Parametric inference of autoregressive heteroscedastic models with errors in variables

We propose a consistent and asymptotically normal parametric estimator for autoregressive heteroscedastic models with errors in variables based on contrast minimization and give an example for a discrete time observed CIR process with additive noises.