Toward End-to-end Reliable Robot Learning for Autonomy and Interaction

Abstract: Robots must behave safely and reliably if we are to confidently deploy them in the real world around humans. To complete tasks, robots must manage a complex, interconnected autonomy stack of perception, planning, and control software. While machine learning has unlocked the potential for full-stack end-to-end control in the real world, these methods can be catastrophically unreliable. In contrast, model-based safety-critical control provides rigorous guarantees, but struggles to scale to real systems, where common assumptions, e.g., perfect task specification and perception, break down.

However, we need not choose between real-world utility and safety. By taking an end-to-end approach to safety-critical control that builds and leverages knowledge of where learned components can be trusted, we can build practical yet rigorous algorithms that can make real robots more reliable. I will first discuss how to make task specification easier and safer by learning hard constraints from human task demonstrations, and how we can plan safely with these learned specifications despite uncertainty. Then, given a task specification, I will discuss how we can reliably leverage learned dynamics and perception for planning and control by estimating where these learned models are accurate, enabling probabilistic guarantees for end-to-end vision-based control. Finally, I will provide perspectives on ongoing work and future opportunities in reliable and secure data-driven autonomy, including adversarially-robust perception-based control and integration with pre-trained models.
 

Bio: Glen Chou is an assistant professor at Georgia Tech in the College of Computing, within the School of Cybersecurity & Privacy (SCP), and in the College of Engineering, within the School of Aerospace Engineering (AE). He joined Georgia Tech in November 2024. Glen directs the Trustworthy Robotics Lab, which focuses on the design of algorithms that can enable general-purpose robots and autonomous systems to operate capably, safely, and securely, while remaining resilient to real-world failures and uncertainty. To achieve this, the lab leverages control theory and machine learning, while connecting to optimization, perception, formal methods, motion planning, human-robot interaction, and statistics. Glen received dual B.S. degrees in EECS and ME from UC Berkeley in 2017 and M.S. and Ph.D. degrees in ECE from the University of Michigan in 2019 and 2022, respectively. Prior to joining Georgia Tech in 2024, Glen spent two years as a postdoc at MIT CSAIL.