Passing Q matrices to UKF's RTS smoother

[mcphatty]

Hi,

I think I spotted a "bug", maybe it is not that important for most of the people.

In UKF's RTS smoothing function there is a parameter for passing Q matrices to the smoother ('Qs'), but the variable seems not to be referenced, instead self.Q is always passed to the unscented transform.

Thank you and kind regards,
Philipp

[mcphatty]

Just an addition, for the sake of completeness: It would be nice to be able to pass Q to UKF's predict() function as an override of the internal one, just like in KalmanFilter.