An attitude and heading reference system (AHRS) is a nonlinear state estimator unit for computing orientation in 3D space.This paper designs an AHRS using three approaches:an invariant observer,an invariant extended Kalman filter (IEKF),and a conventional extended Kalman filter (EKF).The three designs are validated in experiment versus a ground truth,demonstrating the practical interest of the invariant observer methodology and the advantage of the IEKF over the EKF under model uncertainty.