Abstract:

In this paper, a novel adaptive filter algorithm is proposed to improve position accuracy of the GPS/BeiDou-2(BD-2) integrated satellite navigation system for a maneuver vehicle in the presence of model and noise statistics uncertainty. Firstly, a new adaptive UKF is proposed to solve the problem of noise uncertainty by use of an adaptive factor obtained from the smoothing filter based on the residual error treated as the uncertainty of process noise.This adaptive method significantly reduces the effect of noise uncertainty on filtering performance.Secondly,in our method, a set M of models is taken into consideration and a ‘soft switching’ among the model set based on model probability is carried out by using interactive multiple model (IMM) algorithm, by which the position error caused by the model uncertainty can be reduced. The simulation results verify the effectiveness of the proposed method, and show preferable position accuracy for the complex maneuver vehicle.

Key words: Adaptive； Unscented Kalman filter；Interactive multiple model； Model uncertainty；Noise statistics uncertainty, Satellite navigation system