Digital Circuit Network Evolution Based on the Virtual Reconfigurable Platform

Abstract

Evolvable hardware, as a product of the integration of evolutionary computation and reconfigurable device, can change its architecture and function dynamically and autonomously by interacting with environment. Nowadays, three key issues about EHW are circuit representation methods, genetic algorithms and reconfigurable hardware platform. In this paper, based on the similarities between biological neural network and digital circuit network, a circuit network model was constructed. The digital circuits were encoded by using of a matrix group. Corresponding algorithm was designed, including the selection and transposal operator. The structure and configuration bit-stream format of common commercial FPGA are unknown in research layer. In order to get rid of these problems, a reconfigurable multi-layered feed forward circuit network was implemented in a commercial FPGA chip based on the virtual reconfigurable platform. Its internal structure and configuration bit-stream format is known. So, a digital circuit evolution system was implemented based on the genetic algorithm and hardware platform.