Advances in faster and more reliable networking and computing infrastructure, as well as their wider availability, have been driving cyber-physical systems to become increasingly networked. The control systems community has been leading these efforts, with fundamental results in topics such as distributed optimization and control, networked control, and multi-agent systems being developed. Concurrently, there is a revolution in machine learning techniques that allow extraction of useful information and models from data. Given that it is difficult to obtain accurate models for large scale networked cyber-physical systems, over the last years it has been recognized that networked cyber-physical systems will become increasingly data-driven. A natural question then is whether networked cyber-physical system control can exploit developments in machine learning algorithms, and concomitantly, drive advances in machine learning. The purpose of this workshop is to explore the fundamental research questions that arise at the intersection of these areas.

The workshop will take place virtually on 12 December 2021 at 13:00-17:00 UTC (8:00 - 12:00 EST) during the 2021 60th IEEE Conference on Decision and Control.

Registration for the workshop can be made through this link at the IEEE Conference on Decision and Control website.

Tamer Başar
University of Illinois Urbana-Champaign
Thinh Doan
Virginia Tech
Hamed Hassani
University of Pennsylvania

Na (Lina) Li
Harvard University
Angelia Nedich
Arizona State University
Sebastian Trimpe
RWTH Aachen University
Time (UTC) Topic (Click on title to open abstract)
13:00 - 13:05 Introduction by organizers
13:05 - 14:35

Session 1: Distributed and Federated Learning - Chair: Vijay Gupta

13:05 - 13:35 Distributed Algorithms for Optimization in Networks - Angelia Nedich

13:35 - 14:05 Achieving Linear Convergence in Federated Learning under Objective and Systems Heterogeneity - Hamed Hassani

14:05 - 14:35 Two-time-scale methods for distributed learning over multi-agent systems under imperfect communication - Thinh Doan

14:35 - 14:50 Break
14:50 - 16:20

Session 2: Learning for Multi-Agent Systems - Chair: Konstantinos Gatsis

14:50 - 15:20 Decentralized Multi-agent Reinforcement Learning with Networked Agents - Tamer Başar

15:20 - 15:50 Real-time Distributed Decision Making in Network Systems - Na Li

15:50 - 16:20 Event-triggered Learning - Sebastian Trimpe

16:20 - 16:30 Break
16:30 - 17:00

Session 3: Roundtable Discussion on Open Challenges and Opportunities

Konstantinos Gatsis
University of Oxford
Vijay Gupta
University of Notre Dame

Should you have any questions, please do not hesitate to contact the organizers.