We submitted a paper titled: “Noise-Induced Limitations to the Scalability of Distributed Integral Control” to the 2018 MTNS in Hong Kong. You can find the preprint here.
This paper nuances our previous results which have shown that distributed integral control may fundamentally improve performance of large-scale networked systems, compared to standard distributed proportional control. In fact, if the controller’s measurements are noisy, performance may scale very badly to large networks. We demonstrate that the requirement for inter-nodal controller alignment will be so large, that centralized integral control will in principle be necessary in large and sparse networks.