DSP-based neural-network control system with finite time convergence method

Abstract

Neural network (NN) is a poplar intelligent control scheme for wildly application. It is usually adopt the back propagation as their learning method to reduce the resource on calculation. Furthermore, the terminal steepest descent algorithm (TSDA) has been adopted as the adjustment of network parameters in this study. The TSDA provide finite-time convergence ability to solve slow convergence performance of steepest descent algorithm. The development of control strategies will surround the model-free strategies, and then establish a control system that the digital signal processor is adopted as computation core. Finally, there are comparisons of performance of all strategies that previous mentioned will compared in two different conditions. The simulation and experimentation of proposed strategies performance results will be compared at last. The terminal steepest descent algorithm (TSDA) has been adopted as the adjustment of network parameters in this study. The TSDA provide finite-time convergence ability to solve slow convergence performance of steepest descent algorithm.