Explore a novel approach to scene understanding for autonomous systems in this 58-minute talk. Delve into the challenges of combining diverse information sources, background knowledge, and sensor data to comprehend dynamic environments. Examine current computer vision and deep learning techniques for object detection and localization, and understand their limitations in complex driving scenarios. Discover a new perspective that leverages external knowledge, representation learning, and neurosymbolic AI to address these challenges. Learn about potential future research directions and applications of this technology to enhance machine perception in autonomous systems. Gain insights from Ruwan Wickramarachchi, a Ph.D. candidate at the AI Institute, University of South Carolina, whose research focuses on improving context understanding in autonomous systems through expressive knowledge representation and neurosymbolic AI techniques.