Low-cost Inertial Sensors Calibration : Algorithm, Design, Performance and Cases Study

Abstract

Inertial measurement units (IMU) have long been used in the personal and vehicular navi gation field. The sensor is robust because it provides kinematic information of the object it is attached to. Inertial navigation technique uses the kinematic information to navigate the object with very low computational resource requirement. The common drawback of the sensor is that its measurements contain multiple sources of errors and random noise. These sources decreases the navigation accuracy, especially if low-cost IMU is used. This thesis paper introduces errors in the IMU and the robust Unscented Kalman Filter algorithm to calibrate IMU. The sources of sensors are IMU, GNSS and wheel-tick odometer. A software architecture design of calibration algorithm for these sensors is demonstrated in this work. It is shown that the proposed algorithm could estimate the error in the IMU with high accuracy while it could fit very low computational and memory resources requirement, for example, on hand-held devices. The thesis will present real case studies, which applied the proposed algorithm to navigate and estimate error of IMU. It is shown in the case studies that the algorithm could perform in real-time with very low-end compu tation processing unit. That result opens the possibility to apply the algorithm on portable devices later.