A proposed attitude estimator with reliability test criteria for sensor data fusion

Abstract

The attitude computed from angular integration is priori, which should be optimized to the optimal attitude determination. Previous researches mainly focus on compensating for angular velocities or quaternions’ derivative before the integration operation. This paper presents a new-structure attitude estimator with direct compensation for the priori estimate. A gradient descent method is designed as the attitude optimizer, which generates the attitude compensation by fusing measurements of accelerometers and magnetometers. The orientation of homogeneous field is adopted as the absolute reference. That ensures the attitude estimation is not affected by the varied size of homogeneous field. Moreover, the proposed attitude estimator is embedded with a set of reliability test criteria for accelerometers and magnetometers. It ensures the reliability of sensors’ data. Several experiments are carried out on real hardware. Results show that the proposed attitude estimator is effective. The computational efficiency of the non-Kalman filters with higher accuracy can be achieved.