RESEARCH ON ELEVATOR GROUP SCHEDULING STRATEGY AND SIMULATION BASED ON REINFORCEMENT LEARNING ALGORITHM

Rui Tian and Weimin Gao

Keywords

Elevator group control systems, adaptive multi-objective optimisation, reinforcement learning algorithm

Abstract

The multi-objective elevator group optimisation problem has attracted widespread attention due to its high practical significance. The elevator group control system (EGCS) is a typical multi- objective system designed to increase passenger service and reducing costs, such as energy consumption (EC). Based on the characteristics of elevator group scheduling problem, this paper treats EGCS as multiple reinforcement learning agents that cooperate with each other. Each agent controls the operation of an elevator according to the average reward reinforcement learning algorithm (RLA), aiming to minimise the average waiting time (AWT) of passengers. Additionally, a neural network is employed to store and update the state behaviour value functions while dynamically classifying the state of unknown environments. The simulation results show that the reinforcement learning based-scheduling algorithm has better scheduling effects than the static partition scheduling algorithm (SPSA) and cultural algorithm (CA).

Important Links:



Go Back