Identification of multiple mode models via Distributed Particle Swarm Optimization

摘要: Abstract This paper considers the identification of multiple-mode systems, and introduces a new method to estimate subsystem parameters piece-wise affine systems. First, notion linear regression model way reduce its problem an optimization one are introduced. Second, since introduced is inherently ill-conditioned nonconvex, technique named distributed PSO (particle swarm optimization) developed avoid being trapped in suboptimal solutions. The proposed scheme can handle systems without any prior knowledge about their mode transitions has no difficulty large number data samples, which distinguished feature method. Finally, experiment with set I/O from DC motor system given demonstrate effectiveness evaluate performance technique.