5G/GNSS Integrated Vehicle Localization with Adaptive Step Size Kalman Filter

This paper investigates an adaptive step size Kalman filter for 5G/GNSS integrated positioning of land-vehicles to deal with model uncertainties and multi-frequency measurements. A set of system and measurement models are established including a nonlinear dynamic vehicle model as the prediction model for the fusion filter. A robust stepwise fusion method is introduced to address model uncertainties, which is based on an adaptive Kalman filter with a step size adaptation mechanism involved to deal with multi-frequency measurements. The resulted fusion-based localization algorithm achieves a balance of positioning accuracy and computational cost. Both simulation and experimental results demonstrate that the proposed method outperforms the state-of-the-art methods in urban environments.