Neuro-fuzzy state modeling of flexible robotic arm employing dynamically varying cognitive and social component based PSO

The present paper proposes the development of a neuro-fuzzy state-space model for flexible robotic arm on the basis of real sensor data acquired. The training problem of the neuro-fuzzy architecture has been configured as a highly multidimensional stochastic global optimization problem and improved variants of particle swarm optimization (PSO) techniques have been successfully implemented for it. The effects of dynamically varying the “cognitive” and the “social” components of the improved PSOs on the training performance have been studied in detail. The practical utility of such a model development procedure is aptly demonstrated by employing the best trained model to design a stable fuzzy state controller and implementing it in real life for the same flexible robotic arm.