A multi-level search framework for asynchronous cooperation of multiple hyper-heuristics
Djamila Ouelhadj, Sanja Petrovic, Ender Özcan · 2009
In this paper, we propose an agent-based multi-level search framework for the asynchronous cooperation of hyper-heuristics. This framework contains a population of different hyper-heuristic agents and a coordinator agent. Each hyper-heuristic agent operates on the same set of low level heuristics, while the coordinator agent operates on top of all the hyper-heuristic agents. Starting from the same initial solution, each hyper-heuristic agent performs a search over the space generated by the low level heuristics. The hyper-heuristic agents cooperate asynchronously through the coordinator agent by exchanging their elite solutions. The coordinator agent maintains a pool of elite solutions and manages the communication between the hyper-heuristics agents.