SOPC-Based Architecture for Discrete Particle Swarm Optimization

Abstract

The complexity of modern chips is rising and fundamental changes in system design are necessary. System- on-a-Programmable-Chip (SOPC) concept is bringing a major revolution in the design of integrated circuits due to the flexibility it provides and the complexity it caters to. Particle Swarm Optimization (PSO) is a powerful function optimizer that is successfully used to solve problems in numerous fields. The main downside of PSO is that it has significant computation time because of sequential execution of software implementations. In this paper, an SOPC-based PSO framework is proposed. By implementing a hardware/software co-design of PSO, most of the computations can simultaneously be performed using hardware to reduce the computation time, while keeping the flexibility of software. The results indicate a speed-up of up to 100 times in the elapsed computation time.