Adaptive high precision position control for a flexible joint with friction and parameter uncertainties using neural networks

Dynamic position-control of a flexible joint is proposed by applying adaptive control and Artificial Neural Networks theory. A flexible joint is modeled, including Coulomb and static frictions and the model is represented as an ANN. The control strategy is based on a dual loop strategy. An outer load state feedback is used to compute desired load torque and motor state. An inner motor state feedback loop is used to control the motor. Both loops use feedforward compensation of friction. The controllers are represented as ANN, the system parameters being the weights of the output layer. Parameter identification is achieved using recursive least squared algorithm. Simulation results show that the proposed controller can suppress vibrations.