Integrated task and motion planning (TAMP) is an approach for solving robotics problems by conducting a two-level search over solutions, first over discrete actions and then over continuous commands to implement the robot movements. Recent approaches employ machine learning techniques to restrict the search space over continuous commands. Current methods do not account for future actions that the robot will need to execute to solve the overall task; for instance, if a robot is tasked with stacking books on a shelf, it would place the first books anywhere on the shelf and not consider leaving room for future books. My work will incorporate foresight into such learning-based methods to enable more efficient planning.

I am participating in the SuperUROP because I want to gain more experience in the research of Robotics and Machine Learning. I have taken Machine Learning and Robotics courses in the past and want to expand my experience in more cutting-edge technologies in those fields. I hope to publish a paper by the end of the SuperUROP, if I have meaningful results to display.