PILLAR is a robotics research project that aims to integrate model-based planning methods with experience-based learning methods into one general, robust, and adaptive framework with sound theoretical guarantees.
Despite recent advances in robotics, learning, and planning, robots still face significant challenges operating outside of the lab and in the real world, with unknown environment conditions, unmodeled objects and disturbances, and tasks that need skills which the robots do not yet possess.
Our goal is to tackle these challenges with the PILLAR framework and enable robots to effectively leverage prior knowledge, perform life-long learning of models and skills, and explore and accomplish tasks in unstructured environments.
This joint project is led by Professor Maxim Likhachev and Professor Oliver Kroemer from Carnegie Mellon University’s Robotics Institute, in collaboration with Professor Dieter Fox from University of Washington.
PhD students on the project include: