Using Java Mobile Agents and PSO for Implementing a Distributed Antenna-Optimization Platform [EM Programmer's Notebook]

Various techniques and platforms have been introduced for the effective use of distributed computational resources when using specialized computational electromagnetics (CEM) simulation software for designing antennas. These techniques usually impose significant entry barriers to researchers, in the form of a steep learning curve for specialized application programming interfaces (APIs), or limited access to resources such as clusters or computational grids. The current paper discusses a sophisticated approach for alleviating these barriers, with the fusion of the programming paradigms of object-oriented programming and mobile agents. A case study concerning the development of a distributed Particle Swarm Optimization (PSO) platform - an increasingly popular algorithm within the antenna-research community - is thoroughly analyzed and presented as an example of data-level parallelism. Various aspects of computational efficiency in terms of scaling have been examined, in order to estimate the merits of the proposed approach. The aim of the analysis is to stress the advantages of the above-mentioned techniques for the effective loose coupling of CEM programs with distributed resources, demonstrating ease of application, improved scalability, and future expandability.