Autonomous Decentralized FMS and its Efficient AGVs Control

Abstract

This paper describes Autonomous Decentralized FMS (AD-FMS) and its method to control automatic guided vehicles (AGVs) by using a memory. The aim is to increase the reasoning efficiency of a system the authors call reasoning to anticipate the future (RAF) which controls AGVs in AD-FMS. This RAF applies hypothetical reasoning to the number of next actions that can be considered for the AGV (competing hypotheses). However, if the number of agents included in the hypothetical reasoning process in the RAF is increased, the number of next actions that are considered as competing hypotheses also increases. As a result, the replacement of true and false hypotheses and number of repetitions of discrete production simulations produced by these replacements are increased, giving rise to the problem of decreased reasoning efficiency of the RAF. The present article reports a method to solve these problems. The reported method, the authors call ranking by oblivion and memory (ROM), is based on the idea that when a production situation occurs that is the same as one in the past, the same destination as in the past is more likely to be selected; that is, it has a high probability of being selected as the true hypothesis. By applying the ROM to AD-FMSs constructed on a computer, it was found that under all conditions the ROM reduced the number of hypothesis replacements to half that of a conventional system, demonstrating the validity of this system.