Inertial Sensor-Based Attitude Estimation Combining a Kinematic Constraint for Elimination of Acceleration-Induced Inaccuracy

Regarding attitude estimation Kalman filter (KF) based on inertial sensor signals, one of the major issues that affect the estimation accuracy is the acceleration-induced inaccuracy associated with accelerometer signals. To deal with this issue, researchers have proposed various types of acceleration-compensating mechanisms. Because those mechanisms, however, are basically stochastic approaches, such as Markov chain models, they cannot thoroughly eliminate the acceleration-induced inaccuracy. Resultantly, improvement in estimation performance is limited. In this paper, the kinematic constraint associated with a ball joint is formulated to eliminate the acceleration-induced inaccuracy. Then, the constraint is augmented in a state-of-the-art attitude estimation KF. In our experimental results, the proposed KF with constraint was superior to the conventional KF by an average of 6.31 degrees and a maximum of 11.22 degrees.