New developments in neurocontrol

Abstract

The complexity of modern day systems is increasing and performance requirements for industrial and military systems are becoming more stringent. With new results based on nonlinear stability theory, neural network (NN) based controllers are now able to provide guaranteed closed-loop stability, performance, and robustness for such complex dynamical systems. This paper presents a family of NN controllers developed for robotic systems, force control, backstepping control of industrial motors, friction compensation, deadzone compensation of actuators, etc. Some high-level NN control architectures are discussed, including reinforcement techniques and optimal design.