Skip to Main Content
This paper focuses on an extensible framework for the development of parallel/distributed population-based algorithms. This framework uses mobile agents launched into different hosts on available networked PCs and cooperating among them to solve large combinatorial problems efficiently. The execution environment used to realize our framework is based on the JADE technology. In addition, we define a new information exchange strategy based on a dynamic migration window method and a selective migration model. A parameters adaptation model is also proposed. This model is used to adjust different parameters/operators of the genetic algorithm executed by each mobile agent. The proposed framework has been experimented on an extended set of Earliness and Tardiness Production Scheduling and Planning Problem (ETPSP). Several experiments are carried out on different computer networks of different sizes. Results obtained show the advantages and efficiency of our approach.