For most of the transportation system in semiconductor fabricating bay, the requests for transportation are given in every few seconds. It is necessary to determine the allocation of the transportation requests to AGVs and the route planning of AGVs without collision and deadlock in a dynamic environment. In this paper, we propose an autonomous decentralized route planning method for dynamic route planning. In the proposed system, each AGV individually derives the route for itself by repeating the generation of the route for each AGV and the wireless communication with other AGVs. Each AGV has a function of optimizing the route for itself. However, it is desirable that a rescheduling is executed locally as much as possible. From that viewpoint, we propose a rescheduling algorithm for dynamic environment. From the numerical experiments, it has been demonstrated that the proposed method can reduce 80 percent of the computation time compared with the conventional autonomous decentralized method without lowering the performance level. Moreover, the proposed method has been implemented on a parallel experimental system and it has been demonstrated that each robot can successively travel without deadlock and collision among AGVs.