Acceleration of Experiment-Based Evolutionary Multi-objective Optimization of Internal-Combustion Engine Controllers Using Fitness Estimation

摘要

Control parameters of automotive engine have to be adjusted adequately and simultaneously to achieve plural criteria such as environmental emissions, fuel-consumption and engine torque. Evolutionary Multi-objective Optimization (EMO) is expected to be a powerful optimization framework for these engineering designs. Additionally, a smart environment called Hardware In the Loop Simulation (HILS) has recently become available for the engine calibration. To make Experiment-Based EMO (EBEMO) using the HILS environment feasible, the most important pre-requisite is reduction of the number of necessary fitness evaluations. In this paper, we apply an acceleration method using fitness estimation to overcome the aforementioned problem for EBEMO of real internal-combustion engines, and the effectiveness of our proposal is examined through real engine experiments.