Supervised Learning via Optimal Control Labeling for Criticality Classification in Vehicle Active Safety

摘要: A core component of vehicle active safety algo-rithms is the estimation criticality, which a measure threat or danger traffic situation. Based on criticality esti-mate, an system can significantly increase passenger by triggering collision avoidance mitigation maneuvers like emergency braking steering. Interpreting as intensity evasion maneuver, we formulate MinMax optimal control problem incorporates moving obstacles and clothoidal lane constraints. We show how solution this be used labeling function to generate reference data sets for scenes. In order achieve fast execution speeds, present supervised classification approach estimation. Using Random Forest classifier with feature selection, that combined steering predicted high precision.