Understanding MEMS Accelerometers & Gyroscopes: A Builder\’s Guide

From drone flight controllers to self-balancing two-wheeled robots, 6-Axis Inertial Measurement Units (like MPU6050, BMI160, and LSM6DSO) form the sensory core of modern robotics. But raw sensor readings are noisy and prone to drift. Here is how sensor fusion bridges the gap.

1. Accelerometers vs Gyroscopes: Complementary Strengths

An accelerometer measures linear acceleration (including gravitational acceleration 1G), allowing static tilt angle estimation. However, any physical vibration or sudden movement corrupts the angle estimate.

Conversely, a gyroscope measures angular rate ($^circ/text{sec}$). Integrating angular velocity over time gives a clean, rapid angle response unaffected by linear bumps. However, continuous integration accumulates numerical drift within seconds.

2. Implementing a Complementary Filter

A complementary filter marries both signals: a high-pass filter cleans gyroscope integration, while a low-pass filter gently corrects long-term drift with accelerometer gravity vectors:

// Complementary Filter Implementation
float angle_pitch = 0.0;
float dt = 0.01; // 100 Hz sampling loop (10ms)
float alpha = 0.96; // Filter weighting constant

void update_orientation(float gyro_pitch_rate, float accel_pitch_angle) {
  // Integrate gyro rate
  float gyro_delta = angle_pitch + (gyro_pitch_rate * dt);
  
  // Complementary blend
  angle_pitch = alpha * gyro_delta + (1.0 - alpha) * accel_pitch_angle;
}