Iterative learning control of perspective dynamic systems

This paper introduces the problem of iterative learning control (ILC) for perspective dynamic systems (PDS), referred to as ILC-PDS, where the task is to track a 3-D desired trajectory from information observed on the image plane of an imaging system, such as a camera. Unlike many ILC implementations in motion control applications, which use feedback from encoders, we assume measurements from the image plane are used as feedback to the ILC system. We focus our analysis on 2-D motion with a 1-D perspective measurement. It is shown analytically and via simulation that under the standard ILC assumption of identical initial conditions from trial-to-trial, tracking in the 2-D space can be achieved using the 1-D PDS image feedback.