Abstract
Risk predictive driving is essential to maintain safety while passing a parked car (passing-a-parked-car scenario), which is one of typical high risk traffic situations because a pedestrian might suddenly dash out from behind the car. In this study, we propose a novel risk prediction method for generating a reference trajectory in the passing-a-parked-car scenario. This paper defines the risk as collision speed with AEB (Autonomous Emergency Brake) against a simulated pedestrian dashing out from behind the parked car. Because the collision speed is theoretically determined depending on vehicle position and speed, the risk is defined in vehicle state (position and speed) space. Plotting vehicle state sequences of actual driving data in the space, we quantify risk at each moment in passing-a-parked-car scenario and show that vehicle state sequences of risk predictive driving can be plotted in zero-collision-speed region in the space.
