Many-Objective Optimization of a Hybrid Car Controller

Abstract

Hybrid cars are a promising approach for providing individual mobility with lower $CO_2$-emissions and without compromising on affordability and driving range. In order to reach these targets a highly efficient control (energy management) is required. In mass production vehicles control is often organized using simple, quick, and easy to understand rule-based systems. Such a rule-base typically contains a moderate number of parameters which can be tuned using methods like evolutionary algorithms to improve performance. However, prior work basically targets a minimization of fuel consumption. In this work we present a many-objective evolutionary optimization that considers up to 7 objectives in parallel. We outline the additional optimization challenges that arise due to the large number of objectives and demonstrate that a substantial performance increase, over all objectives, can be achieved.