This paper aims to present a method for improving the performance of the visual-inertial navigation system (VINS) by using a bio-inspired polarized light compass.
The measurement model of each sensor module is derived, and a robust stochastic cloning extended Kalman filter (RSC-EKF) is implemented for data fusion. This fusion framework can not only handle multiple relative and absolute measurements, but can also deal with outliers, sensor outages of each measurement module.
The paper tests the approach on data sets acquired by a land vehicle moving in different environments and compares its performance against other methods. The results demonstrate the effectiveness of the proposed method for reducing the error growth of the VINS in the long run.
The main contribution of this paper lies in the design/implementation of the RSC-EKF for incorporating the homemade polarized light compass into visual-inertial navigation pipeline. The real-world tests in different environments demonstrate the effectiveness and feasibility of the proposed approach.
