221 Hybrid Extended /Unscented Square Root Particle Filter with Vehicle Dynamic Based Switching

MEMS and GPS are in use since decades for navigation of autonomous land and airborne vehicles. To overcome non linearity of navigation mechanization equations unscented filters (UKF) were developed. UKF are only good for Gaussian noise. If the noise depict non Gaussian behavior particle filters are best to estimate. A particle filters which uses unscented filter to propagate system state matrix and covariance is called unscented particle filter (UPF).UPF is in use for Navigation and Self Alignment. The paper has used Particle filter with Unscented Square root (SRUKF) filter called unscented square root particle filter or SRUPF. To reduce computational burden of filter a switching based on low, high dynamics (LHDS) of the trajectory is utilized. In the regions where vehicle dynamics is slow enough to be considered as linear extended filter is used and in the regions where vehicle dynamic is fast SRUPF is utilized. The MEMS sensors simulated are of automotive grade with high noises and large initial attitude errors .GPS outages of 1–2 minutes are simulated to verify performance. The simulation results showed that due to increased stability of SRUKF, this system is more efficient towards increased noise levels and due to LHDS technique; the averaged computation time is relatively 30–50 % less than UPF.