Attitude Estimation

Fuses gyroscope, accelerometer, and (optionally) magnetometer data to compute roll, pitch, and yaw angles in real time.

Attitude estimation computes an object’s orientation in space from inertial sensor data, typically expressed as roll, pitch, and yaw angles or as a quaternion. AHRS (attitude and heading reference system) was originally an aviation instrument, built from a three-axis gyroscope, accelerometer, and magnetometer plus an onboard computer that outputs attitude and heading directly; a plain IMU usually only outputs raw angular velocity and acceleration. In principle, integrating the gyroscope gives an angle that’s accurate short-term but drifts; the accelerometer can sense the direction of gravity, which corrects roll and pitch; yaw has no gravity reference, so it needs correction from a magnetometer, vision, or another source. Common fusion algorithms include the Kalman filter, the complementary filter, the Madgwick algorithm (proposed during a 2009 PhD project, now open-sourced under the name Fusion), and the Mahony filter. Control policies for legged and humanoid robots often take the estimated attitude (or projected gravity) and angular velocity as input, so attitude error directly affects balance control.

ExampleWhen a quadruped robot is shoved from the side, IMU attitude estimation gives the real-time change in body roll angle, and the locomotion control policy steps to recover balance based on it.

Also called

Attitude and Heading Reference System, AHRS, IMU Attitude Fusion

Related

Inertial Measurement Unit · Complementary Filter · Kalman Filter · Projected Gravity · Roll-Pitch-Yaw (RPY) · Quaternion

Sources

Wikipedia: Attitude and heading reference system
x-io Technologies: Open source IMU and AHRS algorithms (Madgwick / Fusion)