Abstract: A new intelligent optimized dispatching method is proposed, and reinforcement learning control is applied in elevator group control system, in which CMAC neural network based on traffic pattern recognition is designed as the controller, in order to optimize the passengers’ average waiting time. This method can train weights in neural network on-line, not only without many expert knowledge and learning samples, but also with stronger adaptive ability. As a result, the system efficiency is improved, and the system performance is optimized. The simulation is performed under the pattern of interbedded traffic, and the results show that the method is feasible and effective.