Social Force Aggregation Control for Autonomous Driving with Connected Preview

Abstract

Vehicular social force modeling (SFM) aims at capturing social driving behavior of human drivers in traffic by drawing analogies to models of psychological forces that govern the navigation behavior of humans walking in crowds. In addition, the extended preview afforded by vehicular connectivity can be exploited to bring additional information about downstream traffic to be incorporated in the planning and guidance computations for an autonomous vehicle. This paper outlines a hierarchical vehicular social force control scheme that integrates both ideas. At the upper level, social force aggregation is applied to predictively select the most efficient lane over a long horizon covered by connectivity. This is then passed down to a lower level controller that enforces lane tracking while considering higher fidelity social force resolution and lane-changing dynamics within the shorter horizon captured by the ego vehicle s sensor field of view. The workings and performance of the proposed framework are illustrated via simulations of the connected autonomous vehicle in multi-lane highway scenarios.