A framework for autonomous search in the Eclipse solver

Abstract

Autonomous Search (AS) is a special feature allowing systems to improve their performance by self-adaptation. This approach has been recently adapted to Constraint Programming (CP) reporting promising results. However, as the research lies in a preliminary stage there is a lack of implementation frameworks and architectures. This hinders the research progress, which in particular, requires a considerable work in terms of experimentation. In this paper, we propose a new framework for implementing AS in CP. It allows a dynamic self-adaptation of the classic CP solving process and an easy update of its components, allowing developers to define their own AS-CP approaches. We believe this will help researchers to perform new AS experiments, and as a consequence to improve the current preliminary results. © 2011 Springer-Verlag.