This paper focuses on the motion control for machines used for the production of products that inherently consist of equal features placed in a repetitive pattern. In many cases the repetitiveness of these structures is prone to imperfections, for example due to thermal expansion, such that the distance between successive features deviates. As a consequence the metric positions of the features of such near-repetitive structures are unknown a priori such that setpoints cannot be created a priori. The considered motion task in this paper is to position a tool relative to the features of a near-repetitive structure with an accuracy of View the MathML source. Instead of metric positions novel two-dimensional feature-based positions will be used that are obtained from a camera capturing images at 1 kHz for feedback, resulting in a direct visual servoing control approach. The robustness with respect to imperfections in the repetitiveness is investigated and the design is validated on an experimental setup.