Machine Learning for Constraint-based Configuration: A Survey

Abstract

Constraint-based configuration is a successful industrial application of symbolic Artificial Intelligence. It involves selecting a set of components, features, or services that satisfy a given set of user requirements. These requirements, specified by an individual user or a group, guide the configuration system in identifying a solution that aligns with both, user requirements and the constraints defined in the configuration knowledge base. As configuration tasks grow in size and complexity, there is a growing need to integrate machine learning (ML) for increasing algorithmic efficiency and quality of user interaction. This survey provides a comprehensive overview of approaches that combine ML with constraint-based configuration techniques. We highlight key developments including new developments related to the integration of Large Language Models (LLMs) and identify open research challenges.