About
I’m a second year PhD student in the Robotics Department at The University of Michigan Ann Arbor. I work in the Laboratory for Perceptive Robotics and Grounded Reasoning Systems (PROGRESS), advised by Dr. Chad Jenkins, and funded by the National Science Foundation’s Graduate Research Fellowship Program (NSF GRFP). I previously studied Mechanical Engineering at the University of Alabama, where I obtained a Bachelor and Master of Science degrees, studying non-linear control and rehabilitative robotics under Dr. Christian Cousin and the Randall Research Scholars Program. I also interned multiple times at Southwest Research Institute (SwRI), working in the Robotics and AI division.
My recent research focuses on the intersection between robotic control and planning, including ways to model uncertainty within these problems. I’m especially interested in ways to run more effective and optimal planning/control strategies on non-linear, high degree-of-freedom, and complex systems, paving the way for them to become more assistive to humanity.
Specifically, my work focuses on combining Stein Variational Inference with state-of-the-art planning and control methods, to increase their performance, efficiency, and uncertainty-handling capabilities.
Spatial Object-Centric Representation Network (SORNet, Yuan et al., 2021) is a network architecture that takes an RBG image with several canonical object views and outputs object-centric embeddings. The authors of the original paper trained and tested SORnet on their custom Leonardo and Kitchen data sets, as well as the CLEVR dataset (Compositional Language and Elementary Visual Reasoning). We expanded SORnet’s capability by training it on PROPS Dataset (Progress Robot Object Perception Samples), which was extensively used throughout this course. Training SORNet with PROPS dataset allow us to test its capabilities to a real-world dataset in order to better understand how it performs in real-life applications.
Strokes, Parkinson’s disease, and many other debilitating neurological disorders affect millions across the world, resulting in many cases of disability and reduced overall quality of life. There are several rehabilitation methods for these diseases and conditions, but functional electrical stimulation (FES) has become a popular choice recently due to its ability to induce muscle contractions. Further, FES is often combined with robotic components to improve its control and reduce fatigue, forming hybrid exoskeletons. These exoskeletons, while effective, can often hinder user engagement and immersion, which prevents them from reaching their full potential. Thus, to combat the lack of engagement in rehabilitation exercises, this work describes a design for a large scale two-degree-of-freedom motion platform to be integrated with virtual reality. The platform can support an exoskeleton and its user, safely and stably actuating to a range of desired pitch and roll angles.
Work from Southwest Research Institute (SwRI): This demonstration highlights the abilities of robots to actively perceive and act on dynamic environments, by providing an example case of building a Jenga tower on a rotating turn table. It actively perceives/estimates the state of the blocks, tower, and turntable to build a full Jenga tower (18 layers or 54 blocks). This work is simply a demonstration of work that can be applied to several tasks, particularly those involving On-orbit Servicing, Assembly, and Manufacturing (OSAM)
Functional electrical stimulation (FES) is a viable rehabilitation method for individuals affected by neurological injuries. When FES is combined with a powered exoskeleton, a hybrid exoskeleton is created. To control hybrid exoskeletons, a user’s muscles must be stimulated while simultaneously activating the exoskeleton’s motors. However, this stimulation can become uncomfortable at higher levels, and looses, effectiveness. Thus, in this work, we stimulate a user’s biceps and triceps muscle groups on an upper-limb hybrid exoskeleton, but saturate the stimulation input and redirect remaining control input into the exoskeleton’s motor. The motor is actuated using a robust feedback controller along with the excess input from the user’s saturated muscle controller. Additionally, a Lyapunov stability analysis was conducted to prove that the closed-loop position error system is uniformly ultimately bounded using the two controllers. This approach showcases a novel way to safely limit stimulation without sacrificing overall performance via a conditionally offloaded neural network on hybrid exoskeletons. Experiments were conducted on four participants without injuries for validation and to demonstrate the efficacy of the proposed approach.
This robot climbs up and down stairs while safely carrying a load(i.e., a portable oxygen tank) and is be portable, so that the robot is not confined to a singular environment. Through two linear actuators, four active wheels, and four passive wheels, the robot can successfully ascend and descend stairs in a safe and stable fashion. As the robot drives forward, it uses a combination of sensors (i.e., infrared, sonar, limit switches) to determine if it has arrived at a set of stairs. A linear actuator is used to pull the front set of passive wheels up to the height of the stairs, at which point the robot drives forward until it has established stable points of contact with the two front driven wheels and both sets of passive wheels. Another linear actuator is used to pull the back set of driven wheels up to the height of the stairs, at which point the robot drives forward again until it hits the next stair. The process will then repeat, with the front set of passive wheels and back set of small wheels being raised using linear actuators, until the robot reaches the top of the stairs.
![]()
Motorized functional electrical stimulation (FES) cycling can serve as a rehabilitation strategy for individuals affected by neurological injuries. It is unique among human-robot interaction tasks because the cycle’s motor must be simultaneously controlled alongside the rider’s leg muscles (using neuromuscular electrical stimulation). In this work, two tracking objectives are proposed for the FES cycle: 1) stimulate the rider’s leg muscles to track a desired cadence, and 2) use the cycle’s motor to indirectly track a desired torque with an adaptive admittance controller. A combined Lyapunov and passivity based switched systems stability analysis was conducted to prove the cycle’s motor is able to globally exponentially track the admittance trajectory and stabilize the overall system. Additionally, this work showcases a method for the rider to smoothly enable and disable torque tracking.
