Technology Commercialization in Partnership with Össur
We have worked with Össur, one of the leading manufacturers in prosthesis technology, to test our control algorithms on their commercial hardware. The video above shows our data-driven variable impedance controller deployed on the Össur Power Knee. We’re working on quantifying the clinical benefits of the Power Knee and investigating how these benfits depend on the control policy used.
Aug 15, 2024
Contact-Implicit Differential Dynamic Programming
This project was developed for my Legged Robots course at the University of Michigan taught by Prof. Yanran Ding in the Winter 2024 semester. This work mainly follows the publication from Kim et al. “Contact-Implicit Differential Dynamic Programming for Model Predictive Control with Relaxed Complementarity Constraints”, IROS 2022.
Apr 30, 2024
Data-Driven Impedance Control For Knee-Ankle Prostheses
The prevailing paradigm in lower-limb robotic prosthesis control is to use hand-designed controllers with heuristically tuned behaviors. While these approaches can be very effective, they are labor-intensive in both the design of the behaviors and in individualizing them to each individual at the point-of-care. In this work, we instead developed a data-driven control architecture that allowed the prosthesis to work over continuums of tasks without requiring manual tuning.
Jun 1, 2023
Control of a Planar Ballbot with Obstacle Avoidance
This project was developed for an Applied Optimal Control course taught by Dr. Christian Hubicki in the Winter 2023 semester. I wrote a controller for a ballbot, which is a small, wheeled robot that balances on top of a basketball. The controller uses trajectory optimization and convex model predictive control (MPC) to navigate the world while staying balanced and avoiding obstacles.
Apr 28, 2023
Open Source Leg
The Open-Source Leg project aims to accelerate research in the field of lower-limb robotic prostheses by providing a common research platform to the field. Senthur Ayyappan, a rockstar mechanical designer and software engineer, currently maintains and leads the project along with PI Elliott Rouse at the University of Michigan. Full CAD design files, software libraries, and other resources are available to the public to use free of charge. See www.opensourceleg.org for more information.
Jan 1, 2022
A Machine Learning Approach to Predicting Distal Leg Muscle Activation for Real-time Control
This work explores three machine learning approaches to predict muscle activations in the lower limbs during walking. This was a group project for our Machine Learning Course (EECS 545) in Spring 2021.
Apr 30, 2021