Ian Abraham and Todd Murphey Win the 2019 Best Paper Award for the IEEE Transactions on Robotics

The paper "Active Learning of Dynamics for Data-Driven Control Using Koopman Operators," by PhD student Ian Abraham and his advisor Todd Murphey, was selected the recipient of the 2019 King-Sun Fu Best Paper Award for the best paper appearing in the IEEE Transactions on Robotics in 2019.  The editorial board was "impressed by the principled approach to robot learning and the depth and breadth of the experimental results."  

The award-winning paper describes an innovative algorithm for fast robot learning. Under Abraham and Murphey's approach, the robot performs actions that trade off the objective of quickly learning the dynamics of the task ("active learning" by exploratory actions) with the objective of performing the task efficiently. Fast learning is critical for robots, since experiments in the real world are expensive, unlike in simulation. The paper demonstrates the value of the active learning approach on examples such as stabilization of a falling quadrotor drone, control of a spherical robot rolling on a sandy surface, and control of a seven-degree-of-freedom robot arm.