We present a proper motion measurement for the halo globular cluster Pyxis, using HST/ACS data as the first epoch, and GeMS/GSAOI Adaptive Optics data as the second, separated by a baseline of about 5 years. This is both the first measurement of the proper motion of Pyxis and the first calibration and use of Multi-Conjugate Adaptive Optics data to measure an absolute proper motion for a faint, distant halo object. Consequently, we present our analysis of the Adaptive Optics data in detail. We obtain a proper motion of mu_alpha cos(delta)=1.09+/-0.31 mas/yr and mu_delta=0.68+/-0.29 mas/yr. From the proper motion and the line-of-sight velocity we find the orbit of Pyxis is rather eccentric with its apocenter at more than 100 kpc and its pericenter at about 30 kpc. We also investigate two literature-proposed associations for Pyxis with the recently discovered ATLAS stream and the Magellanic system. Combining our measurements with dynamical modeling and cosmological numerical simulations we find it unlikely Pyxis is associated with either system. We examine other Milky Way satellites for possible association using the orbit, eccentricity, metallicity, and age as constraints and find no likely matches in satellites down to the mass of Leo II. We propose that Pyxis probably originated in an unknown galaxy, which today is fully disrupted. Assuming that Pyxis is bound and not on a first approach, we derive a 68% lower limit on the mass of the Milky Way of 0.95*10^12 M_sun.

We present in this study a first analysis of the astrometric error budget of absolute astrometry relative to background galaxies using adaptive optics. We use for this analysis multi-conjugated adaptive optics (MCAO) images obtained with GeMS/GSAOI at Gemini South. We find that it is possible to obtain 0.3 mas reference precision in a random field with 1 hour on source using faint background galaxies. Systematic errors are correctable below that level, such that the overall error is approximately 0.4 mas. Because the reference sources are extended, we find it necessary to correct for the dependency of the PSF centroid on the used aperture size, which would otherwise cause an important bias. This effect needs also to be considered for Extremely Large Telescopes (ELTs). When this effect is corrected, ELTs have the potential to measure proper motions of dwarfs galaxies around M31 with 10 km/s accuracy over a baseline of 5 years.