Precise position control of tubular linear motors with neural networks and composite learning

This paper examines an adaptive control scheme for tubular linear motors with micro-metric positioning tolerances. Uncertainties such as friction and other electro-magnetic phenomena are approximated with a radial basis function neural network, which is trained online using a learning law based on Lyapunov design. Differently from related literature, the approximator is trained using a composite adaptation law combining the tracking error and the model prediction error. Stability analysis and bounds for both errors are established, and an extensive experimental investigation is performed to assess the practical advantages of the proposed scheme.