Based on Petri net models of flexible manufacturing systems and embedding the optimal deadlock avoidance policy into the particle swarm optimization(PSO) algorithm, a deadlock-free modified PSO scheduling algorithm is proposed to minimize the makespan. A particle is corresponded to a part sequence which is a permutation with repetition of parts. Real numbers in components of a position vector indicate priorities of the corresponding part operations. The feasibility of
sequences is checked with the one-step look-ahead method in the optimal deadlock control policy, and a strategy for jumping out local optima is proposed. Experimental results show the effectiveness and feasibility of the particle swarm optimization for solving deadlock-free scheduling of flexible manufacturing systems and the modified particle swarm scheduling algorithm can perform much better.