Troubleshooting with Dependent Actions, Conditional Costs, and Questions

We study decision-theoretic troubleshooting and search for effective heuristic ap-proaches that can repair malfunctioning systems at low cost. We adopt the single-fault assumption and address questions, dependent actions and conditional costs. Our main idea is defining a measure of goodness for a troubleshooting decision and optimizing it by looking at only the next step. This approach is implemented in the 1-SEA troubleshooter and its performance is compared to two other troubleshoot-ers from the literature. The results indicate that 1-SEA is very effective and time efficient, and promises successful practices in complex systems. Key words: Decision-theoretic troubleshooting, fault analysis, probabilistic network