Ethan (E) H. Harrison

Hi! I'm an undergraduate at UC Berkeley studying Electrical Engineering and Computer Science. I am a undergraduate researcher at Berkeley's Robotic AI & Learning Lab, advised by Professor Sergey Levine and Andrew Wagenmaker, where I work on training robot policies with reinforcement learning.

Email  /  Resume  /  Google Scholar  /  LinkedIn  /  Github

Research

I'm interested in robot learning, reinforcement learning, and deploying learned policies on real hardware. My current work focuses on asynchronous RL for vision-language-action models under inference latency.

Investigated finetuning action-chunking VLA policies under inference time constraints. Proposed Asynchronous RL with Intermediate Information (ARLI), achieving state-of-the-art performance on simulated and real-world manipulation tasks.

Projects

Engineered a physical harness with IMUs and an ESP32 for teleoperating a UR7e robot via arm movements, supporting pick-and-place and dexterous tasks. Integrated MediaPipe and ArUco detection for hand tracking and built a ROS/RViz GUI for defining and executing robot punches.

Trained a PPO-based drone racing policy with custom reward and termination design in Isaac Sim. Used parallel simulation and curriculum-style resets to improve convergence; achieved the fastest policy improvement rate and fastest average lap time among RL-based teams.

Built an agentic AI system that designs and executes ML experiments in WandB from a user's codebase and request. Developed a tool-calling backend using OpenAI's tool API to interact with WandB, read/write code, and summarize results.

Miscellanea

Website template from Jon Barron.