Research of elevator group scheduling system based on reinforcement learning algorithm

Abstract

Elevator group control system (EGCS) is a complex decision-making system, which has characteristics of multi-objective, randomness and nonlinear. It is difficult to adopt precise mathematical models describing. This paper introduces a new elevator dynamic scheduling system based on reinforcement learning algorithm. We trade reinforcement learning algorithm as the way to learn the optimal strategy in the course of interacting with the environment. Average waiting time and average riding time are optimized indicators. Combine with the value iteration algorithm called Q-learning to construct the whole algorithm for elevator group scheduling. The simulation result shows great superior and feasibility for elevator dynamic scheduling system based on reinforcement learning algorithm.