A recent breakthrough in motion planning via trajectory optimization finds smooth paths in high-dimensional complex configuration spaces while adhering to dynamic constraints and avoiding obstacles. While the algorithm can find globally optimal paths faster than sampling-based methods, it heavily relies on the more computationally expensive region generation. We believe we can reuse some static regions, such as the robot’s self-collisions, and add simpler obstacles more efficiently. The latter would allow fast planning when new static or maybe even dynamic obstacles get added.

Through this SuperUROP, I want to explore conducting research and publishing before diving blindly into graduate school. I have worked on the motion planning approach my work would be based on and experienced the need for such methods in mobile manipulators in the industry. I hope that regardless of the outcome, I can publish a paper and directly impact how things are done in the industry.