Machine Learning for Enhancing Autonomous Vehicle System Robustness

Abstract

Autonomous vehicles (AVs) have started to attract notable attention due to various potential benefits that can revolutionize our current transportation system. Generally, an autonomous vehicle can be defined as an automated vehicle capable of sensing its environment using an array of sensors like radar, lidar, GPS, odometry, IMU, etc., and navigating without human input. This requires a rich variety of sensor technologies, significant computational power for processing the various input data, and an extremely reliable perception system for analysis. While the convenience of moving from point A to point B with minimal human input is attractive, the enabling technology required in next-generation vehicles is based on more advanced machine learning techniques for robust decision-making to control these next autonomous vehicles. Given the complexity of engineered systems that are governed by multiple variables, machine learning has attracted attention in the engineering domain for optimizing vehicle performance or conducting predictive analysis for various vehicle system failures or relevant diagnostics.