Synthesis of adaptive hardware system on field programmable gate arrays.

摘要

Recently, research in Evolutionary Computation has switched focus to applications in Electrical Engineering, leading to the field of study called Evolvable Hardware. The ultimate goal is to develop hardware systems with on-line adaptation capabilities. Applications requiring optimum performance due to changing environments can be integrated with an evolutionary component capable of fine-tuning system parameters during real-time operation. The proposed Adaptive Hardware System consists of an Evolutionary Adaptation System, Configurable Working System, and On-line Monitoring and Diagnostic System. Focus is primarily concentrated on the Evolutionary Adaptation System called the Autonomous Genetic Machine. Synthesis results for Field Programmable Gate Arrays from various manufacturers are studied. Performance measurements relating chromosome length, population size, fitness representation and system clock frequency are investigated. As an extension, the Fitness Evaluation Subsystem is developed. Capable of configuring and evaluating an application using chromosomes in conjunction with training data, this additional subsystem increases the scope of potential applications.