Parameter Estimation for ARX Models with Missing Data

Abstract This paper examines the problem of parameter estimation for discrete-time, linear, time-invariant stochastic ARX models in the presence of missing data. A recursive parameter estimation algorithm based on pseudo-linear regression methods is developed, and a theorem for the global convergence of the parameter estimates to their true values is presented A sufficient condition for convergence is the very strict passivity of a linear time-varying operator A novel frequency domain condition for verifying the passivity of the operator is presented, in the case of periodically missing data The performance of the proposed algorithm is illustrated via simulations.