Balancing Triangle
A reaction-wheel controlled balancing robot using embedded control systems.
Overview
A self-balancing triangular robot that uses an IMU and a single reaction wheel to stabilize around one pivot point. Unlike a two-wheeled balancing robot that balances by driving wheels along the ground, this project balances entirely through reaction-wheel momentum exchange — closer in spirit to spacecraft attitude control than to a conventional wheeled robot.
Problem
Balancing on a single pivot point using only a reaction wheel is an inherently unstable control problem: the system is open-loop unstable, and the only actuator available is a wheel that exchanges angular momentum with the body rather than pushing against the ground. That means the controller has to work entirely through momentum transfer, and available control torque is limited by motor and wheel inertia.
The project needed:
- Accurate real-time tilt-angle estimation from noisy IMU data
- A control loop fast and stable enough to catch the system before it falls
- Embedded firmware that can hit a hard real-time control loop rate on resource-constrained hardware
Technical Architecture
- Sensing — An IMU (accelerometer + gyroscope) provides raw tilt-rate and gravity-vector measurements, fused into a stable tilt-angle estimate with a complementary filter to combine the gyro's short-term accuracy with the accelerometer's long-term stability.
- Control (STM32, C) — A PID controller running on a fixed-rate timer interrupt computes the corrective torque command from the estimated tilt angle and rate.
- Actuation — The torque command drives the reaction wheel motor; the resulting change in wheel angular momentum applies an equal and opposite reaction torque to the body, per conservation of angular momentum.
Implementation Details
The control loop runs on a hardware timer interrupt on the STM32 to guarantee a consistent sample rate, since PID gains tuned at one control rate don't transfer cleanly to another. IMU readings are fused with a complementary filter rather than a full Kalman filter — simpler to implement and tune on embedded hardware, and sufficient given the IMU noise characteristics observed on this platform.
Reaction wheel speed is monitored to avoid actuator saturation: if the wheel approaches its maximum RPM, it has less headroom left to exchange momentum with the body, so the mechanical design biased toward a wheel/motor combination with enough headroom for the expected disturbance range.
Challenges
Getting the complementary filter's tuning right — balancing responsiveness to fast tilt changes against susceptibility to accelerometer noise — took several iterations. PID tuning was similarly iterative: gains that looked stable in a bench test would sometimes oscillate once the full chassis mass and center of gravity were accounted for, requiring re-tuning with the real hardware rather than an idealized model.
Results
The final system reliably self-balances from a near-vertical starting position and rejects moderate manual disturbances (a light push) without falling, demonstrating stable closed-loop control purely through reaction-wheel momentum exchange.
Lessons Learned
Tuning a real embedded control loop is a different exercise from tuning a simulated one — sensor noise, actuator saturation, and timing jitter all show up in ways that are easy to abstract away in simulation. Building the mechanical platform with generous reaction-wheel headroom made the control problem meaningfully easier than a tightly-constrained design would have.
Future Work
- Add a second reaction wheel for full planar (not just single-axis) balance
- Move from a complementary filter to a Kalman filter for improved estimation under faster disturbances
- Add telemetry logging over wireless link for post-test analysis