Simulation-based optimization of a controller for multi-car elevators using a genetic algorithm for noisy fitness function

摘要

Recognizing progress in linear motor technology, multi-car elevators (MCE) that have several cars in a single elevator shaft attract attention as a novel transportation system for high-rise buildings. Sudo et al. have demonstrated that a genetic algorithm can find good control strategies through simulation-based optimization for MCE systems. However, it takes very long computation time because evaluation of a control strategy is obtained through complex discrete event simulation of a MCE system. This paper discusses acceleration of the simulation-based optimization of MCE controller using a genetic algorithm for noisy fitness function and a PC cluster system. With these devices, optimization of the controller and evaluation of the obtained results are carried out within practical computation time, and it enables detailed investigation of a control scheme for MCE.