FLOW-SHOP PROBLEM WITH JOB TRANSFER ROBOTS USING REINFROCEMENT LEARNING
This paper considers a Flow-Shop problem with N jobs, M Machines and K job transport robots. Identical jobs are processed by exactly M machines in a fixed order. The objective is to develop scheduling algorithms with minimum makespan for robots that transport the jobs between successive machines returning empty to move the next job. Variations based on equal/no equal processing times and zero/nonzero empty return times are considered. When the optimal solution can be achieved in polynomial time, propositions are offered. Otherwise, reinforcement learning is used to obtain job-robot schedules. Results of a two robot problem in which the working area is partitioned so that each robot is designated to an equal subset of machines (m) are compared to a maximum lower-bound (LB) and errors (UB) are computed (Min (UB) -Max (LB))/Max (LB).
Job transfer robots Reinforcement learning Flow-Shop Scheduling.
Kfir Arviv Helman Stern Yael Edan
Department of Industrial Engineering and Management, Ben-Gurion University of the Negev Beersheba, I Department of Industrial Engineering and Management, Ben-Gurion University of the NegevBeersheba, Is
国际会议
上海
英文
1-6
2009-08-02（万方平台首次上网日期，不代表论文的发表时间）