CushSense is a soft, stretchable fabric-based tactile sensing skin designed for physical human-robot interaction tasks like robotic caregiving. CushSense uses stretchable fabric and hyper-elastic polymer to detect contacts by monitoring capacitive changes from skin deformation, offering high accuracy, durability, low cost, and easy fabrication. 

The SMArt bracelet is a smart wearable interface integrated with Shape-Memory Alloy (SMA) springs aimed at promoting well-being by allowing self-regulation of stress and anxiety through mindful breathing practices. It comprises a customizable woven band with integrated hardware that reacts to elevated heart rates by providing visual and haptic feedback to remind the user to regulate their breathing. While integrating the hardware into the soft fabric is challenging, the potential benefits of this on-skin interface for health monitoring and improved well-being present promising opportunities.

We taught a WidowX robot to flip pancakes using a setup with 3 depth cameras, AR tagged pancakes, a 3D printed pan, and a surrounding cage. By recording joint states, velocities, and pancake poses during demo flips, we trained a model that took the robot's joint/velocity states and pancake pose as input to output joint velocity commands, enabling the robot to learn pancake flipping.

Utilizing the SolidWorks software, our team engineered a robust and heavy-duty firewood processor, reminiscent of a trailer's design. This innovative machine is capable of accommodating massive logs through its side-loading mechanism. Once the logs are positioned, the processor employs precise cutting mechanisms to segment them into manageable pieces. Subsequently, a combination of conveyor systems and hydraulic pushers facilitates the splitting process, transforming the log segments into neatly divided firewood pieces, ready for distribution and sale in retail stores.

I utilized the open-source Multi-agent System for non-Holonomic Racing (MuSHR) robotic hardware and software platform to explore AI in autonomous vehicles and mobile robotics. I programmed ROS Python scripts to launch the MuSHR car's controller, planner, and localization modules, enabling surveillance of three rooms in 27 seconds using set waypoints. I implemented particle filters and extended Kalman filters for precise localization based on car kinematics in simulation and real-world experiments. Additionally, I applied A* and RRT path planning algorithms in Python for optimal roadmap searching with obstacle avoidance. I designed PID and MPC controllers, tuning parameters for accurate path tracking performance based on the planners' desired paths. Finally, I built maps of the real-world environment utilizing LIDAR data and photo processing techniques.

SPINDLE is a tabletop device designed to assist stroke patients in performing wrist motor function physical therapy. This innovative apparatus can simulate activities of daily living, such as turning knobs or opening jars, by providing adjustable resistance through three flat motors integrated into a Spherical Parallel structure. By mimicking these common tasks, SPINDLE aims to facilitate the rehabilitation process and improve the functional capabilities of individuals recovering from stroke, enabling them to regain independence in performing routine activities.