Full-Day Workshop at the 65th IEEE Conference on Decision and Control (CDC 2026)
December 2026 | Honolulu, Hawaii
Collaborative control is emerging as a compelling new direction within control engineering, bringing ideas long explored in machine learning—such as federated learning, multi-task learning, and shared representation learning—into the analysis and control of dynamical systems. As modern control applications scale to unprecedented levels of complexity—spanning large fleets of interconnected devices, geographically distributed infrastructures, and data-rich cyber-physical systems, the traditional paradigm of centralized learning and control becomes impractical due to data volume, communication constraints, latency, and privacy considerations. Collaboration across agents, tasks, or datasets is therefore not merely advantageous but essential.
However, this shift introduces a fundamental performance tradeoff: while increasing the number of agents can improve statistical efficiency and coverage, heterogeneity across agents and tasks can induce bias and degrade control performance, necessitating careful coordination and structure sharing. This workshop aims to position collaborative control as a principled framework that navigates these tradeoffs, touching upon the themes of stability, sample-complexity, robustness, information constraints, and the fundamental implications of sharing data generated by distinct (heterogeneous) systems. By unifying insights from control theory and contemporary machine learning, the goal is to chart a path toward scalable, reliable control methodologies suited to the next generation of complex systems.
Assistant Professor
North Carolina State University
amitra2@ncsu.edu
Associate Professor
Columbia University
james.anderson@columbia.edu