Due to the competition for limited resources by many concurrent processes in large scale automated manufacturing systems (AMS), one has to resolve a deadlock issue in order to reach their production goal without disruption and downtime. Monolithic resolution is a conventional approach for optimal or acceptable solutions, but suffers from computational difficulty. On the other hand, some decentralized methods are more powerful in finding approximate solutions, but most are application-dependent. By modeling AMS as Petri nets, we develop an innovative distributed control approach, which can create a trajectory leading to a desired destination and are adaptable to different kinds of constraints. Control strategies are applied to processes locally such that they can concurrently proceed efficiently. Global destinations are always reachable through the local observation upon processes without knowing external and extra information. Efficient algorithms are proposed to find such distributed controllers.