Recursive estimation of the continuous-time process parameters

The problem of continuous-time process parameter identification is considered. Filtered input-output process signals are used to create a linear differential equation governed by the same continuous-time process parameters. The estimation scheme is implemented by sampling the filtered signals and using a recursive least squares algorithm (RLS). The choice of filter leads to different parameter convergence properties. Conditions for parameter convergence are established in terms of frequency content of the input signal. The convergence rate is also analysed and an upper bound on the parameter error norm is given. The relation between choice of filter, sampling time selection and quality of the estimates is discussed and exemplified with simulation examples.