Readings in model-based diagnosis
Readings in model-based diagnosis
Bucket elimination: a unifying framework for reasoning
Artificial Intelligence
A Partially Solved Form for Heterogeneous Constraints in Disjunctive Normal Form
CP '02 Proceedings of the 8th International Conference on Principles and Practice of Constraint Programming
| Hi-index | 0.00 |
Model-based applications in engineering, such as diagnosis, configuration or interactive decision-support systems, require embedded constraint solvers with challenging capabilities. Not only consistency checking and solving, but also the computation of (minimal) conflicts and explanations are required. Moreover, realistic models of engineered systems often require the usage of very expressive constraint languages, which mix continuous and discrete variable domains, linear and non-linear equations, inequations, and even procedural constraints. A positive feature of the models of typical engineered systems is, however, that their corresponding constraint problems have a bounded and even relatively small density (induced width).We present here our relational constraint solver RCS that has been specifically designed to address these requirements. RCS is based on variable elimination, exploiting the low-density property. To analyze a set of constraints, RCS builds a so-called aggregation tree by joining the input constraints and eliminating certain variables after every single join. The aggregation tree is then used to compute solutions, as well as explanations and conflicts. We also report some preliminary experimental results obtained with a prototype implementation of this framework.