State estimation as switching between target simulated and driving data?

[hyunjimoon]

@tomfid put the importance of state estimation on my radar earlier.

Is making two model, first estimate a scalar $\theta$ using vector $Y$ (min $Y$ and $\tilde{Y}$, then replace the type of $Y$ from target simulated and driving data. Hazhir thinks this is more reliable but it may be against data is one realization of truth.