Model Predictive Control for Automated Driving and Collision Avoidance

Abstract

The automotive sector has seen a rapid transition towards autonomous driving with an aim to achieve SAE Level 5 vehicle. The incessant drive for innovation has resulted in modern passenger cars equipped with plethora of control technologies such as ABS, VSC, AFS via EPS assist etc. all working respectively in various critical and non-critical scenarios to ensure safe and smooth driving at all times. These efforts have reduced the number of road accidents significantly over the past years, yet it was observed that the number of rear-end crashes have not decreased but has rather maintained its proportion which is about 1/3rd of all road accidents. An evasive maneuver, a limit-handling situation, involves quick steering or braking action to avoid colliding with the vehicle in-front (thus avoiding a rear-end crash). A normal human driver is not trained to drive and control car in such demanding and high stress inducing task. This warrants the needs of an autonomous controller design that can itself take over the control of car and perform the maneuver successfully. The aim of this research was to verify this statement by designing a novel control scheme that can steer and brake simultaneously ensuring collision avoidance and passenger safety at all times. The controller designed uses the state-of-the-art Model Predictive Control (MPC) scheme such that best possible performance can be extracted even in critical driving scenarios. Variety of simulations were performed to successfully conclude that MPC worked well and is robust in design as well.