DisCoRaP

Distributed Coordination of Socio-Technical Networks: Toward Resilience and Performance

Year: 2025
Project leader: Mohamed Maghenem
Laboratory: GIPSA-Lab

 

Scientific and/or technological background:

The main objective of this project is to examine resilience through the analysis and design of innovative and efficient decision-making algorithms based on feedback. These algorithms are designed to address coordination tasks in certain classes of interconnected systems. The coordination tasks under consideration are formulated as the collaborative reconstruction by the network of the minimum of a global objective function. Applications include:

  • sustainable energy supply for a community through the coordination of wind turbines, solar power systems, and batteries;
  • reducing congestion by coordinating a group of autonomous vehicles;
  • optimizing comfort and energy efficiency by regulating temperature or air quality in buildings;
  • and training the parameters of machine learning-enhanced applications while respecting user privacy (by exchanging users’ local estimates rather than their actual data).

The DisCoRaP project focuses on scenarios where potential interactions between nodes are governed by a communication graph. The innovative aspect of this project lies in the fact that it allows nodes to leave or join the network, which can occur as a result of packet loss or link failures. Nodes can also be voluntarily removed (in response to the detection of an attack) or intentionally added to strengthen the network in the face of global disruptions. In such dynamic and adaptive networks, we study adjusted objectives that remain close to the nominal objectives according to specific metrics. This ensures resilience in classes of interconnected systems known as open networks.


Project objectives and positioning:

The DisCoRaP project aims to advance the study of distributed optimization in the presence of disturbances, failures, and environmental changes. These are critical control challenges driven by contemporary socio-technical issues.
To achieve this, the DisCoRaP project will integrate three distinct areas that will complement and enrich one another:

Adaptive Networks: Classes of linear, nonlinear, and hybrid networks have been studied in the literature, primarily to achieve synchronization. The interconnection protocols used often require tuning based on global information, such as graph connectivity metrics. The design of fully distributed protocols using adaptive techniques remains an active area of research.

Distributed optimization: Much research has focused on the interconnection of dynamic nodes to collectively reconstruct the minimum of an objective function. This research has addressed both continuous-time and discrete-time contexts, considering both constrained and unconstrained problems, under various assumptions regarding the objective function and the interaction graph. However, to the best of our knowledge, aspects of resilience—particularly through the addition or removal of agents—have not yet been explored.

Open networks: In this context, nodes can potentially join or leave the network, making the system’s size time-dependent. This type of system, also known as a living system, is a relatively new field, and many related issues are still being investigated.

Some recent work, including studies conducted by members of this consortium, has begun to establish links between the first two areas. These studies propose theoretical frameworks that reinterpret certain distributed optimization algorithms using new concepts drawn from the theories of adaptive and hybrid systems. However, only simple optimization problems have been considered so far. Furthermore, resilience to failures and the injection of false data—through the addition and removal of nodes—remains to be rigorously studied within a framework that guarantees robust results.


Scientific publications:

 

Published on April 24, 2025

Updated on May 13, 2026