Planning for long-horizon, high level tasks poses the interesting problem of learning to combine simple primitives in an efficient, yet flexible manner in order to overcome any problems that arise in the world. For example, take a robotic arm whose sensorimotor primitives include moving its hand, grasping items, and rotating its hand. How would a model learn to compose these primitives in order to make a cup of coffee? Much improvement upon existing, generalized solutions is needed before these algorithms can realistically be applied to an actual robot. The goal of my project is to improve upon existing solutions to the composition of sensorimotor primitives in robotics and find novel solutions to the challenges faced when applying these algorithms to a real robot.

“My academic experience at MIT has provided me with a solid foundation in various topics in the field of EECS. By participating in SuperUROP, I have the opportunity to apply my academic experience to real-world problems that are awaiting solutions, without the guidance found in the classroom environment. My time will be spent coming up with novel solutions to unsolved problems, and that is exciting to me.”