Critical Reasoning

Model-based diagnosis algorithms face a combinatorial explosion. To combat this explosion, this paper presents a fundamentally new architecture, IMPLODE, which constructs an abstract representation of the environment, the conflict, and the diagnosis spaces using a sensitivity analysis of assumptions. Experimental results show that the most dramatic improvement is obtained for circuits which are the most difficult to diagnose using previous algorithms. Moreover, typical sources of combinatorial explosion, such as reconvergent fanout, are a source of combinatorial implosion for IMPLODE.