Project Overview

The Adaptive Constraint Engine (ACE) seeks to automate the application of constraint programming expertise and the extraction of domain-specific expertise. Under the aegis of FORR, an architecture for learning and problem solving, ACE learns search-order heuristics from problem solving experience. ACE automates the construction of problem-class-specific solvers using machine learning techniques: each solver will be an adapted, collaborative "community" of heuristics appropriate for its problem class. This will allow more efficient and appropriate selection of methods for specific problem classes, and the discovery of novel solution strategies. ACE is both a test-bed for CSP research and a discovery environment for new algorithms.