Reflective, self-adaptive problem solvers

Abstract

Problem-solving systems, situated in the real world are faced with a great challenge, that is, the dynamic nature of their environment. In any realistic environment the state of the world changes, and therefore, the system's knowledge about the world often becomes incomplete and incorrect. Furthermore, the constraints and the requirements imposed on the system's behavior may also evolve, and as a result, the system's functional architecture may become insufficient to meet the requirements of the evolving task environment. In principle, we would like our systems to be able to adjust themselves in their environments and to sustain quality performance across such environmental changes. To enable a system with the capability of self-adaptation, we have developed a framework for endowing it with the competence of reflection. In this framework, the system's problem-solving behavior is modeled in terms of a SBF model. This model captures a deep comprehension of the system's task structure, world knowledge and their inter-dependencies. The knowledge captured in the SBF model of a system enables it, when it fails, to identify the need to update its world knowledge and also appropriately redesign its functional architecture.

This work has been supported by the National Science Foundation (research grant IRI-92-10925), the Office of Naval Research (research contract N00014-92-J-1234), and the Advanced Projects Research Agency. In addition, Stroulia's work has been supported by an IBM graduate fellowship.