Dispatching algorithm design for elevator group control system with Q-learning based on a recurrent neural network

A dispatching algorithm of elevator group control system is proposed based on reinforcement learning. Elevator dispatching is modeled by a Markov Decision Process. Then an internally recurrent neural network based reinforcement learning method is designed to find the optimal dispatching policy while the state-action value function is iteratively approximated. Finally, several simulated experiments are done to compare the trained dispatching policy with other traditional ones. The experimental results demonstrate the effectiveness of proposed dispatching method.