Adaptive parameter estimation using interior point optimization techniques: convergence analysis

Interior Point Optimization techniques have recently emerged as a new tool for developing parameter estimation algorithms [1, 2]. These algorithms aim to take advantage of the fast convergence properties of interior point methods, to yield, in particular, fast transient performance. In this paper we develop a simple analytic center based algorithm, which updates estimates with a constant number of computation (independent of number of samples). The convergence analysis shows that the asymptotic performance of this algorithm matches that of the well-known least squares filter (provided some mild conditions are satisfied). Some numerical simulations are provided to demonstrate the fast transient performance of the interior point algorithm.