A new on-line exponential parameter estimator without persistent excitation

In this paper we propose a new algorithm that estimates on-line the parameters of classical vector linear regression equation Y=Ωθ, where Y∈Rn,Ω∈Rn×q are bounded, measurable signals and θ∈Rq is constant unknown parameters, even when regressor Ω not persistently exciting. Moreover, convergence parameter estimator global exponential given for both, continuous-time discrete-time implementations. As an illustration example consider problem estimation time-invariant system, input signal sufficiently exciting, which known to be necessary sufficient condition solution with standard gradient or least-squares adaptation algorithms.