Colloquium: Teaching robots to explore unknown places using deep reinforcement learning

14 July 2017

Deep reinforcement learning has made some remarkable breakthroughs lately, and has challenged human experts in multiple gaming domains such as Go, Texas Hold'em, and many Atari games. However, the gaming domain has become such a widely used benchmark in the field, that most research seems to focus solely on this domain.

This talk reviews some possibilities and problems of exploiting Deep-RL in the domain of robotic exploration. For the most part, the talk focuses on raising research questions instead of answering them, but some initial insights from simulations are provided.

Friday 14 July 2017, 13:00 s.t. Room R-031, Department of Informatics

Tuomas Välimäki, from Tampere University of Technology