A preprint of our article “On Fundamental Limitations of Dynamic Feedback Control in Large-Scale Networks” can now be found on ArXiv. This paper was recently submitted to IEEE Transactions on Automatic Control.
This paper investigates performance limitations of distributed dynamic feedback controllers in large-scale networks. The main question motivating this work is whether memory in the local controllers can compensate for their lack of global sensing, and thereby improve global performance. We show, however, that dynamic feedback with a single memory state in general suffers from the same performance limitations as static feedback.
In order to derive our results, we develop a technical framework for analyzing networked systems in the limit of infinite networks. We are also able to derive interesting stability conditions, which show that certain classes of controllers only can guarantee stability in small networks, and inevitably destabilize the system as the network grows.