Agent based periodic control of an AGV system in dynamic transportation

Abstract

In this paper, we propose a periodic control strategy for the assignment of the tasks to several AGVs (Automated Guided Vehicles) in dynamic transportation. In the proposed method, all of the transportation tasks that arrive at a time are collected during a given time period and an efficient scheme of the assignment is generated by the optimization procedure, that has to be completely executed during the next period. An meta-heuristic algorithm called pheromone communication method is adopted as an optimization method for minimizing the traveling time of AGVs while maintaining a short computation time. In our study, the optimal number of collecting new tasks is investigated by changing the various timing of sending the requests to AGVs in the numerical experiments. The effectiveness of the proposed method is demonstrated by comparing the total throughput of an AGV system with that of the conventional method for a large scale transportation bay. Automated Guided Vehicles (AGVs) are widely used as a means of transportation system in semiconductor fab- ricating hays. There exists increased interests in opera- tional control of AGV systems in semiconductor fabri- cating hay currently with the expansion of AGV system. The operational transportation control of AGV system consists of two levels. The first level is the assignment of tasks to AGVs when the transportation requests to he performed. The second level is to derive the transporta- tion route of multiple AGVs minimizing the traveling time without the collision and deadlock among AGVs. These problems can he treated as a combinatorial op timization problem which is extremely difficult to solve even in a static case. It has been requested to obtain a near optimal solution in a extremely short computation time for the assignment and route planning problems in dynamic case where the requests are given in real time. Conventional method for the task assignment has been based on several heuristic rules such as Near Neighbor Rule or Longest Idle Vehicle first Rule mini- mizing the computation time from a practical point of view. The simulation based analysis for the assignment rules has been extensively studied by Mantel et. all). However, when the frequency of the arrival of trans- portation tasks is higher than the capacity of the AGV route planning problems. In the proposed method, some of transportation tasks that arrive are collected and an efficient assignment scheme is generated by the opti- mization procedure, that has to be completely executed during the next period. An agent based task assignment algorithm is developed to reduce the computation time. In this algorithm, each AGV agent searches its candi- date of solution by using the value of pheromone. The use of pheromone from the analogy of ant system for the combinational optimization problem has been pro- posed by Dorigo et al.') The pheromone communication method is adopted as a optimization method for gener- ating the solution with minimum computation time. In our study, the optimal number of collecting new task is investigated by changing the Tarious timing of sending the tasks to AGVs. The effectiveness of the proposed method is demonstrated by numerical exper- iments.