In the next generation surveys, the discovery of moving objects can be successful only if an observation strategy and the identification/orbit determination procedure are appropriate for the diverse apparent motions of the target sub-populations. The observations must accurately measure the displacement over a short interval of time; observations believed to belong to the same object have to be connected into tracklets. Information contained in tracklets is in most cases not sufficient to compute an orbit: two or more of them must be identified to provide an orbit. We have developed a method for recursive identification of tracklets allowing an unbiased orbit determination for all sub-populations and efficient enough to cope with the data flow expected from the next generation surveys. The success of the new algorithms can be easily measured only in a simulation, by consulting a posteriori some “ground truth”.
We present here the results of a simulation of the orbit determination for one month of operations of the future Pan-STARRS survey, based upon a Solar System Model with a downsized population of Main Belt asteroids and a full size populations of Trojans, NEO, Centaurs, Comets and TNO. The results indicate that the method already developed and tested to find identifications of NEO and Main Belt asteroids are directly applicable to Trojans. The more distant objects often require modified algorithms, fitting orbits with only 4 parameters in a coordinate system specially adapted to handle very short arcs of observations. These orbits are mostly used as intermediate results, allowing to find full solutions as more tracklets are identified.
When the number density of detections is as large as expected from the next generation surveys, both joining observations into tracklets and identifying tracklets can produce some false results. The only reliable way to remove them is a procedure of tracklet/identification management. It compares the tracklets and the identifications with a complex logic, allowing to discard almost all the false tracklets and all the false identifications. However, the distant objects still present a challenge for orbit determination: they require three tracklets in separate nights. If this requirement is met we have found no problem in achieving an unbiased orbit determination for all populations. Further work will lead to more advanced simulations, in particular by introducing a realistic model for astrometric and photometric errors.