Stable embedding based particle filter for state estimation of rigid body

Abstract

In this paper, a stable embedding technique is applied to a particle filter for estimating the state of a rigid body system which is defined on a manifold. The stable embedding technique extends a system defined on a manifold to an ambient open set in Euclidean space and modifies the system so that the extended system is stable on the manifold. The performance of a particle filter with stable embedding is compared with a standard particle filter. Particle filter with stable embedding shows higher performance than the existing standard particle filter, and performance gap remains even though the number of particles increase. We also compare the performance with extended Kalman filters and show that the performance improvement of the particle filter with stable embedding is on par with the performance improvement of the extended Kalman filter with stable embedding.