Explore the intersection of deep learning and control theory in autonomous systems through this 49-minute lecture by Claire Tomlin from UC Berkeley. Delve into the challenges of efficiently navigating an autonomous system using a monocular RGB camera in unknown environments. Examine success rates, control profiles, and comparisons with geometric mapping-based approaches. Gain valuable insights into safe learning for dynamics and control, and understand key lessons learned in the field of perception for robotics. Part of the Frontiers of Deep Learning series at the Simons Institute, this talk offers a comprehensive perspective on the application of deep learning techniques in autonomous system perception.
Using Deep Learning for Perception in Autonomous Systems - A Perspective from Control Theory
Overview
Syllabus
Using deep learning for perception in robotics: a perspective from control theory
Safe Learning for Dynamics and Control
Outline
How to efficiently navigate an autonomous system with a monocular RGB camera to a goal in an a priori unknown environment?
Success Rate
Control Profile
Comparison with Geometric Mapping-based Approaches
Some lessons learned
Summary
Taught by
Simons Institute
