Learning and Prediction based Control for Robotic Systems

Contributor: Sang-ik An

Making the robotic systems or the articulated and constrained mechanical systems adapt and interact to the uncertain objects or environments is still a hard problem to solve. Specifically, the control policies need to be found that allow the robot to perform given tasks on the existence of uncertainties. The difficulty arose on this topic is that complexity and uncertainty of constrained dynamic systems have to be dealt with on the same time. A possible approach is to use the past and future information when determining the feasible set of the current control policy. The uncertainty can be learned from the history and be used to predict the possible scenarios in the near future. Then, the control policy can be searched among them to achieve the best result for the tasks.


Publications

Peer-reviewed conference papers

  1. S. An and D. Lee, "Prioritized Inverse Kinematics with Multiple Task Definitions," in IEEE International Conference on Robotics and Automation (ICRA), 2015.
  2. M. Saveriano, S. An, and D. Lee, "Incremental Kinesthetic Teaching of End-Effector and Null-Space Motion Primitives," in IEEE International Conference on Robotics and Automation (ICRA), 2015.
  3. S. An and D. Lee, "Prioritized Inverse Kinematics using QR and Cholesky Decompositions," in IEEE International Conference on Robotics and Automation (ICRA), pp. 5062-5069, 2014.

International workshops

  1. S. An and D. Lee, "Inverse Kinematics with Multiple Tasks and Multiple Task Definitions," in International Workshop on Human-Friendly Robotics (HFR), 2015.