Research

University of Pennsylvania (June 2015 to Present)

Semantic Localization, Mapping, and Navigation with the PHD Filter
With Dr. Nikolay Atanasov and Prof. Philip Dames
We are developing an improved way of the semantic SLAM approach by combining the probability hypothesis density filter for pose estimation and deformable part models for object detection. This can be applied to any robot, especially autonomous cars, and will be a significant upgrade from the current mapping approaches. Looking to submit paper to upcoming ICRA.

quad picture 

ROS Implementation for Autonomous Control of Pico-Quadrotor
Advised by Professor Vijay Kumar at GRASP Lab, University of Pennsylvania
This project was to implement the Robot Operating System (ROS) on the world's smallest autonomous drone (designed by PhD candidate Yash Mulgaonkar), which gives it an extensive robot ecosystem with toolboxes for communication, control, planning, visualization, mapping, and more, and autonomous fly and control with the Vicon motion capture system. I implemented a new ROS node in C++ using the existing control stack to receive data from Vicon to the computer and send commands back to the robot via wireless USB for autonomous take-off, hover, flight, and landing.

Project article featued on M&T Blog and Penn News Today in Fall 2015.

Stanford University (June 2014 to November 2014)

free-flyer pic 

A Fast Marching Trees Motion Planning Approach to the Autonomous Space Taxi: Algorithmic Experiments for Proximity Maneuver and Docking
Advised by Professor Marco Pavone at Stanford University
This project was to validate the FMT motion planning algorithm on a physical testbed with autonomous free-flying space robots and moving obstacles specifically for maneuvering and docking. I developed a C++ framework integrating a PID control law (tested in Simulink), the FMT motion planning algorithm, obstacle avoidance approaches, and Optitrack position data transmission. In addition, I refurbished the three free-flyers with new pneumatics, microcontrollers & motor controllers, momentum wheels, wiring, and docking components. Finally, I wrote the protocol for hardware, software, and testing methodology.

Named a 2015 Intel Science Talent Search semifinalist for research paper. Presented videos of successful navigation, obstacle avoidance, and docking to NASA.

University of California, Berkeley (July 2013 to August 2013)

labview pic 

Mapping Intensity Distributions with LabView Automation
Advised by Professor Connie Chang-Hasnain at UC Berkeley
This project was to map intensity distributions of the optical field, which will be useful for characterizing optical devices in terms of focal length and mode pattern, using automation of actuators controlled with LabView. I developed a LabView program that initializes the motor, moves the motor like a stepper, reads data from the power meter, transfers data into an array, and converts data into distance vs intensity graphs with time delay consideration. In addition, I built the physical setup with a laser, objective lens, photodetector, stage, actuators, power meter, and motor controllers.

Final presentation given to nano-optoelectronics lab groups. Recorded real-time video of two-dimensional scanning.