Machine Learning for Autonomous Vehicle Behavior Prediction

Abstract

In recent years, significant advancements in sensing technology, computation, and machine learning algorithms have led to breakthroughs in the development of autonomous or self-driving vehicles. State-of-the-art systems can effectively perceive the environment, plan future actions, and execute them without human intervention. One important function of an autonomous driving system is to ensure the safety of the vehicle and its passengers. The aforementioned perceptual stack, which takes inputs from sensors and processes them to form a three-dimensional map of the car's environment, elicits a new set of capabilities: accurate localization of other road participants, from which we can predict their future actions and thus safely navigate among them. Accurate and real-time prediction of other road users has therefore become a key area of interest in autonomous vehicles.