• Enhancement of perception algorithms to estimate object states through the integration of vision and tactile sensor data.

• Development in a simulated environment and ensuring their successful transfer and performance on real-world advanced robotic platforms with multi-fingered hands (Sim2Real).

• Exploration of temporal approaches to improve the stability of the estimates over time.

Master’s Capstone: Applications of the Soft Bubble Grippers: Visuotactile Sensing for Visuomotor Policy Learning

• Re-designed 3D fabricated parts for compatibility with UR5e robot and Robotiq grippers.

• Wrote ROS package to interface with Intel Realsense cameras and implement pose and shear deformation estimation functionalities.

• Trained and evaluated diffusion and transformer-based policy learning models on teleoperation demonstrations of various manipulation tasks including Push T and handover/placement of objects.

Other research:

• Developed a Sim2Real framework to classify objects and derive pose estimations from Gelsight tactile data.

• Evaluated the affect of object-centric representations on a diffusion-based policy learning model.

• Developed a framework for LiDAR-based map generation and re-localization in GPS-denied environments.

• Interfaced with the NVIDIA Isaac ROS hardware acceleration package and an open-source LiDAR-inertial odometry package.

• Built support for offloading maps and positional data from hard-core processors to cloud tools for path planning and area clearing.

• Project 1: Configured MQTT connection of soft robotic compression garment WIFI microcontroller to a central Hub server through Docker. Collaborated with an interdisciplinary team of faculty and students in computer science. Integrated a conversational voice assistant to provide real-time, in-home, unobtrusive sensing and on-body stimulation solutions (e.g., pressure, heat, etc.). Participated in a formal interview of my contribution and in a demonstration of the controllability and functionality of the technology.

• Project 2: Constructed a remotely controllable soft robotic compression garment to assist novice meditators. Assembled and soldered hardware for Bluetooth and MOSFET-driven actuator circuit. Programmed Arduino using C/C++ code to control the system. Developed a user interface in Processing, using Java code, to visualize which body areas stimuli was being presented to. Tested garment system through a guided meditation user study (n=10).

• Project 3: Assisted with data collection and system testing for a hybrid shape memory alloy (SMA)- pneumatic-based wearable upper limb exoskeleton. Recorded power consumption, flection angle, and time of flection for a mannequin arm with various SMA layouts.

• Project 4: Designing neural-network-based controller for SMA-driven model arm flexion/extension using Matlab and Simulink. Comparing performance of various data-based model-free controllers on the system.