A Profile-Based Method for Hardware/Software Co-design Applied in Evolutionary Robotics Using Reconfigurable Computing

Abstract

Evolutionary algorithms are a very common technique on computational intelligence field. Some algorithms need a huge amount of memory, making them not so trivial to apply on embedded systems. In this work a profile-based approach is proposed and applied in an evolutionary algorithm with some characteristics that allow its use on embedded systems: the micro-GA. The main goal is to improve the execution time compared to other software algorithms obtaining also an acceptable design time. The results presents comparisons between implementations and discussions about soft-processor features influence on this type of algorithm.