfer operation strategies, namely the random point goods transfer strategy, the nearest point goods transfer strategy, and the fixed point goods transfer strategy, are proposed, and an accurate goods transfer operation calculation model is established. On this basis, an outbound operation scheduling optimization model is established with the goal of minimizing the outbound operation time. A double-population genetic algorithm is designed to solve the model. By introducing variable neighborhood search and double-population recombination and cooperation optimization strategies, the optimization ability of the algorithm is increased, the solution space is effectively increased, and the search performance of the algorithm is improved. The goods transfer operation calculation model can accurately evaluate the goods transfer operation time; in addition, the double-population genetic algorithm has higher optimization efficiency and can effectively shorten the outbound operation time. Description of the Drawings

[0017] Att Figure 1 is a schematic diagram of the double-depth multi-layer shuttle car storage system; Att Figure 2 is a schematic diagram of the goods transfer operation process; Att Figure 3 is a flowchart of the outbound operation; Att Figure 4It is the flow chart of the DPGA algorithm; Appendix Figure 5 It is the encoding and decoding process; Appendix Figure 6 It is the crossover operation process; Appendix Figure 7 It is three kinds of neighborhood transformation operations; Appendix Figure 8 It is the number of relocating operations under three kinds of relocating strategies; Appendix Figure 9 It is the convergence graph of the comparison algorithm for 60 outbound tasks. Specific implementation mode

[0018] Next, the embodiments given in conjunction with the accompanying drawings will be used to further elaborate on the present invention. The symbolic parameters used in this embodiment are shown in Table 1: Table 1 Main parameter symbols The method of this embodiment is mainly used for the existing double-depth multi-layer shuttle car storage system. This system consists of a double-depth storage rack, shuttle cars, elevators, and a control system, etc. Refer to Figures 1 to 2 As shown, ① is the elevator, ② is the shuttle car, ③ is the storage location, ④ is the I / O position, ⑤ is the idle location, ⑥ is the buffer area, ⑦ is the target location, ⑧ is the blocked location, and ⑨ is the relocating location. The double-depth storage rack is used for storing goods, and each storage location can store only one piece of goods. The horizontal transportation of goods is realized by the shuttle car, and the elevator is installed on the same side of the rack to realize the vertical transportation of goods or shuttle cars. The shuttle car and the elevator can only transport one piece of goods at a time. At the same time, several buffer areas are set in this storage system for the temporary storage of goods.

[0019] Compared with the storage system based on a single-depth storage rack, the double-depth multi-layer shuttle car storage system has a higher space utilization rate. Due to the uniqueness of the system structure, the system may have a relocating operation process during the inbound and outbound operations, which increases the inbound and outbound operation time and thus reduces the system operation efficiency. The relocating operation process is as shown in Appendix Figure 2 As shown, when the system issues an instruction to retrieve the target goods (marked as "A") located in the second column, it is blocked by the goods (marked as "B") located in the first column at this time. First, the shuttle car retrieves the blocked goods and places them in the remaining idle location, which is also called the relocating location; secondly, the shuttle car then retrieves the target goods and hands them over to the elevator to complete the vertical transportation of the target goods; finally, the shuttle car transports the blocked goods from the relocating location to the original location. The distance for the shuttle car to perform the relocating operation is called the relocating distance.

[0020] The inbound and outbound operations of the double-depth multi-layer shuttle car storage system are issued by the control system, and the shuttle car and the elevator cooperate to complete. The flow chart of the outbound operation is as shown in AppendixFigure 3 The specific outbound operation process is as follows: (1) The system issues a storage task instruction, and the shuttle receives the outbound task instruction based on the first-come-first-served principle; (2) If the shuttle is idle, the shuttle moves to the row where the target cargo is located; otherwise, the shuttle waits until it is idle; (3) The shuttle determines whether it is necessary to unload the cargo. If necessary, it will unload the cargo first. Otherwise, the shuttle loads the target cargo and requests the elevator to complete the remaining operation process. (4) If the elevator is in an idle state, the elevator loads the target cargo, otherwise wait for the elevator to be in an idle state; (5) The elevator loads the target goods and returns to the system I / O point, unloads the target goods, and the outbound mission is completed.

[0021] In the unloading process of this embodiment, a random unloading strategy, a nearest point unloading strategy and a fixed point unloading strategy are used. The time models of these three unloading strategies are as follows: Operation time model of random dumping strategy The random cargo transfer strategy (Rearrangement to a random point, RTRP) means that when a cargo transfer operation occurs in the system, the control system randomly selects an empty cargo position as the cargo transfer position according to the aisle where the target cargo is located, and the shuttle car places the blocked cargo at the cargo transfer position first, and then continues to complete the horizontal operation of the outbound task. Finally, the shuttle car returns to the aisle to place the blocked cargo at the original position, and the outbound operation task is completed. Therefore, the cargo transfer operation time under the random cargo transfer strategy is the running time of the shuttle car between the cargo transfer position and the target position.

[0022] According to the above description, the unloading operation time under the random unloading strategy is the time consumed by the shuttle vehicle to transfer the blocked goods between the unloading cargo position and the target cargo position. , the storage locations in the second row of the storage system are all occupied, that is, the unloading locations are located in the first row. Therefore, the unloading operation problem under the random unloading strategy can be transformed into the outbound operation problem in the single-depth storage shelf. The unloading operation time under this strategy is calculated as follows: in, For the unloading operation time, is the unloading operation distance, is the average speed of the shuttle, is the shuttle loading / unloading time, Number the shuttle. is the number of storage columns per row of shelves, The length of a single cargo space.

[0023] Operation time model of nearest point unloading strategy The Rearrangement to the nearest point (RTNP) strategy means that when the outbound operation task requires the outbound operation, the control system selects an idle cargo location closest to the target cargo location as the outbound cargo location according to the aisle where the target cargo is located. The shuttle car first places the blocked cargo at the outbound cargo location, and then continues to complete the horizontal operation of the outbound task. Finally, the shuttle car returns to the aisle to place the blocked cargo at the original location, and the outbound operation task is completed. If there are two idle cargo locations at the same distance from the target cargo location, the idle cargo location closer to the buffer area is selected as the outbound cargo location. When the cargo location occupancy rate in the system is When , the probability of dumping operation is ( ), the unloading operation time is shown in formula (2): in, For the unloading operation time, is the unloading operation distance, is the average speed of the shuttle, is the shuttle loading / unloading time, is the length of a single cargo space, is the cargo space occupancy rate.

[0024] Operation time model of fixed point unloading strategy Rearrangement to the fixed point (RTFP) means that when the outbound operation task requires the outbound operation, the control system selects a fixed free cargo position as the outbound cargo position according to the aisle where the target cargo is located. The shuttle car places the blocked cargo at the outbound cargo position first, and then continues to complete the horizontal operation of the outbound task. Finally, the shuttle car returns to the aisle to place the blocked cargo at the original position. At this point, the outbound operation task is completed. Assume that The fixed cargo unloading location coordinates of the lane are ( ), the coordinates of the target cargo location are ( ), the unloading operation time under this strategy can be expressed as: in, For the unloading operation time, is the unloading operation distance, is the average speed of the shuttle, is the loading / unloading time of the shuttle vehicle, is the roadway number, is the fixed transfer location coordinates of the roadway, the coordinates of the target location, is the length of a single location.

[0025] During the operation of the warehousing system, the overall outbound operation time plays a very important role. Therefore, the method of this embodiment provides a corresponding outbound operation time model, which is specifically as follows: Outbound operation time model In a double-depth multi-layer shuttle vehicle warehousing system, the operation time of the outbound task is determined by the specific number of tasks and is related to the location of the target goods. By analyzing the outbound operation process, it can be seen that the operation time of a single outbound task consists of the operation time of the shuttle vehicle, the operation time of the elevator, and the loading time of the equipment. For outbound operation tasks, the total system outbound operation time can be expressed as: Among them, is the number of outbound operation tasks, and respectively represent the running times of the shuttle vehicle and the elevator when executing the th outbound task, the loading / unloading time of the shuttle vehicle, is the loading / unloading time of the elevator.

[0026] Assume that the coordinates of the th outbound task are , indicating that this outbound task is located on the th floor, the th roadway, and the th column. Then the running time of the shuttle vehicle consists of three parts: the shuttle vehicle handling time, the loading time, and the transfer operation time, which are specifically expressed as follows: Among them, is the transfer operation decision factor. When , it indicates that there is a transfer operation during this outbound task process; when , it indicates that there is no transfer operation during this outbound task process. is the transfer operation time, is the average speed of the shuttle vehicle, is the loading / unloading time of the shuttle vehicle, is the number of roadway columns, is the length of a single storage location.

[0027] According to the operation process, for the th outbound task, the operation time of the elevator consists of the handling time and the loading time of the elevator, as shown in Equation (6): Among them, is the number of storage shelves, the height of a single-layer storage shelf, the average speed of the elevator, is the loading / unloading time of the elevator.

[0028] In a double-depth multi-layer shuttle car storage system, through a detailed analysis of the outbound operation process, the outbound operation scheduling problem of the double-depth multi-layer shuttle car storage system can be described as: Assign outbound tasks to s four-way shuttle cars and t elevators for execution, and set the outbound task set as , the four-way shuttle car set as , the elevator set as , the task set of the zth shuttle car as , and the outbound operation time as . To better describe the scheduling process of the outbound operation, the following mathematical model is established: The constraint conditions are as follows: Equation (8) means that the shuttle car executes all outbound tasks; Equation (9) means that an outbound task can only be executed by one shuttle car; Equation (10) means that each shuttle car executes at least one outbound task.

[0029] After obtaining the mathematical model in the above steps, it is necessary to solve the mathematical model to guide the subsequent selection of the goods transfer strategy and scheduling strategy. In this embodiment, when studying the outbound operation scheduling problem of a double-depth multi-shuttle storage system, it is difficult to find a suitable solution in the search space by conventional methods. In the traditional genetic algorithm, individuals can theoretically exchange information with any individual. As the algorithm iteratively runs, the information of the best individual quickly spreads globally, which is conducive to the dissemination of effective information, but it reduces the diversity of the population and the algorithm is prone to falling into local optimality. In the problem of optimizing job scheduling based on the goods transfer strategy, it is necessary to reduce the job time by adjusting the job task order and consider the precedence relationship of job tasks to reduce the goods transfer operation time, ultimately minimizing the inbound and outbound operation time. In view of the characteristics of this job scheduling optimization problem, a dual population genetic algorithm (DPGA) is designed to solve the model. Combining the problem characteristics, a three-section encoding based on job task allocation and the corresponding decoding scheme are designed. Considering the precedence and random initialization of job tasks to reduce a part of the goods transfer operation times. At the same time, to balance the job task time, a variable neighborhood search strategy is introduced to avoid the algorithm falling into local optimality during the search process, effectively increasing the solution space and improving the algorithm search efficiency.

[0030] Principle of the DPGA algorithm: To obtain a better initial solution, a dual population strategy is introduced. The genetic algorithm is used to perform genetic operations on the initial population to evolve the population. Then, the variable neighborhood search strategy is used to expand the range of solutions to avoid the algorithm falling into local optimality during the search process. Then, a new population is generated according to the recombination and cooperation of the dual populations. The algorithm flow is as follows Figure 4 shown.

[0031] The basic process of the DPGA algorithm is as follows: Step 1: Initialize relevant parameters, mainly including: population size , maximum number of iterations , self-adaptive adjustment parameter , and optimization factor etc.; Step 2: Generate initial populations with sizes of and respectively according to the constraint conditions and the same-layer rule; Step 3: Calculate the fitness of each individual in the two populations and sort them. According to the elitist retention strategy, select a certain proportion of the best individuals for retention; Step 4: Retain the individuals with high fitness in the population according to the tournament selection strategy; Step 5: Combine the three - segment encoding method and adopt the adaptive crossover and mutation operators to perform crossover and mutation operations on the population; Step 6: Adopt the variable neighborhood search strategy to generate new individuals and evaluate the new individuals; Step 7: Determine whether the current situation meets the algorithm termination condition. If it meets, end the algorithm process and output the optimal solution. Otherwise, perform the dual - population collaborative optimization operation to generate new population 1 and new population 2, and then go to Step 3. Therefore, the solution steps of the above - mentioned DPGA algorithm are as follows: (1) Encoding and decoding Adopt integer encoding based on the outbound operation, and design the three - segment chromosome encoding method as shown in the appendix Figure 5 Each chromosome represents a scheduling scheme. The first segment is represented by the outbound task number. The position where the outbound task number appears indicates the execution order of the outbound task. The second segment is represented by the number of outbound tasks executed by the shuttle vehicle. The third segment is represented by the number of outbound tasks executed by the elevator.

[0032] When the three - segment encodings are [4, 1, 5, 7, 3, 2, 6], [3, 2, 2] and [4, 3] respectively, after decoding, it can be obtained that the number of tasks executed by shuttle vehicle No. 1 is 3, and the execution order is 4 - 1 - 5; the number of tasks executed by shuttle vehicle No. 2 is 2, and the execution order is 7 - 3; the number of tasks executed by shuttle vehicle No. 3 is 2, and the execution order is 2 - 6; the number of tasks executed by elevator No. 1 is 4, and the execution order is 4 - 1 - 5 - 7; the number of tasks executed by elevator No. 2 is 3, and the execution order is 3 - 2 - 6.

[0033] (2) Population initialization The quality of population initialization has a direct impact on the solution results of the genetic algorithm. Introduce dual populations, denoted as population 1 and population 2, and perform initialization operations on the two populations using different initialization methods. For population 1, according to the constraint conditions, use the random generation method for initialization. For population 2, first randomly arrange the outbound task numbers to obtain segment 1, and then arrange segment 2 according to the principle that the shuttle vehicle and the goods are on the same layer, so as to balance part of the operation time. Finally, repeat the generation operation to obtain the initial population.

[0034] (3) Fitness evaluation Fitness evaluation first requires defining a fitness function, and then selecting suitable individuals to enter the next generation according to the fitness values. The fitness function is a statistical method for the fitness value of the job - scheduling scheme, and it is necessary to ensure that the fitness value is positive. Taking the minimum job time as the goal, use the defined fitness function to evaluate the job time, and select the reciprocal of the job time as the fitness value of the individual. When the job time is larger, the fitness function value is smaller. The fitness function is shown in Equation (11): (4) Selection operation For both population 1 and population 2, the tournament selection strategy is adopted. Arbitrarily select 2 parent individuals to compare their fitness values, and select the one with the higher fitness value to enter the next generation. Repeat the operation to obtain a population of size N.

[0035] (5) Adaptive crossover and mutation Crossover and mutation are operations to generate new individuals. To avoid the premature convergence and local convergence problems caused by the standard genetic algorithm with fixed crossover rate and mutation rate, an adaptive crossover and mutation operation method is proposed.

[0036] The calculation formulas for the adaptive crossover and mutation probabilities are shown in Eqs. (12) to (14). \(t\) is the current iteration number. \(T\) is the maximum iteration number. \(k\) is a variable related to the iteration number. \(f\) is the self - adaptive adjustment parameter. \(f_1 = \frac{T - t}{T}\) \(P_c\) is the adaptive crossover probability. \(P_m\) is the adaptive mutation probability. \(C\) is a constant value. \(P_c = f_1 \times \frac{\overline{f}}{\overline{f}_{max}}\) \(\overline{f}\) and \(\overline{f}_{max}\) are the average fitness value and the maximum fitness value of the current population respectively. \(f_{max}^c\) is the larger fitness of the parent individuals during the crossover operation. \(f_{max}^m\) is the larger fitness value of the parent individuals during the mutation operation.

[0037] Since the three - segment coding method is adopted, if the individuals are crossed in the way of the standard genetic algorithm, a large number of illegal solutions will be generated. Therefore, only segment 1 is crossed. As shown in the appendix Figure 6 Adopt the two - point crossover strategy. Randomly select two crossover points from the parent individuals, replace the task sorting in the gene segment between the two crossover points of parent 1 with the corresponding task sorting in parent 2 to obtain offspring 1, and perform the same operation on parent 2 to obtain offspring 2.

[0038] Adopt the single - point reverse mutation strategy. Randomly select a mutation point and reverse the gene segment after the mutation point. Since the mutation operation will cause a large change in the number of tasks executed by the shuttle car and the elevator, which is not conducive to further generating new individuals, therefore, only the task sequence is mutated.

[0039] (6) Variable neighborhood search In the search algorithm, the initial solution generates new solutions through a series of moves, and all the moves constitute the neighborhood. The structure of the neighborhood has a great impact on the search performance of the algorithm. To avoid the algorithm falling into local optimum during the search process, a variable neighborhood search strategy is designed, which includes three heuristic rules: swap neighborhood, insertion neighborhood, and inversion neighborhood. Given an optimization factor , a random number is generated. If the random number is less than the given optimization factor , variable neighborhood search is performed on the individual, and the specific method is as follows: Swap neighborhood: Randomly specify two positions in the job task numbers of the individual and the task execution numbers of the shuttle vehicle (i.e., segment 1 and segment 2) for swapping to generate a new individual; Insertion neighborhood: Randomly select an insertion point in the job task numbers of the individual and the task execution numbers of the shuttle vehicle (i.e., segment 1 and segment 2), and insert the adjacent numbers to obtain a new individual; Inversion neighborhood: Randomly select a neighborhood search point in the job task numbers of the individual and the task execution numbers of the shuttle vehicle (i.e., segment 1 and segment 2), and invert the job task numbers and the task execution numbers of the shuttle vehicle after this search point to generate a new individual.

[0040] When solving this job scheduling optimization problem, neighborhood search operations need to be performed on both the job task order and the shuttle vehicle execution order simultaneously. As shown in the appendix Figure 7 , segment 1 represents the job task order, and segment 2 represents the shuttle vehicle execution order. Through operations such as swapping, inserting, and inverting, a new job scheduling scheme will be generated. Taking the swap neighborhood operation as an example, swapping job tasks [1] and [2] gives a new job order [4, 2, 5, 7, 3, 1, 6]; swapping the shuttle vehicle task execution quantities [3] and [2] gives a new allocation scheme [2, 2, 3]. Through variable neighborhood search, both the job task order and the shuttle vehicle task execution quantity change, and then new individuals are generated, which is conducive to the algorithm finding a better solution based on the current solution.

[0041] To further improve the algorithm's solution quality, a dual-population recombination and collaborative optimization strategy is introduced. By exchanging genetic information of the dominant individuals in the two populations, that is, hoping to exchange the dominant solutions of the populations generated by two different methods, the limitation of the single-directional evolution of the population is broken.

[0042] The following examples are provided in this embodiment to further illustrate the advantages of the method of this embodiment: Taking the double-depth multi-shuttle car storage system of an enterprise as the research object, the outbound operation scheduling is optimized. The storage system has 3 floors, with 3 lanes on each floor. One shuttle car is configured on each floor, and there are 3 elevators. The remaining storage system parameter settings are shown in Table 2. According to the statistical data, there are 90 types of goods in this storage system, with a monthly shipment volume of about 12,000 pallets. There are 35 outbound tasks to be executed within a certain time window, and the coordinate information of the outbound tasks is shown in Table 3.

[0043] Table 2 Main input parameter settings of the storage system Table 3 Coordinate points of outbound tasks To verify the impact of the three goods transfer strategies of RTRP, RTNP, and RTFP on the outbound operation, when the occupancy rate of the storage location is 0.8, the above 35 outbound tasks are executed respectively. The selection of control parameters is shown in Table 4. In addition, a comparison is made with the system operation scheduling under the existing goods transfer operation time estimation method. For the RTFP goods transfer strategy, the decision maker selects the middle position of each lane as the fixed goods transfer location.

[0044] Table 4 Algorithm parameter settings According to the program operation, the execution order and task completion time of the outbound operation under the three goods-transfer strategies of RTRP, RTNP, and RTFP are obtained respectively, as shown in Table 5. Under the RTRP strategy, the task execution orders of the three shuttle cars are [20, 11, 29, 23, 17, 5, 9, 27, 3, 15, 31, 26, 14, 33], [7, 16, 13, 18, 22, 35, 1, 21, 30, 4, 10, 25], and [19, 6, 12, 32, 24, 2, 28, 8, 34], and the completion time is 558.25 s; under the RTNP strategy, the task execution orders of the three shuttle cars are [5, 9, 11, 29, 26, 21, 20, 3, 15, 27, 14, 33, 17, 31], [18, 22, 13, 1, 23, 30, 7, 16, 21, 10, 35, 4], and [32, 2, 8, 19, 6, 12, 34, 24, 28], and the completion time is 493.68 s; under the RTFP strategy, the task execution orders of the three shuttle cars are [29, 5, 9, 35, 26, 3, 11, 15, 12, 27, 20, 33, 23, 17], [21, 30, 7, 1, 31, 21, 10, 22, 18, 13, 16, 4], and [14, 28, 2, 34, 8, 32, 6, 19, 24], and the completion time is 520.83 s. The task execution order of the shuttle car operation under the goods-transfer operation time estimation strategy is the same as that of the RTFP strategy, and the operation time is 562.46 s. The task execution orders of the RTFP strategy and the estimation strategy are the same, but the operation times are different. Under the estimation strategy, the decision maker adopts the fixed-point goods-transfer strategy based on experience and makes an approximate estimation of the goods-transfer operation time each time, without precise calculation, so an error is generated. As the number of goods-transfer operations in the system increases, the estimation error also gradually increases, indicating that the derived calculation method of the goods-transfer operation time can accurately and effectively calculate the goods-transfer operation time, thereby reducing the error caused by estimation and achieving the accurate evaluation of the system operation time considering the goods-transfer operation.

[0045] Table 5 Comparison of outbound operation under three goods-transfer strategies The goods location occupancy rate is an important factor affecting the optimization of the outbound task scheduling of this system. The higher the goods location occupancy rate, the more goods are stored in the system, and the greater the probability of goods-transfer operations during the outbound task process. To compare the impacts of the three goods-transfer strategies of RTRP, RTNP, and RTFP on the outbound operation under different goods location occupancy rates, the above 35 outbound tasks are executed under the conditions of goods location occupancy rates of 0.6, 0.7, 0.8, and 0.9 respectively.

[0046] As shown in Table 6, through simulation calculations, the relocating operation time and task completion time of the three relocating operation strategies of RTRP, RTNP, and RTFP are obtained respectively under different slot occupancy rates. As the slot occupancy rate increases, the relocating operation time and task completion time of the three relocating strategies of RTRP, RTNP, and RTFP all increase. The larger the slot occupancy rate, the fewer idle slots in the warehousing system, the more complex the outbound task allocation, and the greater the probability of relocating operations during the outbound task process, resulting in an increase in both the relocating operation time and the task completion time. In addition, as the slot occupancy rate increases, the relocating operation time of the RTFP strategy gradually becomes less than that of the RTRP strategy. Since goods are continuously stored in the idle slots, the probability of finding an idle slot under the RTRP strategy gradually decreases, and the relocating operation time gradually increases, while the RTFP strategy is not affected by this condition. Under different slot occupancy rates, the relocating operation time and task completion time under the RTNP strategy are the smallest, because during the relocating operation process, the relocating distance of the shuttle vehicle is the smallest, indicating that the RTNP strategy is more suitable for the actual operation situation.

[0047] Table 6 Outbound operation conditions under different relocating strategies Under different slot occupancy rates, the number of relocating operations of the three relocating strategies of RTRP, RTNP, and RTFP is as shown in the appendix Figure 8 As shown. When the slot occupancy rate is small, the difference in the number of relocating operations under the three relocating strategies is not significant, because when the slot occupancy rate is low, the number of relocating operations during the outbound task process is small; as the slot occupancy rate increases, the goods stored in the warehousing system increase, and the number of relocating operations also increases. When the slot occupancy rate exceeds 0.75, the number of relocating operations under the RTRP strategy is the largest, and the number of relocating operations under the RTNP strategy is the smallest. Because when the slot occupancy rate increases, the difficulty of determining the relocating slot under the RTRP strategy increases, while it is easier to determine the relocating slot under the RTNP strategy. According to the above results, when the slot occupancy rate is small, any of the RTRP, RTNP, and RTFP strategies can be selected. When the slot occupancy rate exceeds 0.7, the RTNP strategy is the optimal choice.

[0048] To verify the effectiveness of the proposed variable neighborhood search and double-population recombination and cooperation optimization strategy, two groups of comparative experiments were designed, and the experiments were repeated 50 times for the cases where the scale of the outbound task was 35, 60, 80, and 100 respectively. The experimental results are shown in Table 7, where the DPGA-1 algorithm represents the use of only the variable neighborhood search strategy, and the DPGA-2 algorithm represents the use of only the double-population recombination and cooperation optimization strategy. According to the result comparison, under different task scales, the minimum completion time and average completion time of the DPGA algorithm are better than those of the DPGA-1 algorithm, indicating that this strategy can improve the optimization ability of the algorithm, thus verifying the effectiveness of the double-population recombination and cooperation optimization strategy; by comparing the results of the DPGA algorithm and the DPGA-2 algorithm, it can be seen that the use of the variable neighborhood search strategy can effectively increase the solution space and avoid the algorithm falling into local optimum, verifying the effectiveness of the variable neighborhood search strategy.

[0049] Table 7 Comparative experimental results for verifying the effectiveness of variable neighborhood search and population cooperation To further verify the effectiveness of the algorithm, the algorithm was experimentally compared with the improved grey wolf optimization algorithm (MGWO) and the genetic-beam search hybrid optimization algorithm (GA-BS) for the outbound operation scheduling, and the experiments were repeated 50 times for the cases where the scale of the outbound operation task was 35, 60, 80, and 100 respectively.

[0050] Table 8 Comparison of the results of three algorithms The experimental comparison results of the three algorithms are shown in Table 8. Under different task scales, the solution accuracy and stability of the algorithm are better than those of the two comparative algorithms, indicating that the proposed algorithm is an effective solution algorithm. As the task scale increases, the outbound operation task scheduling becomes more complex. Compared with the mGWO and GA-BS algorithms, the advantages of the proposed algorithm in solution accuracy and stability are continuously expanding, and the optimization effect is more obvious. When the task scale ranges from 35 to 100, the optimization efficiency of the DPGA algorithm increases from 13.28% to 24.26%, indicating that this algorithm is more suitable for solving large-scale scheduling optimization problems. Attached Figure 9 Figure 6 shows the convergence comparison diagram of the three algorithms for 60 outbound tasks. The mGWO algorithm can converge faster, but it is easy to fall into local optimum, resulting in low solution accuracy. Compared with the mGWO algorithm, the GA-BS algorithm has a slower convergence speed in the early stage, but the search results are better; according to Attached Figure 9 it can be seen that within 400 iterations, each algorithm can converge, and the proposed algorithm has the highest convergence speed and solution accuracy in the early stage. Thus, it can be seen that compared with the mGWO and GA-BS algorithms, the proposed algorithm has stronger optimal solution solving ability at the cost of sacrificing a certain amount of computing time.

[0051] In summary, it is first verified that the proposed calculation method for the goods transfer operation time can accurately and effectively calculate the goods transfer operation time, reducing the error caused by estimation. Secondly, the results show that when the occupancy rate of storage locations is small, any of the three goods transfer strategies can be selected; when the occupancy rate of storage locations exceeds 0.7, the RTNP strategy is more suitable for the actual operation situation. Finally, through comparative experiments, the superiority of the proposed variable neighborhood search strategy and double-population cooperation optimization strategy is verified. The use of double-population cooperation optimization can strengthen the information exchange between populations and improve the optimization ability of the algorithm, while the variable neighborhood search strategy can effectively expand the solution space and enhance the global search ability of the algorithm.

[0052] In summary, the resilient scheduling method for pulsating access in the port three-dimensional warehousing system of this embodiment proposes three goods transfer operation strategies, namely the random point goods transfer strategy, the nearest point goods transfer strategy, and the fixed point goods transfer strategy, based on the actual operation situation, analyzes the goods transfer operation process and establishes an accurate goods transfer operation time model. An accurate goods transfer operation calculation model and an optimized model for the outbound operation scheduling are established. On the basis of considering the goods transfer operation, the outbound operation is scheduled and optimized, a double-population genetic algorithm is designed to solve the scheduling model, and a double-population recombination and cooperation optimization strategy is proposed to enable the two populations to evolve differently while exchanging individuals in the populations to break the population state and jump out of the local optimum. In addition, the variable neighborhood search strategy is adopted to increase the diversity of the solution space and improve the search efficiency of the algorithm.

[0053] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, several improvements and refinements made without departing from the principle of the present invention should also be regarded as within the protection scope of the present invention.

Claims

1. A pulsating access toughness scheduling method for a port three-dimensional storage system, which is used in a storage system with double-depth storage shelves, shuttles, elevators and a control system, and is characterized by: The steps include: Step 1: The control system issues a storage task instruction, and the shuttle receives the outbound task instruction based on the first-come-first-served principle; Step 2: Determine whether the shuttle is in an idle state. If the shuttle is in an idle state, the shuttle moves to the row where the target goods are located. Otherwise, wait for the shuttle to be idle. Step 3: The shuttle vehicle drives to the row where the target goods are located, and determines whether unloading is required. If unloading is required, unloading is performed first. After unloading is completed, the shuttle vehicle loads the target goods and requests the elevator to complete the remaining operation process. If unloading is not required, the shuttle vehicle directly loads the target goods and requests the elevator to complete the remaining operation process. Step 4: determine whether the elevator is in an idle state. If the elevator is in an idle state, the elevator loads the target goods, otherwise wait for the elevator to be in an idle state; Step 5: The elevator loads the target goods and returns to the system I / O point, unloads the target goods, and the outbound task is completed; Among them, the process of the shuttle car's unloading operation is as follows: first, the shuttle car takes out the blocked goods and places them in the remaining vacant cargo space, which is also called the unloading cargo space; secondly, the shuttle car takes out the target goods and hands them over to the elevator to complete the vertical transportation of the target goods; finally, the shuttle car transports the blocked goods from the unloading cargo space to the original cargo space.

2. The pulsating access toughness scheduling method for a port three-dimensional storage system according to claim 1 is characterized by: The shuttle unloading operation strategy in step 3 is a random unloading strategy. The specific random unloading strategy is: the control system randomly selects an empty cargo position as the unloading cargo position according to the lane where the target cargo is located. The shuttle car first places the blocked cargo at the unloading cargo position, and then continues to complete the horizontal operation of the outbound task. Finally, the shuttle car returns to the lane to place the blocked cargo at the original position. At this point, the outbound operation task is completed. The unloading operation time of the random unloading strategy is the running time of the shuttle car between the unloading cargo position and the target cargo position, which is calculated by the following formula: ； in, For the unloading operation time, is the unloading operation distance, is the average speed of the shuttle, is the shuttle loading / unloading time, Number the shuttle. is the number of storage columns per row of shelves, The length of a single cargo space.

3. The pulsating access resilience scheduling method for a port three-dimensional storage system according to claim 1 is characterized by: The shuttle unloading operation strategy in step 3 is the nearest unloading strategy. The nearest unloading strategy is as follows: the control system selects an idle cargo position closest to the target cargo position as the unloading cargo position according to the lane where the target cargo is located. The shuttle car first places the blocked cargo at the unloading cargo position, and then continues to complete the horizontal operation of the outbound task. Finally, the shuttle car returns to the lane to place the blocked cargo at the original position, and the outbound operation task is completed. Among them, if there are two idle cargo positions with the same distance from the target cargo position at the same time, the idle cargo position closer to the buffer area is selected as the unloading cargo position. The unloading operation time of the nearest unloading strategy is calculated by the following formula: ； in, For the unloading operation time, is the unloading operation distance, is the average speed of the shuttle, is the shuttle loading / unloading time, is the length of a single cargo space, is the cargo space occupancy rate.

4. The pulsating access toughness scheduling method for a port three-dimensional storage system according to claim 1 is characterized by: The shuttle unloading operation strategy in step 3 is a fixed-point unloading strategy. The specific fixed-point unloading strategy is as follows: the control system selects a fixed free cargo position as the unloading cargo position according to the lane where the target cargo is located. The shuttle first places the blocked cargo at the unloading cargo position, and then continues to complete the horizontal operation of the outbound task. Finally, the shuttle returns to the lane to place the blocked cargo at the original position. At this point, the outbound operation task is completed. The unloading operation time of the fixed-point unloading strategy is calculated by the following formula: ； in, For the unloading operation time, is the unloading operation distance, is the average speed of the shuttle, is the shuttle loading / unloading time, Number the lane. for Fixed cargo unloading location coordinates in the aisle, The coordinates of the target location, The length of a single cargo space.

5. The pulsating access toughness scheduling method for a port three-dimensional storage system according to any one of claims 1 to 4, characterized in that: It also includes the total time of outbound operations The calculation steps are as follows: outbound operation tasks, the total outbound operation time of the system It can be expressed as: ； in, is the number of outbound operation tasks, and Respectively indicate execution of The running time of the shuttle and elevator for each outbound task, Shuttle loading / unloading time, It is the loading / unloading time of the elevator.

6. The pulsating access toughness scheduling method for a port three-dimensional storage system according to claim 5 is characterized by: The shuttle vehicle performs The running time of the outbound task is calculated as follows: The coordinates of the outbound task are , indicating that the outbound task is located in Layer, Lane, The shuttle's running time is It consists of three parts: shuttle handling time, loading time and unloading operation time, which are specifically expressed as follows: ； in, is the decision factor for unloading operations. When , it means that there is a reverse operation in the outbound task; when , it means that there is no cargo transfer operation during the outbound task; For the unloading operation time, is the average speed of the shuttle, is the shuttle loading / unloading time, is the number of lane columns, The length of a single cargo space.

7. The pulsating access toughness scheduling method for a port three-dimensional storage system according to claim 6 is characterized by: The hoist performs The running time of the outbound task is calculated as follows: outbound tasks, elevator operation time It consists of the hoist handling time and loading time. As shown below: ； in, is the number of storage shelves, Single-layer storage shelf height, The average speed of the elevator, It is the loading / unloading time of the elevator.

8. The pulsating access resilience scheduling method for a port three-dimensional storage system according to any one of claims 1 to 4, characterized in that: Steps 1 to 5 are performed by constructing a mathematical model, and the construction process is as follows: The outbound tasks are assigned to s four-way shuttles and t elevators for execution. The outbound task set is set as , the four-way shuttle set is , the elevator set is , the task set of the zth shuttle is , the outbound operation time is , the following mathematical model is established: ； The constraints are as follows: ； ； ； Formula (8) indicates that the shuttle performs all outbound tasks; Formula (9) indicates that an outbound task can only be performed by one shuttle; Formula (10) indicates that each shuttle performs at least one outbound task.

9. The pulsating access toughness scheduling method for a port three-dimensional storage system according to claim 8 is characterized by: The constructed mathematical model uses a dual population genetic algorithm to solve the model. The specific process is as follows: Step 1: Use integer coding based on outbound operations to design a three-segment chromosome coding method, where each chromosome represents a scheduling plan; the first segment is represented by the outbound task number, and the number of outbound task numbers indicates the execution order of the outbound task; the second segment is represented by the number of outbound tasks executed by the shuttle; the third segment is represented by the number of outbound tasks executed by the elevator; Step 2: introduce two populations, denoted as population 1 and population 2, and use different initialization methods to initialize the two populations. For population 1, use the random generation method to initialize according to the constraints. For population 2, first number the outbound tasks and randomly arrange them to obtain segment 1. Then, according to the principle that the shuttle and the goods are on the same layer, arrange segment 2 to balance part of the operation time. Finally, repeat the generation operation to obtain the initial population. Step 3, define the fitness function, and then select suitable individuals to enter the next generation according to the fitness value calculated by the fitness function; Step 4: For both population 1 and population 2, the tournament selection strategy is adopted. Two parents are randomly selected for fitness value comparison, and the one with higher fitness value is selected to enter the next generation. The operation is repeated to obtain a population of N. Step 5, perform adaptive crossover and mutation; Step 6: Given the optimization factor , generate a random number, if the random number is less than the given optimization factor , then a variable neighborhood search is performed on the individual.

Images