Comfort-based AI-driven Roundabout Control for Automated Vehicles

Automated vehicles are emerging with the advancement of driving algorithms and improvement in their sensing, computing, and navigating capabilities. All leading stakeholders are developing prototypes and testing them on roads. If safety is of uttermost importance, efficiency and passenger comfort are becoming critical to their success. In that context, roundabouts are particularly challenging due to the difficulties to interpret the situation and inject into traffic, leading to stop-go situations, inefficient waiting time, and uncomfortable jerk values. In this paper, we present an AI-driven comfort-oriented roundabout control mechanism for automated vehicles. We propose a throttle and steering control respecting passenger comfort metrics, an AI-driven situational analysis to optimize yielding, and a model predictive control algorithm to adjust maneuvers and minimize waiting time at roundabouts.