Distributed collaborative parameter estimation based on bias compensation

This paper presents the study of the problem of distributed parameter estimation by bias compensated recursive least squares (BCRLS) algorithm over adaptive networks. The nodes in the distributed network have a common objective to estimate parameter vector in a collaborative strategy. Traditional recursive least squares (RLS) estimator is biased in case that both the regressor and the output response are corrupted by stationary additive noise. A real-time estimation algorithm of noise variance is proposed, which nodes get the estimation of objective parameter bias. Based on collaborative strategy, we propose a diffusion bias compensated recursive least-squares algorithm. Simulation results show that the BCRLS algorithm has better estimation accuracy than traditional RLS algorithm, and compared with the local estimators, the diffusion BCRLS algorithm has lower mean square error (MSE).