Driver Intent Prediction with Barrier Functions

Abstract

Path planning and collision avoidance algorithms for driver assist systems and autonomous vehicles require predictions of the future locations of other vehicles. In the absence of V2X communication, driver intent prediction algorithms provide this information. Intent prediction algorithms generally fall into two categories: those based on vehicle dynamics models and filtering, and machine learning based approaches. Both approaches have advantages and drawbacks, but two disadvantages both approaches share is the relative computational complexity and the likelihood of false positives and negatives. In the literature, barrier functions (BF) are used to enhance safety by providing forward invariance of an admissible set. In this paper, BFs are used to determine driver intent in a comparatively efficient manner with quick detection particularly at high speeds, in a way where the chance of false positives and negatives can be significantly reduced.