Research

Simulating the effects of age on gait and fatigue

I am developing models that explain why older adults walk slower and get tired faster than their younger counterparts. I use optimal control methods to generate plausible gaits for simulated older adults using OpenSim Moco, as well as feedback controlled gait using the SIMULINK based reflex model of walking, as well as the SCONE/HYFYDY implementation of the reflex model.

As part of this work I have built custom goal functions, and custom actuator models for OpenSim Moco in C++ and developed methods to use these custom functions in Matlab.

The broader project includes a collaboration with the MOBL lab at UMass, Amherst to collect experimental data with the eventual goal of using inverse methods to make our models walk like actual humans.

So far, I have presented work on this project at NACOB 2022, ASB 2023 and the Department of Kinesiology seminar at UMass, Amherst.

Studying ankle impedance in stroke survivors

This work is focused on understanding the impact of rehabilitation techniques such as Botox therapy on the gait of stroke survivors. I modified and controlled the Pertuberator robot designed by Prof. Elliott Rouse to measure the mechanical impedance of the ankle joint for healthy as well as impaired individuals. I also collected kinematic, kinetic and EMG data as well as measures of strength and muscle activation for 11 subjects.

While the BoNT side of this research was cut short due to the 2020 pandemic, we determined that - 

Understanding gait stability and human-structure interaction

I designed and conducted experiments to apply perturbations to human participants as they walked on a treadmill. I measured kinematics and ground reaction forces as the participants returned to steady-state walking, and used system identification techniques to build controllers that recreate human-like dynamics in simple models responding to perturbations

We developed controllers that allow simulated bipeds to recover from perturbations in a manner dynamically similar to how humans recover from perturbations.

I studied the large lateral oscillations observed on footbridges and walkways and developed models to explain this phenomenon. These models were built using trajectory optimal control to determine the energy minimal way to walk on a shaky bridge and linear foot placement and push-off controllers to simulate hundreds of bipeds walking on shaky bridges.

We simulated hundreds of bipeds walking across a pedestrian bridge and determined the conditions necessary for the bipeds to shake the bridge and synchronize with each other.

Agent based models of flocking and obstacle avoidance

I developed agent-based models of flocking that allow groups of animals to escape dead-end obstacles after being trapped in them. These agents follow very simple behavioral rules to determine their desired heading:

A combination of these rules allows agents to escape dead-ends while moving in a coherent group, but only when the agents desire to align with each other was very strong. This result is robust to changes in models parameters, obstacles shapes, model formulation, and added gaussian noise 

This project was a cross-disciplinary collaboration with Helen McCreery, Justin Werfel and Stefan Popp.