Evolvable hardware with development

Recent progress in semiconductor technologies enable us to use programmable hardware devices, such as the Field Programmable Gate Array (FPGA). When such devices are combined with evolutionary computing techniques, we can develop hardware capable of evolving its circuit configurations on-the-fly. This is an attractive technology since it simultaneously attains real-time performance and adaptive capability. In this paper, we propose the incorporation of the developmental stage in order to generate a circuit configuration matrix. Using this approach, called the grammar encoding method, we evolve a set of graph rewriting rules, whereas the direct encoding method, a conventional approach, evolves the circuit configuration matrix itself. Our approach allows for non-trivial mapping of genotype and phenotype, as well as structure operation capability. Experimental results demonstrate that our approach consistently outperforms the direct encoding method.