Fundamental bounds in distributed decision-making

Distributed decision making systems arise in many engineering problems.

Due to the nonclassical information structure of these problems, optimal decision and communication strategies are considered to be notoriously difficult and a general design framework is missing.

Promising initial results show that the in information theory recently appeared concept of coordination coding can be used to derive fundamental bounds for distributed decision systems.

This project will extend and broaden those initial results to deepen the understanding, develop and generalize proof techniques to set the basis for a new analysis framework.The project is divided in four work packages.

It is expected that this new approach will provide new fundamental insights on how to design and trade-off optimally the communication and control in vector-valued distributed decision-making problems. In particular, the project will derive new bounds and perhaps the optimal solutions of the Witsenhausen counterexample.

The research results will deepen the relation between information theory and distributed stochastic control and enable the transfer of advanced results in both directions.The PI will collaborate in this 4-year project with a new PhD student, M. Le Treust (CNRS), with whom the initial results have been obtained, and S.

Yüksel (Queens Univ.). We aim for research results of highest international quality published at most relevant conferences and journals.