Discussion on: "Adaptive Tracking for Linear Plants under Fixed Feedback"

The paper by M. Nilsson and B. Egardt [6] presents successful implementations of a 2DOF Adaptive Control technique that is made of a fixed feedback and an adaptive feedforward controller. As they state, their methodology does not assume that the plant is totally unknown and that the Adaptive Controller alone must do all the task. On the contrary, they assume that some basic knowledge is available and divide the plant into two parts. A preliminary LTI (specifically, QFT) design is then performed on that part of the plant that, in spite of its uncertainties, is sufficiently known to allow a basic robust design that ends in a fixed feedback controller. Then, they perform identification of the second, less certain, part and a corresponding control design based on identified parameters. They also apply this technique to two challenging examples from literature. First, we congratulate the authors for the nice presentation and, like them, we also encourage prospective users to try the techniques with their own examples. One must try and get used with these techniques before one can get a good understanding and can be confident enough to use them with real-world systems. Only then can one see how Adaptive Control can add performance that could not be obtained otherwise. As the authors make a good presentation of their 2DOF methodology, we will try to clarify the differences between this approach and the Simple Adaptive Control (SAC) approach that the paper also mentions. The SAC approach was born out of very specific needs, in particular, control of large plant, such as planes, missiles, satellites, large flexible structures, etc, where the order of the plant is not only very large, but also basically unknown. These