Jet transport particle filter for attitude estimation of tumbling space objects

The particle filter (PF) is one of the most potential methods for nonlinear state estimation of spacecraft on account of its accuracy and stability. However, the heavy computational burden limits its application in real-time estimations. In this paper, we propose an improvement for the PF based on the Jet Transport (JT) theory, and apply it to the real-time attitude estimation of tumbling space objects. The main innovation of the Jet Transport Particle Filter (JTPF) is to employ the JT technique in the particle evolution process, rather than using the numerical integration as the classical PF, so as to reduce the computational burden of the algorithm. Furthermore, the proposed JTPF uses the multiplicative error quaternion to avoid further errors in the normalizing process, and the regularization technique to avoid the particle degeneracy. The JTPF is tested in three scenarios with different state and observation dimensions. Monte Carlo simulations demonstrate that the JTPF has a similar accuracy as the classical PF, and costs only 7% ∼ 13% of processing time as the latter. Moreover, some empirical principals are summarized about the optimal JT expansion order and particle number of the JTPF.