Using genetic algorithms to design switched Ethernet industrial networks

The Ethernet network is increasingly used for industrial communications which are strongly time-constrained. This kind of network is intrinsically nondeterministic and does not enable to guarantee the communication end-to-end delays to be bounded. Nevertheless, from the traffic observed between the industrial devices, some network topologies improving the availability and the temporal performances can be designed. This paper describes the use of the genetic algorithm approach for the optimisation and the design of industrial control networks. We propose to solve the combinatorial problem which consists of finding the best distribution of the industrial devices on a switched Ethernet architecture with GA techniques.