Intelligent SINS/RDSS integrated algorithms for land vehicle navigation

This is a discussion of the design of strap-down inertial navigation systems (SINS) and radio determination satellite service (RDSS) integrated navigation algorithms. The research aims at testing the effectiveness of artificial intelligence (AI)-aided Kalman filtering (KF) approaches for land vehicle applications. A back-propagation neural network (BPNN)-aided K*F algorithm and a fuzzy inference-based KF algorithm are presented in order to overcome the time delay of RDSS positioning provided by a double-star positioning system in China. Traditional KF causes biased solutions, and indeed, leads to filter instability easily since the time delay of RDSS positioning, in an active mode, is hard to be modeled and sometimes suffers from RDSS outages. Therefore, a fuzzy inference is used to correct the variance matrix of KE measurement noises adaptively; and a trained BPNN corrects the outputs of the Kalman filter. The algorithms proposed herein have been verified on real SINSIRDSS data. collected in land vehicle tests and are compared with other approaches. The results demonstrate that fuzzy inference-based KF algorithms improve the positioning accuracy to over 40 % better than KF algorithms, and BPNN-aided KF algorithms have the same precision as GPS which is the reference station In dynamic experiments without RDSS outages. The test results with RDSS outages indicate that the fuzzy inference-based KF is feasible but with positioning errors of hundreds of meters, so the BPNN-aided KF is designed to efficiently compensate for RDSS outages and improve system performance.