|
Abstract : |
This paperpresents anewapproach for exploiting Truth Maintenance Systems(TMSs in which the inference engine can convey locality in its knowledge to the TMS. This approach is ideally suited for systems which reason about the physical world because much of knowledge is inherently local, i.e., the constraints for a particular component or process usually only interact with constraints of physically adjacent components and processes. The new TMSs operate with a set of arbitrary propositional formulae and use general Boolean Constraint Propagation(BCP) to answer queries about whether a particular literal follows from the formulae. Our TMS exploits the observation that if propositional formulae are converted to their prime implicates, then BCP is both efficient and logically complete. This observation allows the problem solver to influence the degree of completeness oftheTMS by controllinghowmany prime implicates are constructed. This control is exerted by using the locality in the original task to guide which combinations of formulae should be reduced to their prime implicates. We show that conveying locality to the TMS is an important strategy for qualitative physics problem solvers. For example, at a minimum formulae corresponding to a single component (or commonly occurring combinations) model should be compiled into prime implicates in order to minimize run-time cost. When confluence models are used, the results of using ourTMS subsume those of the qualitative reasolution rule. This approach has been implemented and tested both within Assumption-Based Truth Maintenance Systems and Logic-Based Truth Maintenance Systems. 1, |
