An agent-based model of hierarchic genetic search

An effective exploration of the large search space by single population genetic-based metaheuristics may be a very time consuming and complex process, especially in the case of dynamic changes in the system states. Speeding up the search process by the metaheuristic parallelisation must have a significant negative impact on the search accuracy.There is still a lack of complete formal models for parallel genetic and evolutionary techniques, which might support the parameter setting and improve the whole (often very complex) structure management.In this paper, we define a mathematical model of Hierarchical Genetic Search (HGS) based on the genetic multi-agent system paradigm. The model has a decentralised population management mechanism and the relationship among the parallel genetic processes has a multi-level tree structure. Each process in this tree is Markov-type and the conditions of the commutation of the Markovian kernels in HGS branches are formulated.