ECAI 2004 Conference Paper

[full paper]

A Troubleshooting Approach with Dependent Actions

Eylem Koca, Taner Bilgic

Man-made systems often break, and it is highly desired that they are fixed as fast and as cheap as possible. However, achieving this optimally is a very complex problem if there is action dependency, that some actions may fix the same fault. When this is the case, ordering of ordinary efficiencies fails to guarantee an optimal action sequence. Thus, it is aimed to find a heuristic method that adjusts action efficiencies for better results. For this end, dependency sets are defined, and it is shown that these sets are what complicate the problem and that an optimal sequence is achievable once they are solved optimally. When adjusting efficiencies, the value in failing an action is utilized, which is modeled as the increase in the weighted efficiencies of remaining actions. The weight for each action is assumed to be its scaled efficiency, which is found by normalizing all efficiencies. This approach is implemented in the 1-SEA troubleshooter, and its performance is tested on the basis of expected cost of repair in 14 problems and compared to two other troubleshooters from the literature. The results indicate that the approach is very effective and time-efficient, and promise successful practices in complex systems.

Keywords:

Citation: Eylem Koca, Taner Bilgic: A Troubleshooting Approach with Dependent Actions. In R.López de Mántaras and L.Saitta (eds.): ECAI2004, Proceedings of the 16th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2004, pp.1041-1042.