Bandit based monte-carlo planning
ECML'06 Proceedings of the 17th European conference on Machine Learning
Hi-index | 0.00 |
Advanced Driver Assistance Systems (ADAS) operate more and more autonomously and take over essential parts of the driving task e.g. keeping safe distance or detecting hazards. Thereby they change the structure of the driver's task and thus induce a change in driver's behavior. Nevertheless it is still the driver who is ultimately responsible for the safe operation of the vehicle. Therefore it is necessary to ensure that the behavioral changes neither reduce the controllability of the vehicle nor the controllability of the hazardous events. We introduce the Threshold Uncertainty Tree Search (TUTS) algorithm as a simulation based approach to explore rare but critical driver behavior in interaction with an assistance system. We present first results obtained with a validated driver model in a simple driving scenario.