Adaptive Robust Federal Kalman Filter for Multisensor Fusion Positioning Systems of Intelligent Vehicles

Authors:
Publisher website
Google Scholar
Abstract
Multi-sensor fusion positioning is an important technology for achieving high-precision positioning of intelligent vehicles in complex road scenes. However, existing fusion positioning algorithms are difficult to guarantee the positioning accuracy and robustness of intelligent vehicles in uncertain abnormal noise interference environments. Therefore, this paper proposes an adaptive robust federated Kalman filter based on student’s t-distribution (ARFKF-ST). In this method, to better describe the positioning system with heavy tail-non-Gaussian noise under uncertain interference, a hierarchical Gaussian state space system model is constructed based on the student’s t-distribution. Meanwhile, considering that complex and variable noise interference can exacerbate the uncertainty of the system, leading to the fluctuation of noise parameters in the traditional Kalman recursive process. Therefore, the variational Bayes inference method (VB) is used to estimate the state and noise parameters of subsystems in real-time and online, thereby improving the adaptive performance of the algorithm to abnormal noise interference. In addition, a diagonally weighted fusion strategy based on state independence is designed to estimate global positioning information, so as to eliminate the effect of disturbed subsystems on the localization performance of other subsystems. The simulation and experimental results show that ARFKF-ST has stronger abnormal noise suppression performance and filtering stability, and can ensure positioning accuracy and the robust performance of intelligent vehicles under different levels of abnormal noise pollution and interference intensity.
Funding Information
  • National Science Foundation of China (52072175)
  • Jiangsu Outstanding Youth Fund Project (BK20220078)
  • Jiangsu Key Research and Development Plan (BE2022053)
  • Postgraduate Research and Practice Innovation Program of Jiangsu Province (KYCX21_0187)

This publication has 32 references indexed in Scilit: