Decentralized Fixed-Interval Smoothing Algorithms

Abstract

The Weineret-Desai smoother formula is applied to derive new decentralized fixed-interval smoothing algorithms for a decentralized estimation structure consisting of a central processor and of M local processors. Such algorithms are based on decentralizing the estimates of global backward information filter and obtained from the use of the superposition principle in scattering framework. The smoothing update problem is also investigated to illustrate the application of the proposed algorithms. The emphasis is on computational efficiency, independence of local a priori statistics, and flexibility of implementation.