Reinforcement Learning for Autonomous Dexterous Manipulation Used state-of-the-art reinforcement learning techniques to improve sim-to-real transfer of autonomous dexterous manipulation skills. Embed to Control Implementation of Embed to Control, a locally linear latent dynamics model for control from raw images. Ergodic Control in Exploration An active learning agent that uses ergodic control and iLQR to find a hidden box. ICORR 2025 Paper I created a model that estimates upper limb function in stroke patients, using features extracted from IMU data. I published a paper to a top rehab robotics conference, which was nominated for best paper. Deep RL from Scratch Custom Deep RL, including a simple Gymnasium implementation, neural net from scratch, and the Deep A2C + vanilla tabular Q-Learning algorithms. Pick-and-Place Mobile Manipulation Pick-and-place mobile manipulation tasks with a KUKA youBot, starting from an initial condition with an arbitrary error relative to a desired trajectory. Filtering and State Estimation Implementation of Bayesian filtering algorithms, including EKF, regular + Augmented UKF, and Particle Filter. Autonomous Quadrotor Control Manual and autonomous control of a quadrotor, using PID control and computer vision. Simulated Soccer Juggling A physically reasonable representation of soccer juggling with Lagrangian dynamics, with a custom feedback control system. PID Motor Controller and Client A robust PID motor controller in C on the PIC32MX270F256B microcontroller, with a Python client.
