An approach for predicting vehicle velocity in combination with driver turns

Abstract

Statistical information such as average traffic flow on a road network link are not precise enough to enable detailed in-vehicle optimization strategies. Energy efficiency measures are further restricted if the vehicle is not aware of the destination of the trip. A modular and dynamic approach for predicting the vehicle speed in combination with driver turns is introduced. The objective is to adapt the vehicle components to the upcoming speed and state. The cloud-based approach foregoes the user-based selection of a route, trip destination or individual points of interest but is mainly based on historic speed profiles of different drivers. The concept provides a two-stage approach. The first stage is the prediction of upcoming turns and trip segments based on the historical features and the currently driven road segments. The second stage uses this information to predict the vehicle's speed.