Closed-loop identification of continuous-time systems from non-uniformly sampled data

In this paper, the instrumental variable (IV) and expectation-maximization (EM) methods are combined to identify a continuous-time (CT) transfer function model from non-uniformly sampled data obtained from a closed-loop system. A simple version of Box-Jenkins (BJ) model is considered, where the noise process is parameterized as a CT autoregressive (CAR) model. The advantage of considering CT models is to get a invariant solution while handling non-uniformly sampled data. The performance of the proposed method is evaluated by a simulation example.