Censored regression distributed functional link adaptive filtering algorithm over nonlinear networks

摘要

Wireless sensor network (WSN) is an important part of the Internet of Things (IoT) and has emerged in various new forms, such as smart home, smart city, and intelligent manufacturing system. Due to its high reliability, distributed estimation over nonlinear WSNs is one of the most active fields in recent years. In this paper, a novel distributed functional link least mean square (DFLMS) algorithm based on rblackthe diffusion strategy is proposed, in which the diffusion functional link network (DFLN) is used to model the nonlinear dynamic behavior of the distributed system. In particular, by using different orthogonal polynomials, we develop four types of DFLNs, i.e., trigonometric DFLN (TDFLN), Legendre DFLN (LDFLN), Chebyshev DFLN (CDFLN), and Hermite DFLN (HDFLN). However, the censored measurement caused by the range of sensors brings great challenges to the traditional distributed nonlinear estimation. To tackle this problem, a censored regression-distributed functional link adaptive filtering (CR-DFLAF) algorithm is further proposed. Compared with the DFLMS algorithm, the CR-DFLAF algorithm can compensate the estimated bias in the CR scenario at the price of slightly increased computational complexity. Simulations involving two distributed nonlinear networks verify the effectiveness of the proposed algorithms.