Cost Reference Particle Filtering Approach to High-Bandwidth Tilt Estimation

Abstract

In this paper, cost reference particle filter (CRPF) approach in estimating 1-D “tilt” of a vehicle attitude is proposed. CRPF has a couple of advantageous features compared to standard particle filtering; particularly, it does not require noise statistics in its application. H_ filter (HF) has common features as that of CRPF. The extended HF (EHF) is employed, which uses the approximate linearization of the nonlinear measurement function as the extended Kalman filter is extended. The performance of both approaches is investigated and compared in this paper. Low-cost “accelerometer” and “gyroscope” sensors are cooperatively employed instead of inclinometer in measuring the tilt. Simulation results show that CRPF outperforms EHF in estimating the tilt due to its robustness against the nonlinearity of the measurement equation, whereas EHF outperforms CRPF in estimating the tilt rate whose measurement equation is linear. Notably, an efficient CRPF outperforms EHF in tracking the tilt with just ten particles.