A Possibilistic Risk Assessment Framework for Unmanned Electric Vehicles With Predict of Uncertainty Traffic

Abstract

At present, electric vehicles (EV) have entered a stage of rapid development. Meanwhile, with artificial intelligence (AI) technology fast improving and implementing many inventions in electric vehicles (EV), almost all EV sold in China are equipped with automatic driving technology to achieve safer and more energy-saving driving. In order to solve the problem of anti-collision in self-driving Smart EV under complex traffic, especially at intersections, most of the existing methods make sequential predictions for the driving level of vehicles, and it becomes difficult to deal with sudden changes in intentions of other vehicles. Therefore, a collision risk assessment framework based on other vehicles’ trajectory prediction is proposed. The framework integrates the solutions of other vehicles’ expected path planning, uncertainty description of driving process, trajectory change caused by obstacle intrusion, etc., as well as adopts the Gaussian mixture model to evaluate the risk according to the probability of collision. It realizes the real-time evaluation of the probability of collision and makes safe decisions and trajectory planning of the vehicles. After simulation and verification, it effectively solves the decision-making planning problem of autonomous vehicles under complicated traffic flow and demonstrates that the method is better than the current sequential prediction method (SORT\Karlman filter, etc.).