Knowledge representation and reasoning for fault identification in a space robot arm

The construction of diagnostic systems able to manage tasks like fault detection, fault localization or fault identification in autonomous spacecraft is currently considered a big challenge for Artificial Intelligence techniques. In the present paper we report on the work done inside a project sponsored by ASI (the Italian Space Agency) aimed at building an intelligent multi-agent system for the control and supervision of the SPIDER Manipulation System with some form of interaction with the human operator. In particular, we will discuss knowledge representation and reasoning issues related to the construction of a model-based diagnostic component which has to co-operate with other modules of the system. An in-depth analysis of FMECA documents has guided the modeling of the domain knowledge on the faulty behavior of SPIDER. In this paper, problems related to the choice of the suitable modeling formalism involving abstractions and interaction among components are formally addressed, as well as the definition of innovative diagnostic strategies able to deal with the huge number of possible diagnoses that may arise during the diagnostic activity. The paper reports some preliminary results of the prototypical version of the diagnostic module on simulated data.