Incorporating Geometric Constraints into Rule-Based Systems Using Nonlinear Optimization.

摘要

A transformation procedure is presented for converting rule-based systems with nonlinear constraints into nonlinear optimization problems. The transformation procedure is then applied to an illustrative site identification system. Nonlinear optimization problems are constructed both for the subset of rules consisting only of logical predicates and for the complete system. Experimental results obtained by solving the nonlinear optimization problems for the site identification system yield a complete description of valid variable bindings for all rules in the system. This procedure was developed in response to the need to solve logical deployment problems for the U.S. Army. In this application domain, numerical, geometric, and geographic constraints must be incorporated with logical constraints in a uniform framework as in human inference. Rules in the system must be able to express various types of relationships, including relationships in the form of nonlinear constraints. Modeling rule-based systems as nonlinear optimization problems provides a powerful, uniform framework with the flexibility to handle mixed data types and numerical and geometric relationships. Rule-based systems, Nonlinear constraints, Site identification, Logical deployment problems, Geographic constraints.