Jaeyoun Choi

I'm a Ph.D. student at MIT advised by Professor Chuchu Fan. Before I joined MIT, I served as the Chief Technical Officer at Neumafit Corporation, where I focused on developing an AI-powered wearable mask for cardiorespiratory analysis. Prior to that, I completed my B.S. and M.S. in Mechanical Engineering at Seoul National University, conducting research at the Soft Robotics and Bionics Laboratory.

My long-term research goal is to develop robots as a universal solution for all tasks requiring physical interaction. Multi-robot systems hold immense potential for enhancing robotic capabilities. Specifically, in my Ph.D. studies, I aim to explore multi-robot collaborative systems where multiple robots physically cooperate to transport and manipulate objects with safety, robustness, and adaptability.

Email  /  CV  /  Scholar

A team of drones carries a shared payload on compliant cables and delivers it to a goal pose while avoiding obstacles sensed onboard by LiDAR. I trained the policy with a CBF-based safe reinforcement learning method, randomizing both cable stiffness and team size during training, so that a single policy scales to any team size and transfers to hardware zero-shot with no real-world fine-tuning.

A sampling-based predictive controller that stays safe while it learns unknown system parameters online, hedging against parametric uncertainty instead of assuming a nominal model.

A modeling approach that couples known physics with a learned residual network, so that the model remains physically consistent where the dynamics are known and data-driven only where they are not.

I developed a wearable robot that assists various lifting tasks and enhances spinal safety. I verified the BBEX with human subject experiments and musculoskeletal analysis using OpenSim. I theoretically analyzed the potential to reduce risk factors for back injuries.

I derived the theoretical model of the novel compliant linear actuator named TERRA and analyzed its characteristics.

I fabricated the self-folding elastomer sheets using polymer molding and CNC machining. I conducted tensile tests and baking experiments to analyze the repeatability of the polymer.

I developed deep-learning algorithm based on gated recurrent unit (GRU) for VO2 and VCO2 estimation. This algorithm is implemented in Neumafit Pacer.

I designed an inertia-driven swimming robot mimicking the C-start maneuvers of fish. I integrated a LiDAR sensor into the swimming robot, enabling it to recognize and rapidly avoid approaching obstacles.