Evolvable Hardware Design of Digital Circuits Based on Adaptive Genetic Algorithm

摘要

Evolvable hardware (EHW) is a thriving area of research which uses the genetic algorithm (GA) to construct novel circuits without manual engineering. GA has been widely implemented using software but have not gained an appreciable edge because of the huge computation time involved and easy to fall into local optimum. This has been a major hindrance to real-time applications. In order to improve GA, this paper proposes an adaptive algorithm for adjusting the probability value of genetic operator, called adaptive genetic algorithm (AGA). Adding an elite strategy to the algorithm further accelerates the convergence speed of the algorithm. In addition, a complete hardware evolution system based on Field Programmable SoCs (FPSoCs) for design of digital circuits is developed. The experimental results show that the proposed algorithm can accelerate convergence, reduce the generation of evolutionary circuits, and increase the success rate of evolution.