ADAPTIVE TASK SEQUENCING FOR A FLEXIBLE MANUFACTURING SYSTEM
This paper presents a scheduling reinforcement learning algorithm designed for solving a job sequencing problem presented by a flexible manufacturing system. This is achieved by tailoring a unique sequence for every manufacturing order according to its specific characteristics. The algorithm addresses the problem of sequencing tasks executed by a single transfer agent (a robot), aiming to achieve minimal completion times of manufacturing orders. The performance of the algorithm was evaluated in simulation under various combinations of manufacturing order sizes, products mixes, and part inter-arrival times. Comparison of the policies generated by the algorithm to the FIFO policy currently implemented on the system, demonstrated the superiority of the new algorithm in all tested combinations. The presented algorithm can be adjusted to suit other sequencing problems since it does not require any predefined sequencing rules or problem specific information.
Reinforcement Learning Adaptive Sequencing.
A. Gil S. Berman H. Stern Y. Edan
Department of Industrial Engineering and Management, Ben-Gurion University of the Negev, Beer-Sheva Department of Industrial Engineering and Management, Ben-Gurion University of the Negev,Beer-Sheva 8
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）