HHmax

HHmax is an open-source software framework designed to support hyper-heuristic optimization for combinatorial problems. It provides a platform to develop, test, and benchmark hyper-heuristics, allowing researchers and practitioners to evaluate high-level control strategies that select among low-level heuristics across various problem domains. The project aims to separate problem modeling from heuristic selection to enable fair comparison of different approaches on common instances.

The architecture of HHmax centers on a modular engine that integrates several components. The core engine manages execution flow, time and resource constraints, and result logging. A problem model layer encodes instance data and evaluation functions for specific domains. A library of low-level heuristics offers basic operations such as local search moves or constructive procedures. A selector component implements the hyper-heuristic logic that chooses which low-level heuristic to apply next, while a benchmarking suite provides standardized experiment protocols, statistics collection, and reproducibility tools.

Key features include a plugin-based design for easy extension, multi-language bindings (notably Python and Java), and built-in templates for common domains such as vehicle routing, scheduling, and timetabling. HHmax supports parallel evaluation to speed up experiment runs and includes facilities for reproducible experiments, versioned configurations, and detailed performance logging.

Development history notes the project’s trajectory from its initial release to recent updates. Early versions introduced core architecture and parallel processing, with subsequent releases adding autoscheduling, improved configurability, and enhanced benchmarking capabilities. HHmax is used in academic research and industry settings to explore combinations of base heuristics and high-level control strategies while maintaining transparent experimental workflows.

The project is MIT licensed and hosted on a public repository with documentation, tutorials, and example experiments to help new users get started. See also hyper-heuristic, metaheuristic, and optimization library concepts.