I’ve been at ICRA 2017 presenting my paper on “Information Theoretic MPC for Model-Based Reinforcement Learning”. This paper is one of five nominated for best conference paper at ICRA. This paper looks at combining an information theoretic variant of model predictive control with learning dynamics using neural networks. This enables us to control systems without having explicit knowledge of the system dynamics. The paper is available here:
Link: http://www.cc.gatech.edu/~bboots3/files/InformationTheoreticMPC.pdf
And here’s a video of our results:
This is another video focusing on the just the AutoRally results:
And here’s all of my co-authors after our paper presentation:
And some more of my pictures from the conference and the city of Singapore:
Grady Williams 