Short Term Prediction of a Vehicle's Velocity Trajectory Using ITS
Sprache des Titels:
SAE World Congress
Modern cars feature a variety of different driving assistance systems, which aim to improve driving comfort and safety as well as fuel
consumption. Due to the technical advances and the possibility to consider vehicle-to-vehicle (V2V) and vehicle-to-infrastructure
(V2I) communication, cooperative adaptive cruise control (CACC) strategies have received significant attention from both research
and industrial communities.
The performance of such systems can be enhanced if the future velocity of the surrounding traffic can be predicted. Generally, human
driving behavior is a complex process and influenced by several environmental impacts. In this work a stochastic model of the velocity
of a preceding vehicle based on the incorporation of available information sources such as V2I, V2V and radar information is
presented. The main influences on the velocity prediction considered in this approach are current and previous velocity measurements
and traffic light signals. For practical applications a model must capture the driver's reaction on the traffic situation as well as the
vehicle dynamics. Here a Bayesian network approach is followed which provides a compact representation of the variable
dependencies and allows inferring the mean value and the confidence interval of the prediction. The proposed model is parameterized
with real traffic scenarios recorded during road tests. The validation results show that the prediction achieves a high accuracy within a
prediction horizon useful for a variety of different driver assistance functions.