An extremum seeking approach to sampled-data iterative learning control of continuous-time nonlinear systems

Iterative learning control (ILC) of continuous-time nonlinear plants with periodic sampled-data inputs is considered via an extremum seeking approach. ILC is performed without exploiting knowledge about any plant model, whereby the input signal is constructed recursively so that the corresponding plant output tracks a prescribed reference trajectory as closely as possible on a finite horizon. The ILC is formulated in terms of a non-model-based extremum seeking control problem, to which local optimisation methods such as gradient descent and Newton are applicable. Sufficient conditions on convergence to a neighbourhood of the reference trajectory are given.