Although Automated Guided Vehicle Systems (AGVS) have been around for more than thirty years, designing an AGVS is still a difficult process because of the interaction of numerous system decisions. The many important elements and variables that must be considered when designing AGVS include the number and location of pickup and delivery stations, the number of vehicles, the routes used by vehicles, the dispatching rules, and the guide path layout. There are two basic categories of AGVS control: static and dynamic. A static control system requires the automated guided vehicle (AGV) to run the same route continuously with stops at each pickup/delivery station. On the other hand, vehicles in the dynamic control system can be routed to different stations using different paths. There are two types of dynamic vehicle control: workcenter-initiated and vehicle-initiated dispatching rules. The system invokes the workcenter-initiated dispatching rule when one workcenter has a pending request and more than one vehicle is available to pickup the request. Vehicle-initiated dispatching rule is employed when one vehicle is free and there is more than one outstanding request in the system. Most research to date analyzes only the static aspect of AGVS. This research attempts to find the minimum number of vehicles needed in an AGVS, using dynamic control of the vehicles, such that the chance of a vehicle-initiated situation occurring is less than a given very small threshold. Under these conditions, load requests will have the least chance of waiting to be picked up. Due to the stochastic behavior of the AGV systems, the proportion of time that the system is in either workcenter- or vehicle-initiated rules is unknown. In order to minimize the waiting time for the load requests while at the same time maintaining a minimum number of vehicles in the system, this research utilizes only workcenter-initiated dispatching rules. A model using queueing theory and Markovian processes is developed to investigate the relationships among empty vehicle travel time, number of vehicles, and the various types of workcenter-initiated dispatching rules. This model is then used to formulate a dispatching-rule based algorithm (DRBA) to determine the minimum number of vehicles. Also, given the number of vehicles required in the system, this research investigates the impact of the nearest-vehicle and farthest-vehicle dispatching rules on the steady-state system performance, e.g., the average waiting time of the load request. The model and algorithm are able to (1) improve the estimate of the number of vehicles required in an AGVS in the early design phase; (2) provide a better understanding of the efficiency of using the different types of workcenter-initiated dispatching rules for various workload conditions; and (3) generate analytical results that can be used as initial estimates for more detailed simulation studies.Ph. D.