(A) study on dispatching algorithms for automated guided vehicle systems

Abstract

Automated Guided Vehicle Systems(AGVSs) have emerged as an important part of material handling in Flexible Manufacturing Systems(FMSs). This is due to the fact that AGVS provides the flexibility and adaptability required from material handling systems in FMS environment. This thesis deals with dispatching algorithms which are recognized as a key element for efficient operations of AGVS. In the first part, this thesis develops a new automated guided vehicle (AGV) dispatching algorithm based on a bidding concept. In the suggested algorithm, the information of work-in-process(WIP) in incoming and outgoing buffers of workstation, and the travel time of an AGV become input parameters of mathematical functions. These functions will be called "bidding functions". To integrate the values coming from these bidding functions, "dispatching function" is suggested which provides a basis for the selection of the most urgent among competitive move requests. Four variations of the dispatching function are suggested and examined. The performance of the algorithm is compared with some well-known existing rules in terms of the system throughput through simulation on a hypothetical job shop type manufacturing system. The second part presents an another type of dispatching algorithm which has an adaptive control capability. The parameter values of the bidding functions are updated with an evolutionary process either on a regular time interval or at the time blocking occurs. Thus the algorithm can respond to changes of the system environment, i.e., addition/elimination of some workstations and/or vehicles, and variation of material flow pattern. It can also reflect on-going situation of each workstation, i.e., busy or not, and frequent blockings or starvations, etc. A set of proper parameter values is adjusted by the evolutionary process which utilizes the historical information on the average WIP level of each workstation and the occurrences of blocking in workstations. The perfo...