Towards a distributed hyperheuristic deploy architecture

The hyperheuristic term is known in the optimization field as an automated methodology for selecting or generating heuristics to solve hard computational search problems. From the design perspective, it is based on decoupling the solving intelligence from the domain expertise, allowing to reuse the same solver for multiple, usually related problem domains. There are few works in which hyperheuristics have been designed and evaluated in distributed environments. In this paper, we propose a conceptual design of a distributed hyperheuristic architecture, from the problem domain deploying perspective, which allows to communicate different optimization environments (such as solver and domain) and to offering a "solving service". Different problems domains could be addressed using an encapsulated hyperheuristic solver, and through well defined interfaces, users can provide different heuristic components to perform the optimization process. The proposed architecture is only an initial step for which different modeling, design and implementation issues must be addressed. Such research should be focused on defining how conceptual design contributions must be leveraged to implement well defined interfaces, capable of connecting hyperheuristic solvers and problem domains within distributed environments.