I am a Ph.D. candidate in Mechanical Engineering at the University of Michigan working in the ROAHM Lab under Professor Ram Vasudevan. I'm interested in using control theoretic techniques to analyze human motion, as well as computer vision methods for extracting models of motion. Currently, my research focuses on building models of human motion and using reachability analysis to characterize human stability during walking and sit-to-stand. I also conduct perturbative biomechanical experiments to verify the accuracy of our stability estimates.

I'm originally from Goleta, California and received a B.S. in Mechanical Engineering from UC Berkeley in May 2015.

[Google Scholar]


Modeling human motion

I work on building personalized dynamic models of human motion from kinematic data. Currently, this type of analysis is confined to a laboratory environment with expensive motion capture equipment. I'm interested in developing markerless and inexpensive camera-based methods for modeling human motion.

Sit-to-stand stability

Fall risk can be assessed by examining an individual's stability. However, stability is difficult to objectively quantify. Here, I used reachability analysis to characterize the stable region of the sit-to-stand motion for 11 test subjects. We perturbed subjects with a motor-driven cable pull as they stood up from a chair, and confirmed that our analysis accurately predicted the instances where the subject stepped or sat back down in response to perturbation.

Walking stability

I'm working on extending the techniques we developed for sit-to-stand to analyze stability during walking. Walking is a more complicated motion, and requires a model that is hybrid in nature due to the person's intermittent contact with the ground. I'm analyzing data from an experiment in which subjects were pulled laterally at random points in their stride cycle, under a number of different step width and velocity constraints.


ACL reconstruction surgery requires extensive rehabilitative care performed under the eye of a physical therapist or clinician. I want to help automate the rehabilitation process by providing healthcare specialists with quantitative estimates of important rehabilitation criteria. In the picture above, we coupled a stereo camera with publicly available pose estimation software to accurately estimate the forces produced by patients as they completed a leg press exercise.


"Automated Camera-Based Estimation of Rehabilitation Criteria Following ACL Reconstruction." (Choong Hee Kim, Shannon M. Danforth, Patrick D. Holmes, Daphna Raz, Darlene Yao, Asheesh Bedi, Ram Vasudevan), arXiv preprint, 2018. [PDF]

"Stability basin estimates fall risk from observed kinematics, demonstrated on the Sit-to-Stand task." (Victor Shia, Talia Yuki Moore, Patrick Holmes, Ruzena Bajcsy, Ram Vasudevan), in Journal of Biomechanics, 2018. [PDF]

"Convex estimation of the α-confidence reachable set for systems with parametric uncertainty." (Patrick Holmes, Shreyas Kousik, Shankar Mohan, Ram Vasudevan), in IEEE Conference on Decision and Control (CDC), 2016. [PDF]


"Perturbative Sit-to-Stand experiment for validating quantitative method of stability estimation." (Patrick Holmes, Shannon Danforth, Talia Yuki Moore, Xiao-Yu Fu, Ram Vasudevan), in Dynamic Walking, 2018 and in World Congress of Biomechanics, 2018. [video]

"Human Feedback Control to Maintain Trajectories of Task-Relevant Variables During Sit-to-Stand Motion" (Patrick Holmes, Shannon Danforth, Talia Yuki Moore, Xiao-Yu Fu, Ram Vasudevan), in Dynamic Walking, 2017

"Direct perturbation on humans performing sit-to-stand motion reveals corrective feedback control strategy." (Patrick Holmes, Victor Shia, Talia Yuki Moore, Ram Vasudevan), in Society for Integrative and Comparative Biology Annual Meeting, 2017.

Contact Me

Feel free to contact me at pdholmes@umich.edu