Qiaozhi Wang, P.E.
(Prounouced "Chi-ao-Jhi")
Results-driven Professional Engineer specializing in mechatronics, systems integration, and autonomous systems. Demonstrated expertise leading end-to-end development of complex electro-mechanical systems — from concept design through prototype build, validation, and public debut.
OffWorld Gym:
Real-world Robotics Environment for Reinforcement Learning Research
(2019 - 2024)
Skills: Mechanical Design, System integration, Lab Testing
Solidworks, Fast-Prototyping, Fabrication, Machining, 3D Printing
Role: Mechatronics integrator, operation and test support
OffWorld Gym (OWGym) is a robot playground, a collection of real-world environments for reinforcement learning in robotics with free public remote access. Close integration into the existing ecosystem allows users to start using OffWorld Gym without any prior experience in robotics and removes the burden of managing a physical robotics system, abstracting it under a familiar API.
This project not only highlights the collaborative efforts of the ML and mechatronics teams but also underscores my personal contributions to its success.
As the lead in charge of the modification and enhancement of the Commercial Off-The-Shelf (COTS) small robot ROSbot 2 PRO(https://store.husarion.com/products/rosbot-pro), my focus was to tailor it specifically for the OWGym application. The original robot chassis underwent significant upgrades, evolving into a more robust structure that improved balance and incorporated features to prevent rocks from hindering wheel movement. Additionally, I increased the robot's height and introduced higher-torque motors to enhance mobility, particularly in navigating pebble piles simulating a moon-surface environment.
The onboard sensor suite was also reconfigured to accommodate a stereo camera, lidar, and HTC Vive tracker. This adjustment provided the ML team with more meaningful data for learning. A slip ring was also integrated for power cord management, which allows the robot to connect to an overhead power source, liberating it from the constraints of limited battery power without compromising mobility or obstructing the sensors' field of view (FOV). I also revamped the robot's color scheme to orange and black, aligning it with the signature look of all OffWorld robots.
ROSbot upgrade
My involvement extended beyond hardware modifications. I created a detailed CAD model of the entire robot and generated drawings for fabrication, utilizing 3D printing for small sensor fixtures and quick design validations. Additionally, I contributed to the construction of the "Gym," a wooden structure filled with pebbles and equipped with monitoring cameras, also providing essential data and power interfaces.
During the beta testing phase, I took charge of organizing the mechatronics team, ensuring timely support and conducting rapid design and fabrication iterations to align with the software team's specifications.
For a more in-depth understanding of the ML and reinforcement learning (RL) aspects, I invite you to explore the program's GitHub page: OffWorld Gym GitHub.
OffWorld Gym: reality vs simulation
