Madgwick

Madgwick refers to a real-time orientation estimation algorithm commonly known as the Madgwick filter or Madgwick’s AHRS filter. It is used to fuse data from inertial measurement units, combining gyroscope, accelerometer, and magnetometer measurements to estimate three-dimensional orientation in quaternion form. The algorithm was developed by Sebastian O. Madgwick in the early 2010s and has since become widely used in both research and hobbyist projects due to its computational efficiency and suitability for embedded systems.

The Madgwick filter operates by integrating gyroscope data to update orientation and then applying a gradient-descent based correction term to align the estimated gravity and magnetic field directions with the sensor measurements. This approach yields a lightweight alternative to more resource-intensive filters such as extended Kalman filters, while maintaining reasonable accuracy for many practical applications. The method supports both nine-axis fusion (gyroscope, accelerometer, and magnetometer) and reduced configurations (e.g., six-axis using gyro and accelerometer).

Open-source implementations are readily available in multiple programming languages and platforms, including C, C++, MATLAB, Python, and hardware targets like Arduino and ESP32. As a result, the Madgwick filter is commonly used in robotics, drone flight controllers, virtual reality and augmented reality tracking, motion capture, and other applications requiring real-time attitude estimation from inexpensive MEMS sensors.

Limitations include sensitivity to magnetometer disturbances and the need for proper calibration of both accelerometer and magnetometer. Dynamic accelerations can also affect accuracy, since the algorithm relies on gravity and magnetic field references for correction.

See also: Attitude estimation, quaternion, IMU, Kalman filter.