A neural network based multi-state scheduling algorithm for multi-AGV system in FMS

Abstract

Due to its numerous advantages in terms of flexibility, maintainability and construction cost, the Flexible Manufacturing System (FMS) has received increasing attention over the last few years. The application of multi automated guided vehicle (AGV) transportation system promotes the flexibility of the FMS, which also imposes higher requirements for AGV scheduling algorithms. This paper develops a novel multi-state scheduling algorithm (MSSA) for AGV that makes a good trade-off between AGV utilization rate and total processing make-span in the FMS. Compared with classic idle-scheduling strategy, MSSA schedules more AGVs and tasks in each calculation, making its optimization target closer to the global optimization target. A neural network based travel time prediction is applied to improve the performance of the proposed algorithm on time accuracy. Five factors are used as neural network inputs to express the impact of vehicle state, distance travelled, AGV trajectory and multiple AGV collisions on travel time. Simulation experiments indicate the proposed algorithm has advantages in terms of total processing make-span, AGV load rate, AGV utilization and task execution time. The proposed algorithm has been applied in an actual air conditioner production line.