This paper presents a colored resource-oriented Petri net (CROPN) modeling method to deal with conflict and deadlock arising in automated guided vehicles (AGV) systems. It can handle both bidirectional and unidirectional paths. The former offer additional flexibility, efficiency, and cost saving when compared with the latter. Yet, they exhibit more challenging AGV management problems. Unlike jobs that can enter and leave automated manufacturing systems, AGVs always stay in the system. By modeling nodes with places and lanes with transitions, the proposed method can construct CROPN models for changing AGV routes. A control policy suitable for real-time implementation is presented.