The Integrated Optimization Of Location Assignment And Sequencing In Multi-shuttle AS/RS Under MMC Pattern

In recent years, because of the rapid development of e-commerce, the traditional logistics systems are encountering with a huge challenge from automation and information. Various kinds of new warehousing systems are emerging. Multi-shuttle automated storage/retrieval system(AS/RS) can efficiently reduce the empty-travel time and improve the overall system efficiency. Therefore, the studies on location assignment and sequencing problems in multi-shuttle AS/RS attract many researchers’ interests.This paper deals with the integrated optimization of location assignment and sequencing in multi-shuttle AS/RS under MMC pattern.This paper is comprised of two problems.In problem I, each retrieval request is designated with a retrieval location, and storage location assignment and sequencing problems will be jointly optimized. Based on problem I, retrieval location assignment is also considered in problem II. Therefore, problem II explores the integrated optimization of storage/retrieval location assignment and sequencing. Integer programming models and heuristics are constructed for each problem considering their own characteristics. Problem I involves two heuristics. The first one is the First-come-first-served(FCFS) strategy. The second one is referred to as FCFS+TSA, in which the solution of FCFS is improved by tabu search algorithm(TSA). In addition to FCFS and FCFS+TSA, nearest neighbor heuristic(NN) and NN+TSA are also involved in Problem II. In NN+TSA, the solution generated by NN will be improved by TSA.The performances of mathematical model and heuristics are investigated in numerical experiments. We construct 48 different instances for problem I in the form of(L,m,n), and each instance is solved by CPLEX, FCFS and FCFS+TSA. According to the obtained results, several conclusions can be figured out.(1) The solution quality of FCFS becomes worse when dealing with large-sized instances.(2) CPLEX is only suitable for solving small-sized instances.(3)FCFS+TSA can solve an instance with the same solution as CPLEX in less computation time.(4)When dealing with large-sized instances, FCFS+TSA performs better than CPLEX in terms of solution quality and computation efficiency.We formulate 800 instances for problem II, and each instance can be solved by five methods: CPLEX, FCFS, FCFS+TSA, NN and NN+TSA. CPLEX is only used to solve small-sized instances with optimal solutions, and the solutions can be used as a reference to that of other heuristics. As for the large-sized instances, four heuristics except for CPLEX can be adopted. Several conclusions can be obtained based on the results:(1) TSA can improve the solution of FCFS and NN obviously.(2) When dealing with larger instances, NN+TSA performs better than FCFS+TSA in terms of solution quality and computation efficiency.(3) The computation efficiency of FCFS+TSA/NN+TSA becomes worse. Nonetheless, both of them performs better than CPLEX in terms of efficiency.(4) When dealing with instances with large rack and numerous shuttles, TSA’s performance become worse with the increase of the number of scheduling cycles. Then a recommended parameter list is given according to the results and figures. Finally, the author makes a conclusion of this paper and gives some future research topics.