Abstract: Autonomous driving has made significant progress in structured settings such as highways and well-marked urban roads. However, the real challenge lies in enabling self-driving vehicles to operate safely and reliably in dense, heterogeneous, and unstructured environments—contexts where traffic patterns are unpredictable, road infrastructure is incomplete, and interactions with diverse agents such as pedestrians, cyclists, and non-standard vehicles are frequent.
In this talk, Dr. Saket, Dr. Sanjit, and Dr. Rohan Chandra from the University of Texas at Austin will explore the latest research directions aimed at addressing these complexities. The discussion will highlight how advanced perception systems, robust decision-making algorithms, and adaptive control strategies can be combined to navigate environments that defy conventional assumptions about order and predictability.
Key themes include:
Challenges of perception and prediction in crowded, multimodal traffic scenarios.
Designing algorithms that can adapt to uncertainty and incomplete information.
Leveraging reinforcement learning and simulation frameworks to train autonomous systems for real-world variability.
Safety, reliability, and ethical considerations in deploying autonomous vehicles in unstructured settings.
By bridging theoretical advances with practical case studies, the speakers will illustrate how autonomous driving research is evolving to meet the demands of complex environments. The presentation will emphasize the importance of resilience, adaptability, and human-centric design in shaping the future of autonomous mobility.