Data fusion of an attitude estimator for global localization of a robot

This paper focuses on the design and test results of an estimator based on fusing data from Inertial Measurement Unit (AHRS). Therefore in order to improve the quality of the attitude estimates, the covariance matrix of measurement noise is estimated in real time upon information obtained from the differential measurements, so that the estimator continually is “tuned” as well as possible. No a priori knowledge on the direction of the gravity vector in the inertial frame is required as these parameters can be also identified by the KF, relieving any need for calibration. With this approach, only the measurements of at least two non-collinear directional sensors are needed. Since the control laws are highly simple and a model based in an observer for angular velocity reconstruction is not needed, the proposed new strategy is very suitable for embedded implementations. Test results are presented showing the performance of the integrated AHRS to estimate the attitude of a mobile robot moving across uneven terrain. The global convergence of the estimation techniques is proved.