Genetic Algorithm Based Aggregation for Federated Learning in Industrial Cyber Physical Systems

During the last decade, Industrial Cyber-Physical Systems (ICPS) have attracted a significant amount of interest from industries as well as academic institutions. These kinds of systems have proved to be very complicated, and it may be a difficult task to get a handle on their architecture and make sure everything works properly. By putting up a framework for federated learning that we've dubbed FedGA-ICPS the purpose of this study is to address some of the difficulties that are associated with the performance and decision-making aids provided by ICPS. To begin, we launch an ICPS modeling formalism with the goal of specifying the structure and behaviour of such systems. FedGA-ICPS then conducts an analysis of the performance of the industrial sensors based on the data supplied by the ICPS from the industrial sensors by putting forth locally integrated learning models. Following that, a genetic algorithm drives federated learning in order to quicken and enhance the aggregation process. In the end, transfer learning is used so that the learned parameters of the models may be distributed across a variety of limited entities. FedGA-ICPS has been implemented on MNIST, and the results have been rather significant.