STORY: How Olin First-Years Built a Viral Self-Balancing Bicycle (VIDEO)
This bike should be falling over. But it’s not...
The autonomous self-balancing bicycle built by Justin (Jonghoo) Yoon '29, Pia Swarup '29, and Aryan Banerjee '29, is pictured balancing on the two stone spheres in front of Milas Hall.
That’s because a team of first-year Olin students built it as an autonomous self-balancing bicycle for their Independent Study Research (ISR) project.
Working with field roboticist and Olin alumna, Dr. Victoria Preston ‘16, assistant professor of engineering, as an advisor, this open source research project explores the design and implementation of an autonomous, self-balancing bicycle using a reaction wheel for stabilization.
The system combines embedded control, power electronics, and robotic autonomy to achieve balance, perception, and navigation on a dynamically unstable platform.
"This is a project I've had in the back of my mind since high school," says Justin (Jonghoo) Yoon '29, one of the bike's creators.
"Building a self-balancing vehicle was always a fascinating topic for me. With the support and resources that Olin offers for student research projects, I was finally able to make it a reality."
Along with fellow first years, Pia Swarup and Aryan Banerjee, the trio of students with Preston were able to build something amazing, plus, the entire project is built to be a low-cost, modular platform that other researchers can use for two-wheeled autonomous vehicle research.
Research of a platform that is proving to be of huge interest as a video created by Olin of the bicycle has since gone viral on Instagram, generating 1.7 million views and thousands of shares and engagements in a week.
Justin (Jonghoo) Yoon '29, one of the bike's creators, is pictured in the Library during Spring Expo 2026.
The Bicycle
The bicycle uses an active reaction wheel controlled by a STM32. The STM32 is a popular family of 32- bit ARM Cortex-M microcontrollers and microprocessor integrated circuits used to generate the stabilizing torque needed for the bike to stay upright.
Based on cascade Proportional-Integral-Derivative (PID) control, the inner PID loop continuously reads roll angle from an Inertial Measurement Unit (IMU) and adjusts the angle of the bicycle, while the outer PID loop dynamically corrects the balance point to keep the reaction wheel from being saturated.
Saturation (i.e. reaching maximum designed rotational speed, thereby losing the ability to maintain stability control) of the motor was the main issue the students were facing with the wheel before they initiated the balance point adjustments.
The bicycle moves using a chain-driven DC motor mounted to the frame, and steers using a high-torque servo motor with a custom bevel gear.
Autonomy is achieved using visual SLAM (Simultaneous Localization and Mapping) with the front-facing ZED 2i stereo depth camera running on a Jetson Orin Nano. An additional 2D LiDAR sensor is used for low-latency obstacle detection.
The most eye-catching parts of the bicycle, it can balance reliably even on narrow railings (see video), and is capable of basic autonomous navigation. Even the phrase, "Ghost Bike" was heard from attendees at the Olin Spring Expo while the autonomous navigation was being demonstrated.
See the team's research paper: Design and Development of a Modular Open-Source Platform for Two-Wheeled Self-Balancing Vehicles
How'd they build it?
See the bill of materials, CAD files and more
(L to R): Pia Swarup '29, Justin (Jonghoo) Yoon '29, and Aryan Banerjee '29, pose in front of the self-balancing bicycle they built as part of an Independent Study Research (ISR) project advised by Dr. Victoria Preston '16, assistant professor of engineering. Still from video by Adam P. Coulter
What's next?
In terms of what's next, Pia says the team is looking to expand on its autonomous driving capabilities. "The goal is for it to autonomously figure out how to get to a certain point using SLAM."
The team is also looking at Reinforcement Learning (RL) to allow it to balance using just steering.
"We are currently looking at RL for a hybrid stabilization model that combines steering-based balancing and the active reaction wheel for ideal stabilizing characteristics in both low and high speed scenarios," says Justin.