Knowledge Representation and Reasoning for Fault Identification in a Space Robot Arm

The constrnction of diagnostic systems able to manage tasks like fault detection, fault localization or fault identificat.ion in autonomous \pacecraft is currently considered a txg c-hallenge for Artific~al Intell~gence technique:, I11 the present paper we report on thc work done inside a project sponsored 11y AS1 (the Italian Space Agency) aimed at h~~ildinagn intelliger~t multi-agent syste~ n for the control and superv~sion of thc Sl'IDF:I< Manipulation Systerri with some for111 of int,eract,ion with the human operat. os. III part,icular, we will discuss knowledge represe~lt.at,iona nd reasoning issues related to t hc cmlst,ruct,ion of a modcl-based diagnost ic c~omponent which has t,o co-operate \vit,ll ot8her modules of the ~yst~emA.1 1 indrpt, h analysis of FMECA docunlent~s has ~ItdIed t he inodeling of the domain knowledge on the faulty behavior of SPIDER. I11 th~sp aper, prohlcnls related to the choice of the st~ltatm~ol~de l~ngfo rrnal~sniln volving abstrac-t~oi~asn d ~nteractiona mong compoticnts arc formally addressed, as well as the definrt~ono f Innovat~ved ~agnost~sctr atrgleb ablf. to deal w~thth e huge riunlber of possi- 111~ diagnos(>sthat mag arise during t,he diagr~ ostic act,ivit,y. The paper report,s some prelilninar); result,s of t,he prot,ot,ypical version of the diagnostic module on simulated data