Neural network based model reference adaptive control structure for a flexible joint with hard nonlinearities

This paper proposes a control strategy based on artificial neural networks (ANN) for a positioning system with a flexible transmission element, taking into account Coulomb friction for both motor and load, and using a variable learning rate for adaptation to parameter changes and to accelerate convergence. The inverse model of this system is unrealizable. The control structure consists of a feedforward ANN that approximates the inverse of the model, an ANN feedback control law, a reference model and the adaptation process of the ANNs with variable learning rate. In this structure, the learning rate of the feedback ANN is sensitive to load inertia variations. The contribution of this paper is to resolve this weakness by proposing a supervisor that adapts the neural networks learning rate. Simulation results highlight the performance of the controller to compensate the nonlinear friction terms, in particular Coulomb friction, and flexibility, and its robustness to the load and drive motor inertia parameter changes. Internal stability, a potential problem with such a system, is also verified.