Robust Autonomous Driving

Robust autonomous driving research aims to create vehicles capable of safe and reliable navigation in diverse and unpredictable environments. Current efforts focus on improving sensor fusion (e.g., integrating LiDAR, camera, and radar data using techniques like dynamic adjustment and masked BEV perception), enhancing model robustness against adversarial attacks (both physical and digital) through methods such as adversarial training and certification frameworks, and developing more efficient data collection and training strategies (including active learning and federated learning). These advancements are crucial for ensuring the safety and reliability of autonomous vehicles, paving the way for wider deployment and impacting transportation systems significantly.