Scheduling method for multi-layer and multi-port elevators

Abstract

A scheduling method for a multi-layer multi-port elevator comprises the following steps: obtaining elevator call tasks; if there is only one group of elevator call tasks, directly executing them; if the number of elevator call tasks is greater than 1 and not greater than M, using a full permutation algorithm to determine the execution order of each elevator call task with the shortest total length of the cargo receiving distance as the goal, and executing multiple groups of elevator call tasks according to the determined execution order, M is equal to 5 or 6; if the number of elevator call tasks is greater than M, using an adaptive genetic algorithm to optimize the execution order of the elevator call tasks with the shortest total length of the cargo receiving distance as the optimization goal, and executing multiple groups of elevator call tasks according to the optimized execution order. The present invention can realize the efficient allocation of elevator call tasks of elevators and improve the operating efficiency of multi-layer multi-port elevators.

Description

Technical Field

[0001] The invention relates to the technical field of logistics, and in particular to a dispatching method for an elevator. Background Art

[0002] With the optimization and upgrading of industrial production and logistics distribution, the efficiency of logistics elevators has become a key issue of concern to the industrial and logistics industries. The operating efficiency of the elevator has always been a pain point in the entire automated transportation system. Improving the operating efficiency of the elevator can effectively make up for the efficiency shortcomings of intelligent logistics transportation and achieve an overall increase in the speed of automated logistics transportation.

[0003] The essence of the optimization control of multi-layer, multi-port, and multi-task elevators is a scheduling problem, that is, the problem of finding the optimal solution to the multi-task running path. Among the optimal solution algorithms, common algorithms include genetic algorithms, simulated annealing algorithms, hill climbing algorithms, and particle swarm algorithms. The simulated annealing algorithm has strong local search capabilities and short running time, but lacks global search capabilities and is easily affected by parameters. The hill climbing algorithm has a simple principle and high efficiency, but it often cannot obtain the optimal solution for multi-constrained large-scale problems. The particle swarm algorithm is suitable for solving mathematical problems. The algorithm is simple and easy to calculate, but there is a problem of falling into the local optimum. The genetic algorithm can handle constraints well, is not limited by the local optimum, has strong global search capabilities, and can infinitely approach the global optimal solution, but the algorithm is relatively complex and has weak convergence ability. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a scheduling method for a multi-layer multi-port elevator, which can realize the efficient allocation of elevator call tasks and improve the operating efficiency of the multi-layer multi-port elevator.

[0005] The embodiment of the present invention provides a scheduling method for a multi-layer multi-port elevator, comprising the following steps:

[0006] Get the elevator call task;

[0007] If there is only one set of elevator calling tasks, it will be executed directly; if the number of elevator calling tasks is greater than 1 and not greater than M, the full permutation algorithm is used to determine the execution order of each elevator calling task with the goal of minimizing the total length of the cargo receiving distance, and multiple sets of elevator calling tasks are executed according to the determined execution order, where M is equal to 5 or 6; if the number of elevator calling tasks is greater than M, the shortest total length of the cargo receiving distance is used as the optimization goal, and an adaptive genetic algorithm is used to optimize the execution order of the elevator calling tasks, and multiple sets of elevator calling tasks are executed according to the optimized execution order.

[0008] The present invention has at least the following advantages:

[0009] 1. The embodiment of the present invention allocates the scheduling algorithm in real time according to the number of elevator call tasks. When the number of elevator call tasks is small, the full permutation algorithm is adopted. When the number of elevator call tasks is large, the genetic algorithm is adopted. The combined algorithm effectively realizes the optimal design of the control time and efficiency of the multi-layer, multi-port, and multi-task elevator scheduling system, and solves the problem that the genetic algorithm takes too long when the number of tasks is low, affecting the overall working efficiency of the elevator.

[0010] 2. The embodiments of the present invention improve the execution mechanism of the genetic algorithm, fully optimize the screening of genetic operators, effectively improve the repeated convergence characteristics of the genetic algorithm, reduce the maximum error, reduce the probability of premature convergence, and meet the actual control needs, etc., in order to solve the problems of poor repeated convergence characteristics, serious premature phenomenon, and excessive maximum error exhibited by the genetic algorithm in actual control. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A schematic diagram of a multi-layer and multi-port elevator is shown.

[0012] Figure 2 A schematic flow chart of a method for scheduling a multi-layer and multi-port elevator according to an embodiment of the present invention is shown.

[0013] Figure 3 A schematic diagram showing the principle of the full permutation algorithm.

[0014] Figure 4 A schematic diagram of an improved genetic algorithm according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0015] Figure 1 A schematic diagram of a multi-layer multi-port elevator is shown. Figure 1 In the example, the elevator is a multi-layer double-port elevator, and the system has the same bilateral design structure. Figure 1 Only its single-side structure is shown, which includes three parts: the transport end, the delivery end and the operating table. The transport end is the part inside the elevator car, the delivery end is the transitional delivery part, and some sites have multiple delivery sections.

[0016] The elevator adopts the one-cargo-one-elevator mode, and the elevator only transports one trailer of standard cargo at a time. The user enters the cargo transportation floor information through the operation console, and places the cargo on the conveyor line to complete the cargo transportation floor data entry. The elevator transports the cargo to the designated floor according to the cargo path information.

[0017] Assume that the position of the elevator before operation is a, and the coordinates of the starting and ending points of the n-point operation task are (b n , c n ), then all running paths X of n tasks L for:

[0018] X L =|a-b1|+|b1-c1|+|c1-b2|+…+|b n -c n |

[0019] Since the elevator adopts a one-cargo-one-elevator model, each task needs to complete a complete set of transportation processes including receiving, delivering, and shipping goods, and the operating rhythm of the horizontal conveyor line is much greater than the operating rhythm of the elevator. Therefore, the acceleration and deceleration time of the elevator and the waiting time for receiving goods are not considered, and the optimal solution for the elevator is only considered from the perspective of the operating distance.

[0020] According to the elevator operation model, no matter how the scheduling tasks are allocated, the running distance in the elevator task system is fixed, and the distance affected by the scheduling is only the distance from the current delivery port to the next task delivery port. Therefore, it is only necessary to sort the total length of the delivery distance in the best way. i :

[0021] X i =|a-b1|+|c1-b2|+…+|c n-1 -b n |.

[0022] See also Figure 2 According to an embodiment of the present invention, the scheduling method of a multi-layer multi-port elevator comprises the following steps:

[0023] Get the elevator call task;

[0024] If there is only one set of elevator calling tasks, it will be executed directly; if the number of elevator calling tasks is greater than 1 and not greater than M, the full permutation algorithm is used to determine the execution order of each elevator calling task with the goal of minimizing the total length of the cargo receiving distance, and multiple sets of elevator calling tasks are executed according to the determined execution order, where M is equal to 5 or 6; if the number of elevator calling tasks is greater than M, the shortest total length of the cargo receiving distance is used as the optimization goal, and an adaptive genetic algorithm is used to optimize the execution order of the elevator calling tasks, and multiple sets of elevator calling tasks are executed according to the optimized execution order.

[0025] In a specific implementation, the scheduling method of this embodiment establishes a pre-processing queue for elevator call tasks and a task execution queue. After the cargo elevator call is successful, the elevator call task will be dynamically processed according to the actual operation of the elevator. When the elevator is in an idle state, any elevator call command will be directly executed without sorting. If the elevator is in operation, the newly received elevator call request will be temporarily stored in the task pre-processing queue. After the current elevator task is completed, the elevator call task will be sorted. The elevator call task sorting is divided into three cases according to the number of tasks:

[0026] 1. There is only one set of elevator call tasks, which are executed directly;

[0027] 2. When the number of elevator call tasks is greater than 1 and less than 6, the full permutation algorithm is executed, with a maximum of 720 sorting calculations, which is absolutely optimal.

[0028] 3. When the number of elevator call tasks is greater than 6, an improved adaptive genetic algorithm is used for sorting. The number of calculations is mainly determined by the evolutionary generations and the number of tasks, which is relatively optimal.

[0029] The full permutation algorithm generates all possible running paths for all elevator call tasks, and then calculates their path distances. The basic principle of the full permutation algorithm is as follows: Figure 3 As shown in the figure, taking the three numbers 0, 1, and 2 as an example, each number represents an element (i.e., a set of elevator call tasks), and there are two possible arrangements of two elements, namely, 1, 0 / 0, and 1; the full arrangement of three elements has 6 queues. The origin of these 6 queues can be imagined as inserting the element 2 into the two known queues 1, 0 / 0, and 1 in sequence. Therefore, Figure 3 As shown, these 6 queues will be: 2, 1, 0 / 1, 2, 0 / 1, 0, 2 / 2, 0, 1 / 0, 2, 1 / 0, 1, 2. Similarly, when the fourth element appears, insert this element into these 6 queues in turn, and each queue will generate 4 new queues, thus generating 24 new queues. Similarly, the full queue sorting of any bit element can be completed. After completing the full permutation array of all elevator call tasks according to this logic, substitute the path distance and use the bubble algorithm to extract the optimal solution.

[0030] This embodiment adopts a combined algorithm sorting method, which can improve the system calculation and execution efficiency when the number of tasks is low, while ensuring the optimal solution for sorting. When the number of tasks is too large, if the full permutation algorithm is still used, it will take a huge amount of calculation time. Therefore, according to the calculation items, 6 tasks are used as the demarcation point, and the processing algorithm is reasonably allocated according to the number of real-time elevator call tasks.

[0031] In this embodiment, the task of obtaining an elevator call includes:

[0032] The received elevator call request is put into the task queue as an elevator call task, and the elevator call request includes a cargo entry floor, a cargo exit floor, a cargo entry port, and a cargo exit port.

[0033] See also Figure 4 In this embodiment, the optimization of the execution order of elevator call tasks by using an adaptive genetic algorithm includes the following steps:

[0034] S1. Take a set of elevator call task execution order as a genetic individual in the population and generate the initial population N. A ;

[0035] S2, initialize the population N A Optimize and get the optimized initial population Nc ;

[0036] S3, the optimized initial population N c The adaptive genetic operation is iteratively performed until a preset iteration termination condition is met, and the optimal solution when the iteration termination condition is met is output as the execution order of the optimized elevator call task.

[0037] In this embodiment, the above step S1 specifically includes:

[0038] Generating chromosome: P i =(b i ,c i ), bi is the incoming cargo floor of the i-th elevator call task, ci is the outgoing cargo floor of the i-th elevator call task, 1≤i≤n, n is the total number of elevator call tasks;

[0039] According to the order of the calling tasks, the genes as the genetic individuals are generated: P j '={P1,P2…P n};

[0040] Generate the initial population N A :N A = {P1',P2'...P m '}; m is the population size.

[0041] In this embodiment, the above step S2 specifically includes:

[0042] For the initial population N A Perform a single genetic operation to obtain a new population N B , the population size is m A With a population size of m, N B Mix and select the best output of a new initial population N with a population size of m C .

[0043] The above optimal output population size is m, and the new initial population N C The following steps are involved:

[0044] Calculate the average fitness of the mixed population (the size of the mixed population is 2m), and genes with higher fitness than the average will enter the next generation population N C , genes with fitness below the average are discarded; if the population size is met in advance (i.e. m genes have been selected), the selection is terminated; if the total number of remaining genes to be selected in the mixed population is equal to the total number of remaining requirements of the new population (i.e. the number of remaining genes to be selected = m-the number of genes that have been selected), the selection is stopped and all the remaining genes are included in the new population.

[0045] The above initial population N A Perform a single genetic operation to generate population NB The specific steps include:

[0046] Calculate the fitness F of each gene:

[0047] F=1 / X i , X i is the total length of the receiving distance, that is, the fitness F is equal to the inverse of the total length of the receiving distance; 1≤i≤n, n is the total number of elevator call tasks;

[0048] Calculate the selection probability F of each gene p :

[0049] That is, the sum of single gene fitness / population fitness, m is the population size;

[0050] The gene for the next evolution is selected by the ergodic roulette rule, the selected gene is used as the mother, a new gene is obtained by crossover operation, and the selected gene is mutated to obtain a new gene to generate a population N B .

[0051] The method of selecting the next gene for evolution by ergodic roulette selection includes: performing m rounds of selection, randomly generating a random number in the interval [0, 1] in each round of selection, comparing each random number with the probability of selection of m genes in the population in turn, until a gene with a probability of selection greater than the random number is found, then stopping the comparison and selecting the gene to enter the next generation of genetic population; in the next round of selection, starting from the next gene of the gene selected in the previous round, the probability of selection of each gene in the population is compared with the random number in turn, if the gene with a probability of selection greater than the random number is not found after comparing to the last gene in the population, then returning to the first gene in the population for comparison. In the prior art, in each round of selection, the first gene in the population is compared with the random number, and the gene with a probability of selection greater than the random number is found, then the comparison is stopped, and this method easily generates a local optimal solution. In the next round of selection of this embodiment, the next gene of the gene selected in the previous round is compared with the random number, which can avoid the generation of a local optimal solution.

[0052] In this embodiment, the above step S3 specifically includes:

[0053] S31. Calculate the fitness F of each gene as a genetic individual in the population:

[0054] F=1 / X i , X i is the total length of the delivery distance, that is, the fitness F is equal to the reciprocal of the total length of the delivery distance;

[0055] S32. Calculate the selection probability F of each gene P :

[0056] m is the population size, that is, the selection probability F P equals the sum of the fitness of a single gene divided by the fitness of the population;

[0057] S33. Select the genes for the next evolution through traversal-based double-gene roulette wheel selection. Use the selected genes as the parent body, obtain new genes through adaptive crossover operations, and perform adaptive mutation operations on the selected genes to obtain new genes, generating a new population;

[0058] The selection of the genes for the next evolution through traversal-based double-gene roulette wheel selection includes: performing m rounds of selection. In each round of selection, two random numbers are randomly generated in the interval [0, 1]. Each random number is successively compared with the selection probabilities of the m genes in the population until a gene with a selection probability greater than the random number is found. Then, stop the comparison and retain the gene. Compare the two genes obtained according to the two random numbers, and select the gene with the larger selection probability to enter the next-generation genetic population; in the next round of selection, start from the next gene after the gene selected in the previous round and successively compare the selection probabilities of the genes in the population with the random number. If no gene with a selection probability greater than the random number is found after comparing to the last gene in the population, then return to the first gene in the population for comparison.

[0059] In the operator selection of the basic genetic algorithm, after selecting a group of individuals nearby, it will start the next individual selection again, and it is easy to only select the individuals with relatively high fitness in the front part of the initial population. Therefore, assuming that the size of the initial population is m, and the individuals selected by the S-th selection operator are the n-th individuals (n < m) in the initialized population, then the selection of the S1-th operator starts from the (n + 1)-th individual in the initial population until all the initialized population is screened, and then traverse and screen again. In the selection calculation of this embodiment, the selection of the best operator is no longer to directly output as long as a group of operators that meet the constraint conditions are encountered, but to keep the first group of excellent operators, continue to screen a group of excellent operators, and select the best output from the two groups of operators. In this way, the computing efficiency can be fully utilized.

[0060] In this embodiment, the adaptive crossover probability P c for adaptive crossover calculation has the following calculation formula:

[0061]

[0062] The adaptive mutation probability P m for adaptive mutation calculation has the following calculation formula:

[0063]

[0064] Where: P C is the crossover probability, P mis the mutation probability, f′ is the larger fitness value of the two sets of genes to be crossed, avg is the average fitness of the population, f max is the maximum fitness of the population, f is the fitness of the mutated gene, P1, 2, k1, k2 are all control parameters, and the control parameters P1, 2, k1, k2 are all pre-set artificially.

[0065] The adaptive crossover probability P of the existing adaptive genetic algorithm c and adaptive mutation probability P m It is more inclined to change low fitness individuals, and almost no change occurs to high fitness individuals, so that the algorithm converges to the local value and may increase. The adaptive crossover probability P of this embodiment c and adaptive mutation probability P m The exponential calculation formula is used to improve the stability of the crossover probability and mutation probability changes and avoid large changes in the calculation factors.

[0066] S34, add the optimal solution of the previous generation to the population, and select the optimal solution of the population through bubble sorting. If the optimal solution is still the optimal solution of the previous generation, discard the last group of genes, and incorporate the optimal solution of the previous generation into the genetic population and pass it to the next generation; if the optimal solution is not the optimal solution of the previous generation, randomly take out a group of genes to compare with the optimal solution of the previous generation, and select the best one to be included in the population; step S34 corresponds to Figure 4 The optimal fitness genetic factor is added;

[0067] S35, judging whether the newly formed population has satisfied the preset iteration termination condition, if not, returning to step S31, if satisfied, the gene with the largest fitness in the population is output as the optimal solution. In this embodiment, the preset iteration termination condition is that the genetic generation of the population reaches the preset termination genetic generation.

[0068] In summary, the main differences between the improved adaptive genetic algorithm used in this embodiment and the general genetic algorithm are as follows:

[0069] (1). Optimize the initialization group;

[0070] (2) Optimize selection;

[0071] (3). Adaptive parameters

[0072] Replace fixed crossover and mutation probabilities with adaptive genetic parameters;

[0073] (4) Elitism

[0074] In order to prevent the optimal solution from being destroyed by crossover or mutation during the evolution process, the optimal solution with the best fitness is directly used as the final operator to participate in the sorting of the shortest path and is inherited to the next generation.

[0075] The embodiment of the present invention adopts a combined algorithm to effectively reduce the execution time of the system when the number of tasks is low, ensuring that the execution queue of the low number of tasks is absolutely optimal; and when the number of elevator call tasks is large, a genetic algorithm is adopted to output the execution tasks relatively accurately within a limited time, ensuring the execution efficiency of the entire system.

[0076] It has been verified that the combined algorithm can effectively reduce the system calculation time and improve the overall execution efficiency of the system compared to a single algorithm, and optimize the system execution efficiency. The improved genetic algorithm can obtain better repeated convergence, greatly reduce the maximum error, significantly reduce the number of abnormal time consumption, and can adapt to actual control needs.

[0077] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A dispatching method for multi-layer and multi-port elevators, It is characterized in that The following steps are involved: Get the elevator call task; If there is only one set of elevator call tasks, it is executed directly; if the number of elevator call tasks is greater than 1 and not greater than M, the execution order of each elevator call task is determined by using the full permutation algorithm with the shortest total length of the pickup distance as the goal, and multiple sets of elevator call tasks are executed according to the determined execution order, M is equal to 5 or 6; if the number of elevator call tasks is greater than M, the execution order of the elevator call tasks is optimized by using the adaptive genetic algorithm with the shortest total length of the pickup distance as the optimization goal, and multiple sets of elevator call tasks are executed according to the optimized execution order; the optimization of the execution order of elevator call tasks by using the adaptive genetic algorithm includes the following steps: S1. Take a set of elevator call task execution order as a genetic individual in the population and generate the initial population N. A ; The step S1 comprises: generating chromosome: P i =(b i ,c i ), b i is the cargo floor for the i-th elevator call task, c i is the delivery floor of the i-th elevator call task, 1≤i≤n, and n is the total number of elevator call tasks; according to the order of the elevator call tasks, the gene as the genetic individual is generated: P j '={P 1 ,P 2 …P n Generate the initial population N A :N A = {P 1 ',P 2 '...P m '}; m is the population size; S2, for the initial population N A Optimize and get the optimized initial population N c ; S3, the optimized initial population N c The adaptive genetic operation is iteratively performed until a preset iteration termination condition is met, and the optimal solution when the iteration termination condition is met is output as the execution order of the optimized elevator call task.

2. The dispatching method of a multi-layer multi-port elevator according to claim 1, It is characterized in that The adaptive genetic operation includes selection, adaptive crossover and adaptive mutation.

3. The dispatching method of a multi-layer multi-port elevator according to claim 1, It is characterized in that The step S2 comprises: For the initial population N A Perform a single genetic operation to obtain a new population N B , the population size is m A With a population size of m, N B Mix and select the best output of a new initial population N with a population size of m C .

4. The dispatching method of a multi-layer multi-port elevator as claimed in claim 3, It is characterized in that The optimal output population size is m, and the new initial population N is C include: Calculate the average fitness of the mixed population, and genes with higher fitness than the average will enter the next generation population N C , genes with fitness below the average are discarded; if the population size is met in advance, the selection is terminated; if the total number of remaining genes to be selected in the mixed population is equal to the total number of remaining requirements of the new population, the selection is stopped and all remaining individuals are included in the new population.

5. The dispatching method for a multi-layer multi-port elevator according to claim 1, It is characterized in that The step S3 comprises: S31, calculate the fitness F of each gene in the population as the genetic individual: F=1 / X i , X i is the total distance of receiving goods, 1≤i≤n, n is the total number of elevator calling tasks; S32. Calculate the selection probability F of each gene P : m is the population size; S33, selecting the gene for the next evolution through ergodic two-gene roulette, taking the selected gene as the mother, obtaining a new gene through adaptive crossover operation, and performing adaptive mutation operation on the selected gene to obtain a new gene, thereby generating a new population; S34, add the optimal solution of the previous generation to the population and sort it, then select the optimal solution of the population. If the optimal solution is still the optimal solution of the previous generation, discard the last group of genes, incorporate the optimal solution of the previous generation into the genetic population, and pass it on to the next generation; if the optimal solution is not the optimal solution of the previous generation, randomly select a group of genes to compare with the optimal solution of the previous generation, and select the best one to be incorporated into the population; S35, judging whether the newly formed population has satisfied the preset iteration termination condition, if not, returning to step S31, if satisfied, outputting the gene with the largest fitness in the population as the optimal solution.

6. The dispatching method of multi-layer multi-port elevator according to claim 5, It is characterized in that The method for selecting the next evolving gene by traversal double gene roulette comprises: performing m rounds of selection, randomly generating two random numbers in the interval [0, 1] in each round of selection, comparing each random number with the selection probability of m genes in the population in turn, until a gene with a selection probability greater than the random number is found, then stopping the comparison and retaining the gene, comparing the two genes obtained according to the two random numbers, and selecting the gene with a greater selection probability to enter the next generation genetic population; in the next round of selection, starting from the next gene of the gene selected in the previous round, comparing the selection probability of each gene in the population with the random number in turn, if the gene with a selection probability greater than the random number is not found after comparing to the last gene in the population, then returning to the first gene in the population for comparison.

7. The dispatching method for a multi-layer and multi-port elevator according to claim 1, It is characterized in that The task of obtaining an elevator call includes: The received elevator call request is put into a task queue as an elevator call task, wherein the elevator call request includes a cargo entry floor, a cargo exit floor, a cargo entry port, and a cargo exit port.

Images