Will Compton
Doctoral Student | California Institute of Technology
Github
Scholar
arXiv
LinkedIn
I am a PhD candidate in Control and Dynamical Systems at the California Institute of Technology, advised by Professor Aaron D. Ames. My research lies at the intersection of nonlinear control theory, machine learning, and robotics, with a particular emphasis on developing control frameworks for agile and robust locomotion in legged robotic systems. I have explored areas including the control of underactuated systems, model predictive control, reinforcement learning, and layered control architectures, and have applied these approaches to bipedal, quadrupedal, and monopedal robots. My work aims to bridge the gap between theoretical control concepts and practical robotic implementations.
Featured Project: Dynamic Tube MPC
Much of my recent research has focused on using data driven methods to model mismatch between reduced order models used for planning and full order systems. In my work on Dynamic Tube MPC, I optimized paths for the hopping robot through cluttered spacing using a single integrator model, while using a learned model of the tube dynamics to ensure the full system followed collision free paths. This allowed both dynamic and safe behaviors, eliminating the need for significant conservativism in traditional tube MPC.
Beyond Research
Outside of my academic and professional work, I really enjoy staying active and outside -- pickup sports (pretty much all of them), hiking/running, climbing, or any outdoor adventure will pique my interest. I also love to create with my hands -- woodworking, origami, legos, calligraphy, ... any physical creation is a great way for me to relax and recharge. When I'm not writing code or working on control algorithms, you might find me reading science fiction, solving a puzzle (sudoku, chess, go), or planning my next adventure.
