Evolutionary Optimization of a Controller for Multi-Car Elevators

We propose a new optimization method, that is found effective for the difficult problem of controlling multi-car elevators (MCE). The important new feature of our method is the introduction of a non-linear policy expression, which is optimized by simulation, using a genetic algorithm. We show that our method outperforms substantially earlier methods that used linear-sum policy expression. We also provide an analysis of the simulation results, and show that the improved performance is attributable to the more flexible handling of bunching situations. This flexibility was made possible by the non-linear policy.