D3QN-based Elevator Scheduling Algorithm for Robots

Abstract

In this study, we proposed an elevator scheduling algorithm based on a Dueling Double Deep Q Network (D3QN) for robots. The rewards for the elevator car allocation decision are estimated based on the robots’ journey time, the number of floors an empty car traverses, and how the car allocation meets the robots’ priorities. The Robotics Middleware Framework (RMF) was adopted to be the simulator. The performance of the proposed algorithm was compared to an existing LOOK algorithm. The simulation results show that the proposed method outperforms the existing LOOK method in terms of the robots’ journey time and how the car allocation meets the robots’ priorities at the cost of a higher number of floors traversed by an empty car.