Abstract
The unprotected left turn of a connected automated vehicle (CAV) is investigated when it has a potential conflict with a connected human-driven vehicle (CHV) approaching in the opposite lane. A control architecture is proposed that includes interactions between the decision making, motion planning, and control levels. By utilizing the road context and information received via vehicle-to-everything (V2X) communication, a reduced state space representation is determined which allows the CAV to evaluate safety in a fast and efficient manner Using a temporal metric, a safety evaluation algorithm is developed which determines the safety of the decision making at controller level. To evaluate the algorithms, data collected with real vehicles is utilized.
