A Distance-Based Two-Stage Ecological Driving System Using an Estimation of Distribution Algorithm and Model Predictive Control

This paper proposes a distance-based two-stage ecological (eco-) driving scheme by using estimation of distribution algorithms (EDA) and model-based prediction of traffic conditions. Before departure, the optimal speed profile for an entire route is generated by an EDA in combination with speedup approaches for a faster computing time, which can optimize the complex cost function of ecodriving without simplification within a reasonably short computing time. This optimization is performed in a distance domain for localizing changes in the optimal speed profile due to traffic conditions while driving. After departure, by taking the optimal speed profile and actual traffic conditions into consideration, the speed profile for a short term-to only the next location-is adapted. In order to reliably react to actual traffic conditions, additional points are interpolated into the long-term distance step and fine control of speeds at the additional points is established, which is based on a predictive model for estimating the spacing to the preceding vehicle. The proposed ecodriving system is evaluated in two types of route conditions, and its results are compared with the optimization result by the quadratic programming method. This comparison shows that an EDA can generate a speed profile with better optimization results in terms of fuel efficiency and driving time within a shorter computing time.