Evolving hardware with genetic algorithms

Genetic techniques are applied to the problem of electronic circuit design, with an emphasis on VLSI circuits. The goal is to have a tool which has the performance and flexibility to attack a wide range of problems. A genetic algorithm is used to design a circuit specified by the desired input /output characteristics. A software system is implemented to synthesize and optimize circuits using an asynchronous parallel genetic algorithm. The software is designed with object-oriented constructs in order to maintain scalability and provide for future enhancements. The system is executed on a heterogeneous network of workstations ranging from Sun Sparc Ultras to HP multiprocessors. Testing of this software is done with examples of both digital and analog CMOS VLSI circuits. Performance is measured in both the quality of the solutions and in the time it took to evolve them