This paper addresses a multi-AGV flow-shop scheduling problem with a reinforcement learning method. Each AGV equipped with a robotic manipulator, operates on the fixed tracks, transporting semi-finished products between successive machines. The objectives dealt with here is to obtain a AGV schedule that minimize the average job delay and total makespan. After formulating such schedule problem as a Markov problem by defining state features, actions space and reward function, a new scheduling method is proposed, based on reinforcement learning. In this new method AGVs share full information on each machine's instant state and job being executed, making decisions thorough understanding of the entire flow shop. Simulation results demonstrate that this new method learns optimal or near-optimal solution from the past experience and provides better performance than multi-agent scheduling method in a dynamic environment.