7
$\begingroup$

The Rosetta spacecraft has reached a comet in August 2014 and sent a lander on its surface. It was one of the greatest science news of the year.

Automated Planning and Scheduling techniques are really important in the operations of these kinds of missions. At the last ICAPS there was even an invited talk by one of the staff members involved in the automated planning technologies used in the spacecraft.

Is there any published paper on the planning technologies used in the spacecraft? I've only found this one, but it does not appear to have been published in conferences or journals. Has it or something similar been published somewhere?

$\endgroup$
6
$\begingroup$

Well, as far as I can tell, there was no specific automated planning tool for the Rosetta mission as we understand automated planning systems in ICAPS, which is the most prominent conference on Automated Planning and Scheduling.

The hardest part was to plan the journey from earth to a common location where both the spacecraft and the comet could meet and this was done mainly in two stages: one making the spacecraft orbit around some bodies to save energy (cruise phase) and this was mainly done with astronomical computations; after hibernating for a good number of years, it had to wake up and fly (now using its own power) to meet the comet. There was, additionally, another scheduling problem in the final landing operation of Philae. The last two operations were self-automated.

I asked someone with a better knowledge of Rosetta and this person told me (right now) that there were mainly scheduling systems (as they are also called in the paper you mention) and that most of the operations were self-automated (ie., the main difficulty was that Rosetta was so far that it had to be able to take some decisions on its own, but still it received orders from the ground segment chiefs in ESTEC, Germany).

This said, I do not think we will ever see a paper about how to use PDDL (Planning Domain Definition Language) for Rosetta, or how to use LAMA, Fast-downward, FF (or any temporal PDDL planner) or anything of the kind (I mean, the kind of research we all do at ICAPS) but something instead far more domain-dependent.

On the other hand, it is rather unusual for engineers and scientists at ESA to publish papers at scientific venues such as ICAPS, ECAI (European Conference on Artificial Intelligence), AAAI (Association for the Advancement of Artificial Intelligence), IJCAI (the International Joint Conferences on Artificial Intelligence). They instead submit their papers to the SpaceOps symposium. I found a number of papers there:

Geurts K., Fantinati C., Ulamec S., Willnecker R. Rosetta lander: On-comet operations preparation and planning. 13th International Conference on Space Operations, SpaceOps 2014; Pasadena, CA; United States; 5 May 2014 through 9 May 2014

Haddow, CR., Werner D. Mission planning framework - Building the rosetta and bepicolombo planning systems. 13th International Conference on Space Operations, SpaceOps 2014; Pasadena, CA; United States; 5 May 2014 through 9 May 2014

Moussi, A. Fronton JF., Gaudon P., Hallouard D., Mangeret M., Delmas C., Lafaille, V., Durand J. PHILAE lander: A scheduling challenge. 13th International Conference on Space Operations, SpaceOps 2014; Pasadena, CA; United States; 5 May 2014 through 9 May 2014

Eiblmaier MG., Blake R., Williams A., Lodiot S., Godfrey J., Bartesaghi M. Automating ESA's planetary missions: From concept to conclusion. 13th International Conference on Space Operations, SpaceOps 2014; Pasadena, CA; United States; 5 May 2014 through 9 May 2014

Hope this helps,

$\endgroup$
  • $\begingroup$ So if I had to say (in the introduction of a paper or a thesis) something like "automated planning is a fundamental problem for space exploration, see for example the Rosetta mission [1,2,3]", which papers would I refer to? I'd thought of using Rosetta as it is the biggest recent example (Norvig's AI book's introduction cite earlier Mars exploration missions but Rosetta just sounds cooler). The "Building the rosetta and bepicolombo planning systems" seems the best, at least looking at the title. What do you think? $\endgroup$ – gigabytes Jun 6 '15 at 10:08
  • $\begingroup$ P.S I understand they don't use PDDL or "conventional" planners. $\endgroup$ – gigabytes Jun 6 '15 at 10:10
  • $\begingroup$ this is all helpful but plz define some of the acronyms for a common (CS) audience eg ICAPS, PDDL, LAMA, Fast Downward, FF etc and it would be helpful to describe the basic CS oriented planning algorithms/ theory behind industry std methods (NASA etc) $\endgroup$ – vzn Jun 6 '15 at 15:54
  • $\begingroup$ @gigabytes: No, no, ... I expressed myself in very bad terms. Rosetta is actually a planning problem and you will do very well referring to it at the beginning of your PhD dissertation. That's nice indeed! What I meant is that the solution actually considered at ESA does not consist of the planning systems we typically study at ICAPS. I do not think they ever used PDDL for this problem but this does not disqualify it as a planning (&scheduling) problem. It is indeed! I hope I made myself more clear now ... $\endgroup$ – Carlos Linares López Jun 6 '15 at 22:17
  • $\begingroup$ @vzn: that was good advice! Thanks for your message. I have updated my response but please feel free to let me know if there is anything I could to improve it. $\endgroup$ – Carlos Linares López Jun 6 '15 at 22:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.