Adaptive chaotification of robot manipulators via neural networks with experimental evaluations

Chaotification is a problem that has been studied in recent years. It consists in injecting a chaotic behavior by means of a control scheme to a system, which in natural form does not present it. This paper explores the chaotification (also denoted anticontrol of chaos) of robot manipulators. Adaptive neural networks have the advantage of compensating the dynamics of a system with practically null information about this. By using a Lyapunov-like framework, chaotification of robot manipulators is assured with an adaptive neural network control law. A two layer neural network is used. Adaptation of the output weights are designed. Real-time experiments in a two degrees-of-freedom robot are presented. The new neural network-based controller is compared theoretically and experimentally with respect to a regressor-based controller.