Compilation of Non-Contemporaneous Constraints

Abstract

Hierarchical execution of domain knowledge is a useful approach for intelligent, real-time systems in complex domains. In addition, well-known techniques for knowledge compilation allow the reorganization of knowledge hierarchies into more ecient forms. However, these techniques have been developed in the context of systems that work in static domains. Our investigations indicate that it is not straightforward to apply knowledge compilation methods for hierarchical knowledge to systems that generate behavior in dynamic environments. One particular problem involves the compilation of non-contemporaneous constraints. This problem arises when a training instance dynamically changes during execution. After dening the problem, we analyze several theoretical approaches that address non-contemporaneous constraints. We have implemented the most promising of these alternatives within Soar, a software architecture for performance and learning. Our results demonstrate that the ...