Probabilistic collision estimation system for autonomous vehicles

Nearly 1.3 million people die each year in traffic-related accidents, whereas an additional 20-50 million peopleare injured. Introducing autonomous vehicles would aim to reduce these numbers by removing the driver from the loop entirely, and thus removing the human error. Intersections are considered a complex traffic situation for autonomous vehicles. Functions which could accurately foresee future events in those situations, mimicking the situation awareness of humans, would improve autonomous systems and increase traffic safety. To address this a system is designed with two main functionalities: estimate the movements of observed vehicles in a general traffic situation and predict the probability of a collision, given the current ego trajectory. This system could either be used as information and feedback for a trajectoryplanner or as support for decision making at higher level system monitoring. The main contributions are the robust system design, that robustly and consistently estimates the likelihood of a collision and thus preventing future collision, and the intention estimation which determines the probability of which route through an intersection an observed vehicle will take through an intersection by using its current state. The system is validated by controlling the ego vehicle’s velocity with a Velocity Planning Controller to avoid colliding. It is shown that in terms of robustness to noise the system successfully avoids collision.QC 20190429