Self-Balancing Bicycle Stability vs Speed

ISEF Category: Robotics and Intelligent Machines

Subcategory: Control Theory  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A bike can balance itself, but only in a narrow speed range. Slow down too much, and the system can tip. Speed up, and the steering dynamics change again. That makes this project a real control problem, not a toy demo.

What Is It?

A self-balancing bicycle uses sensors, a motor, and a controller to keep the bike upright without a rider. Think of it like a tightrope walker with a hidden helper that nudges the rope position. The controller watches the bike’s tilt and steering angle, then commands the front fork to correct the motion before the bike falls.

The flywheel adds another layer. A spinning mass stores angular momentum, which can help resist sudden changes in orientation. The stepper motor on the steerable front fork gives the system a way to steer itself back under the center of mass. Your job is to study when that help works and when it fails, especially as forward speed changes.

The Whipple bicycle model is a physics model that predicts bicycle stability from geometry, mass, and speed. In plain language, it is the math version of the bike’s natural balance behavior. Comparing your real machine to that model lets you test whether your controller matches theory, where it breaks down, and how much the speed-dependent gain schedule improves performance.

Why This Is a Good Topic

This is a strong science fair topic because you can measure real stability, not just build a cool machine. You can change speed, controller settings, fork response, and flywheel effects, then track how those choices shift the balance window. It connects to robotics, vehicle dynamics, and control systems, which gives it real engineering depth. You can also learn model validation, sensor data analysis, and how to compare theory with experiment.

Research Questions

• How does forward velocity affect the maximum tilt angle the bicycle can recover from?
• What is the effect of gain scheduling on the width of the stable speed range?
• Does adding flywheel angular momentum increase the time the bicycle can remain upright after a disturbance?
• To what extent does the experimental stability boundary match the Whipple bicycle model predictions?
• Which steering control parameter changes balance recovery more at low speed than at high speed?
• How does disturbance size affect the controller's ability to restore upright motion across different speeds?

Basic Materials

• Riderless bicycle frame with accessible steering geometry.
• Stepper motor with driver board for front fork actuation.
• Flywheel assembly with safe mounting hardware.
• Microcontroller such as Arduino or ESP32.
• IMU sensor for tilt and angular rate measurement.
• Wheel speed sensor or optical encoder.
• Battery pack and power regulation components.
• Rigid test stand or safety harness for controlled trials.
• Laptop for logging and analysis.
• Tape measure, level, and basic hand tools.

Advanced Materials

• Instrumented bicycle frame with known mass and geometry measurements.
• Brushless or stepper actuator for steerable fork with position feedback.
• High-rate IMU with synchronized data logging.
• Optical encoders for wheel speed and steering angle.
• Data acquisition system with timestamp synchronization.
• Wind-down or release fixture for repeatable disturbance tests.
• Calibration weights for center of mass estimation.
• High-speed camera for motion validation.
• Finite element or multibody modeling tools if available.
• Safety enclosure or spotting rig for repeated trials.

Software & Tools

• Python: Processes sensor logs, computes stability metrics, and plots speed versus recovery performance.
• MATLAB: Fits control models, compares measured data with the Whipple prediction, and helps tune gains.
• ImageJ: Tracks frame-by-frame tilt or steering angle from video when sensor data needs a check.
• Excel or Google Sheets: Organizes trials, labels conditions, and makes quick comparison charts.
• OpenModelica: Helps build or inspect a bicycle dynamics model before you compare it with experiment.

Experiment Steps

1. Define the stability metric you will use, such as recovery time, maximum recoverable tilt, or sustained upright duration.
2. Map the system variables you will hold fixed and the one variable you will change first, usually forward speed.
3. Build a measurement plan that records tilt, steering angle, wheel speed, and controller output in the same time base.
4. Design a set of disturbance tests that are repeatable enough to compare across speeds and controller gains.
5. Create a model comparison plan that links your data to the Whipple bicycle model with the same geometry and mass assumptions.
6. Plan a way to separate controller performance from the bicycle's passive dynamics so your conclusions stay clear.

Common Pitfalls

• Changing speed and controller gains at the same time, which makes it impossible to tell what caused the stability change.
• Logging tilt, steering angle, and wheel speed on separate clocks, which ruins model comparison.
• Using a flywheel that is not balanced well, which adds vibration and hides the real control effect.
• Testing only one disturbance size, which can make the stability envelope look larger or smaller than it really is.
• Ignoring steering backlash or motor lag, which causes the controller to look weaker than the physics model predicts.

What Makes This Competitive

A class-level version of this project only shows that the bike can balance. A stronger version maps the full stability envelope, then compares measured boundaries against a clean Whipple-model prediction. You can raise the quality further by separating passive bicycle dynamics from active control effects and by testing whether the gain schedule improves low-speed and high-speed performance in different ways. Strong statistics, repeatable disturbances, and careful calibration make the work much more convincing.

Project Variations

• Test how different flywheel mass distributions change the lowest stable speed.
• Compare gain-scheduled control with a fixed-gain controller under the same disturbance set.
• Use video tracking instead of onboard sensors to estimate tilt and steering angle, then compare both measurement methods.

Learn More

• MIT OpenCourseWare, Nonlinear Control or dynamics courses: Search MIT OpenCourseWare for bicycle dynamics, feedback control, and multibody systems lecture notes.
• PubMed: Search for review articles on balance control, human and robotic bicycle stability, and sensorimotor feedback models.
• NASA Technical Reports Server: Search for bicycle dynamics, control systems, and balance modeling reports that include mathematical derivations.
• NOAA National Centers for Environmental Information: Use wind and weather context if you test outdoors and need to explain environmental effects on trials.
• The Whipple bicycle model papers in peer-reviewed journals: Search Google Scholar or journal databases for original and review papers on bicycle stability modeling.