Survey Lecture: Sebastian Trimpe: Event-Triggered Learning
Thursday, May 04, 2023, 12:30pm
Location: Computer Science Center, Building E2, ground floor, B-IT room 5053.2.
Speaker: Sebastian Trimpe
The ability to learn is an essential aspect of autonomous systems facing uncertain and changing environments. However, the process of learning a new model or behavior often does not come for free, but involves a certain cost. For example, gathering informative data often means to deviate from normal operation, it may be challenging due to physical limitations, or updating models can require substantial computation. Hence, the question of "when to learn?" is essential for the efficient and persistent operation of autonomous systems.
In this talk, I will be discussing recent research on the concept of event-triggered learning (ETL), which we have proposed for making principled decisions on when to learn a new dynamics model or controller. We will present some specific trigger designs that trigger learning when needed, and also highlight experimental application leading to significant resource savings. Moreover, I will give a personal "historical" perspective and trace back, how our recent research was itself triggered by my own PhD research and papers. I will also dare to look into the future and discuss what potential "milestone" papers there might still be to come.