Towards risk estimation in automated vehicles using fuzzy logic

Abstract

As vehicles get increasingly automated, they need to properly evaluate different situations and assess threats at run-time. In this scenario automated vehicles should be able to evaluate risks regarding a dynamic environment in order to take proper decisions and modulate their driving behavior accordingly. In order to avoid collisions, in this work we propose a risk estimator based on fuzzy logic which accounts for risk indicators regarding (1) the state of the driver, (2) the behavior of other vehicles and (3) the weather conditions. A scenario with two vehicles in a car-following situation was analyzed, where the main concern is to avoid rear-end collisions. The goal of the presented approach is to effectively estimate critical states and properly assess risk, based on the indicators chosen.