Mar 3, 2017 | Atlanta, GA
School of Interactive Computing Associate Professor Karen Liu earned a Google Research Faculty Award for her research titled Closing the “Reality Gap”: A Machine Learning Approach to Contact Modeling.
The research addresses the problem that robotic applications working in a simulation setting often struggle to learn motor skills. As a result, they can often perform poorly on physical hardware due to inaccurate parameters, idealized dynamic and contact models, or other un-modelled factors. Her research proposes to accurately compute contact states – like sticking, sliding, or breaking – and contact forces such that the simulated results will match the real-world phenomena.
“Our approach constructs a data-driven model that utilizes real-world observations to improve the accuracy of simulation,” Liu wrote in the abstract of her research proposal. “The key insight is that the contact problem can be broken down to two steps: predicting the next state of each contact point and calculating contact forces based on the prediction and current dynamic state.”
As a proof-of-concept demonstration, Liu plans to show that a humanoid can perform tasks involving whole-body dynamic balance in the real world using the control policy trained by the improved simulator.
The award will fund one graduate student for one year. Liu is one of two recipients of the Google Research Faculty Award at the Georgia Institute of Technology, the other being fellow IC faculty member Devi Parikh.