DIKU Bits: Efficient Robot Learning under Uncertainty

Speaker

Hang Yin, Assistant Professor - Tenure Track in the Image Analysis, Computational Modelling, and Geometry section (IMAGE)

Title

Efficient Robot Learning under Uncertainty

Abstract

Imagine a robot opening your dishwasher, taking out cleaned cutlery and placing them back into a cabinet. Such a task is seemingly less intelligence demanding than creating legible texts and image arts. Then why haven’t we got such AI systems? In this talk, I will present my opinions on the difficulties in learning embodied agents that live in a physical world. I will review some recent research efforts on acquiring robot skills through data-driven methods. Sharing with you my own research, I will discuss how our knowledge in modelling and controlling other cyber-physical systems may help in building efficient learning algorithms.

Which courses do you teach?

Vision and Image Processing (VIP)

Simulation-based Reinforcement Learning (SiRL)

Which technology/research/projects/startup are you excited to see the evolution of?

New design, computing methods, and applications involving robots and broader AI systems.