In order to efficiently operate an FMS (flexible automation system) with AGVs, it becomes important to construct a good schedule for AGV driving. This paper presents a method for solving an AGV driving scheduling problem. The FMS model treated has a one-way AGV driving lane. Under this condition, passing and collision among the AGVs are not allowed in the lane except when the AGV remains in machine shop buffer area. The problem is NP-complete and it cannot usually be solved by mathematical programing. A simple genetic algorithm (GA) and simulation method are employed to solve this problem. A GA can obtain a practical solution for discrete and continuous mathematical programing with rapid convergence. Its solutions are not guaranteed to be an optimum ones but they are sufficient in practice as feasible solutions. In the use of GA, formulation of a fitness function occupies an important role. However, behavior of AGVs is not well formulated, and therefore AGV driving simulation is adopted in order to estimate the fitness function. By use of simulation, it becomes unnecessary to formulate the restriction on AGV passing and collision because AGV behavior is taken consideration in simulation rules. The proposed method can deal with a large size of AGV scheduling (5∼30) and numerical experiments show that the proposed method provides solutions sufficient for practical use.