Masters student for West Virginia University (WVU) Interactive Robotics Laboratory (IRL):

• Worked on BrambleBee, an autonomous pollination robot (https://youtu.be/66isrgth7-Q)

• - ROS programming in C++

• - Manipulation maneuvers using MoveIt! Library for Kinova® Jaco Arm

• -- Using Intel® RealSense D435's point cloud, loaded into an Octomap server for obstacle avoidance

• -- Pose estimation for a custom end-effector (eye-in-hand configuration) using ArUco markers & a Kalman filter to update its pose during motion

• --- Transferred to work with artificial & real flowers

• -- Solved inverse kinematics using TRACK-IK library

• -- Implemented solution for a form of the Traveling Salesman Problem (TSP) by finding trajectory minimizing required motion to visit all reachable flowers

• -- Robotic Pollination of QR Flowers: https://youtu.be/0t7t8AScZSE

• -- BrambleBee Robot Pollinating a High-Fidelity Artificial Plant: https://youtu.be/ZbgtP9CHycA

• - Updated & created URDF models & plugins for Gazebo simulation

• - Publications:

• -- https://arxiv.org/abs/1906.09294

• -- https://arxiv.org/abs/1808.10010

• ROS camera streaming between onboard robotic computer & remote desktop

• Implemented autonomous obstacle avoidance, navigation, & semi-autonomous box-pushing (https://youtu.be/pOuQqRNq_zs) on Turtlebot3 Waffle Pi

• Use of Vicon motion capture for robotics ground truth pose

• Participating in University Rover Challenge

• - Using IRL's mobile robot Pathfinder

• -- Low-level control of actuators for driving

• -- Executed autonomous GPS waypoint navigation, obstacle avoidance using Leddar sensor, & tennis ball detecting using OpenCV (https://youtu.be/POSIFwnitT0)

• - Mentoring current team & taught programming fundamentals, C++, ROS, & computer vision introduction

• Exploring robotic Reinforcement Learning techniques including Value Iteration, Policy Iteration, & Q-Learning

Mentored by Rohan Thakker and Co-mentored by Ali-akbar Agha-mohammadi.

• Explored machine learning techniques (specifically Reinforcement Learning (RL) & Meta-RL) for rover to learn terrain dynamics & adapt to unknown terrain

• Applied & tested meta-RL method Gradient-Based Adaptive Learning (GrBAL) described in conference paper, "Learning to Adapt in Dynamic, Real-World Environments Through Meta-Reinforcement Learning"

• Experimented & evaluated performance (e.g., training time, average return, forward progress)

• Applied & tested meta-learning technique Adaptive Learning for Probabilistic Connectionist Architectures (ALPaCA) from "Meta-Learning Priors for Efficient Online Bayesian Regression"

• Enhanced simulated RC car in ROS & Gazebo simulator

• Integrated ALPaCA with novel Bayesian adaptive controller

• Assessed offline & online performance including training time & prediction accuracy

• Co-authored "Bayesian Learning-Based Adaptive Control for Safety Critical Systems", primarily in ALPaCA integration and results: https://ieeexplore.ieee.org/abstract/document/9196709

• Implemented algorithm described in conference paper, "A High-Accuracy Visual Marker Based on a Microlens Array," for pose enhancement using a visual marker

• - Analyzed virtual performance assessing pose estimation error

• Compiled literature review on visual SLAM for WVRTC team

• - Researched & evaluated different open source visual SLAM algorithms

• Analyzed Raspberry Pi 3's performance with ROS & image processing

• Implemented algorithm to detect custom outer border containing embedded inner visual marker