ADVENT: Advance Driver's in-VEhicle movemeNt Tracking algorithm for semi-autonomous assistive driving

Abstract

The alarming data on road accidents reveals the demand for advanced machine learning algorithms and computer vision to ensure adequate road safety. Various reports on road safety suggest that driver’s action accounts for the majority of road accidents. Also, recent research studies on assistive driving indicate that there is no single solution to the problem of how to assist the driver by anticipating his driving action in advance. Early anticipation of a particular action helps in assisting the driver by providing a few additional seconds to his reaction time for safe driving. This paper intends to improve the anticipation time and maneuver accuracy of five driving maneuvers like a left turn, right turn, left lane change, right lane change, and straight. The paper proposes an Advanced Driver in -Vehicle movemeNt Tracking (ADVENT) algorithm using the Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM). The driver’s eye gaze extraction gives better inference about his intentions, which improves the overall effectiveness of our system. Performance merit of ADVENT is compared with previous algorithms, and it shows a significant improvement in performance metrics.