The data compression problem for the “GAIA” astrometric satellite of ESA

Space-based astrometry has a great tradition at ESA. The first space-based astrometric satellite in history, “Hipparcos”, was launched by ESA in 1989 and, in spite of orbital problems, was able to accomplish almost all of its tasks until it was finally shut down in 1993. The results of the Hipparcos mission were published by ESA in 1997 in the form of six CD-ROMs: the Hipparcos Catalogue contains 118,218 entries with median astrometric precision of around 1 milliarcsec, and specific results for double and multiple systems. In practice, Hipparcos drew for the first time the three-dimensional “map” of the spherical region of the Galaxy surrounding the Sun and having a radius of roughly 1,000 light years.

Then, in 1995, ESA launched the study of a new astrometric satellite, named “GAIA” and about a hundred times more powerful than Hipparcos, i.e. with median astrometric precision of around 10 microarcsec. This new satellite is intended to measure the parallaxes of over 50 million stars in the Galaxy, at least for the brightest stars, and this would mean to “draw” the three-dimensional map of the whole Galaxy, reaching out even to the Magellanic Clouds, 180,000 light years away.

The team of European scientists and engineers now designing GAIA, however, is facing hard technological difficulties. One of these is the design and coding of radically new and ultra-powerful mathematical algorithms for the on-board compression of the 50-million-stars data that GAIA will send to Earth from its intended geostationary orbit. Preliminary estimates of the raw data rates from the GAIA focal plane, in fact, are of the order of a few Gigabits per second. To reduce the data stream to the envisaged telemetry link of 1 Megabit per second, on-board data compression with a 1 to 1,000 ratio is the target. Clearly, this is far beyond the capabilities of any lossless compression technique (enabling compression ratios of 1 to some tens), and so some “wise” lossy compression mathematical procedure must be adopted.

In this paper a GAIA-adapted lossy data compression technique is presented, based on the Karhunen-Loève Transform (KLT). The essence of this method was already used by NASA for the Galileo mission when the large antenna got stuck and the mission was rescued by re-programming the on-board computer in terms of the KLT. That transform was officially named ICT — “Integer Cosine Transform” — by the NASA-JPL team led by Dr. Kahr-Ming Cheung. But the KLT here described for GAIA will of course differ from the JPL one in many regards, owing to the advances in computer technology.

Finally, estimates are also given about the possibility of using the KLT for onboard data compression in case GAIA is going to be put into orbit around the Lagrangian point L2 of the Earth-Sun system, and, above all, in case the number of stars to be observed is actually raised from 50 millions to one billion, as ESA currently appears to be likely to pursue.