Safe Control and Learning under Uncertainty

Autonomous systems are being widely deployed in safety-critical applications, for example, in autonomous driving and intelligent industrial and service robotics. As there are large uncertainties in these applications (e.g., autonomous vehicles interacting with unknown road participants), it is important to study how to safely control the autonomous systems, how to learn from the interactions with the environment to minimize uncertainty, and how to perform the learning safely and efficiently. This workshop aims to bring together researchers who work in the field of safe control and learning under uncertainties.

We will discuss recent progress in the development of energy-function-based safe control methods (including control barrier function, safe set algorithms, potential field methods, sliding mode methods, etc.) and safe learning controllers that use these safe control methods as safety monitors or safety shields. Detailed topics include and not limited to: how to guarantee safety under uncertainty, how to efficiently synthesize or learn control barrier functions for unknown dynamic systems or learned dynamics encoded in deep neural networks, how to effectively quantify uncertainty for unknown systems, how to ensure safety during both exploitation and exploration, and how to ensure sufficient exploration during safe learning, etc.

This virtual workshop will take place online prior to the inaugural Modeling, Estimation and Control Conference (MECC 2021).

This workshop is free to students. Please register using this google form by Oct 15.