Using the generalized assignment problem in scheduling the ROSAT space telescope

ROSAT is a satellite borne X-ray observatory. Mission planning consists of scheduling up to half a year in advance a preselected large collection of targets. It is shown that a relaxation of the ROSAT scheduling problem can be modeled as a generalized assignment problem and that the solutions obtained with this method lead to good solutions of the non-relaxed problem. Other heuristics as well as upper bounds on the quality of solutions are discussed.