论文简介 |
This article studies convergence characteristics of first- and second-order P-type (proportional-type) iterative
learning control laws for a class of partially known linear time-invariant systems with a direct feed-through term. In the
study, the tracking error is measured in the sense of Lebesgue-p norm, based on which the concepts of the Qp factor and
the convergence speed are specified. For the first-order updating law, the monotone convergence is achieved by means of
the generalised Young inequality of convolution integral. Moreover, for the second-order updating rule, the convergence is
also verified and its speed is compared with the first-order law. Through analysis, it is quantitatively noted that both the
system dynamics and the learning gains affect the convergence. It is also observed that the iterative learning process with
the second-order law can be Qp-faster, Qp-equivalent or Qp-slower than the system with the first-order rule, depending on the
selected learning gains. Numerical simulation manifests the validity and the effectiveness. |