Generation of optimal routes in a neural network based AGV controller

Abstract

Due to the increasing complexity of factory floor environments coupled with the need for increased flexibility in automated guided vehicle (AGV) systems, it is becoming increasingly important to be able to dynamically alter both the AGV job queue and the AGV path. In a previous paper, a new method based on an artificial neural network model was presented for evaluating the best job assignment so as to achieve better system efficiency. This scheme generated the best set of workstations or loading points for a given AGV to visit, given the current status of both the AGV fleet and the waiting job requests. The next stage in the control strategy is to find the optimal route between the workstations on the list. In this paper, a new navigation technique for AGV's based on the A* search algorithm called `forward-reverse search' is presented which fits into our neural network based control structure.