Modelling of traffic situations at urban intersections
- Ansprechpartner: M.Sc. Hong Quan Tran
Understanding of traffic situations is an essential part of future advanced driver assistance systems (ADAS). Therefore, we are currently doing research on a practical framework for modeling traffic situations at urban intersections, which can handle two problems: maneuver recognition and trajectory prediction of moving vehicles.
Probabilistic non-parametric regression models
The spatio-temporal dependencies of traffic situations are modeled with three-dimensional Gaussian process regression models, which are learned from two dimensional trajectory patterns.
Maneuver recognition and trajectory prediction
Driving maneuver can be recognized by evaluating the data likelihood for each individual regression model. Furthermore, multiple-step ahead trajectory prediction can be performed by employing the Monte Carlo method.