Space missions present many challenges for today's information technology, by providing intrinsically difficult problems. These problems enclose the many critical requirements of space missions, namely efficiency, flexibility and robustness.
Various techniques used in Artificial Intelligence have found an application in the context of space missions, among which Planning and Scheduling. These disciplines are studied by a research group which has now been active for many years at the Institute for Cognitive Science and Technology of the Italian National Research Council (ISTC-CNR). The Planning & Scheduling Team (PST) has developed specific algorithms and methods for the representation and integration of these advancements in software applications.
More specifically, between November 2000 and July 2002, the PST has worked on a specific planning problem for the MARS EXPRESS mission. The objective of this ESA mission is to launch a satellite in 2003, which will reach and begin orbiting Mars in 2004 for a total of two years.
The result of this project (Contract ESA-ESOC 14709/00D/IM) is a software system, called MEXAR, which is currently in advanced experimentation at the ESA-ESOC center in Darmstadt. The system is responsible for synthesizing the transmission operations for the gathered data towards Earth, synchronizing transmission with Earth visibility time windows. The objective is to optimize the use of the on-board memory stores in order to avoid data loss while assuring reasonable access times for scientific data. It has been proven that the underlying problem is of elevated combinatorial complexity.
MEXAR models the problem through an ad-hoc constraint representation which has been developed solely for this purpose.
Based on this representation, a set of algorithms extracts solutions of varying degrees of efficiency. Finally, a module which provides an interface to the resolution module has been developed. The interface empowers the users (the ground-based mission planners) to intervene at various levels of the resolution process. For instance, the user can simply inspect the current problem, choose parameters for its resolution, change the current problem of even compose different algorithms in order to explore the space of solutions.
MEXAR is an example of a methodology for automatic problem resolution which makes use of an integration of various techniques (constraint programming, planning and scheduling heuristics, intelligent interaction) in order to solve real problems. This particular application shows how preserving the role and competences of human planners is a winning strategy because it combines the specific and rather distinct capabilities of humans and automatic problem solvers. This generates a well-integrated synergy which enhances the power to intervene of the human operator.
Vedi anche:
Immagini:
