Self-Adaptive Software

University Of California-Irvine, Irvine CA

Investigators

Abstract

Proposal Number 0430066 Principal Investigator: Richard N. Taylor Self-Adaptive Software Adaptive behavior is required of many classes of software systems in which untimely human responses to change are insufficient. Self-adaptive software systems address this requirement by continually observing themselves and their environment in order to effect autonomous modifications to their behavior. This research aims to develop a methodology for the development of self-adaptive software centered on ensuring a high-level of adaptation policy visibility, enforcing a strict decoupling of policy specification from system implementation, and supporting the independent and dynamic runtime evolution of adaptation policies. To achieve these objectives, the design of self-adaptive systems through the integration of architectural models and knowledge-based techniques is investigated. A specification of adaptation policies based on architectural knowledge independent from system semantics is included in software architecture models; these artifacts are leveraged to implement knowledge-based reasoning to determine necessary system modifications. Research results are evaluated in the context of space exploration systems, which provide a rich set of examples sorely needing self-adaptive capabilities. This research enables the principled architecting of self-adaptive systems by elevating and specifying adaptation policies as first-class architectural elements supported by the novel integration of knowledge-based techniques for reasoning over the space defined by architectural knowledge, adaptation policy, and system modifications.