Liste des projets

Collective Consensus via Multi-scale Feedback Systems

Encadrants

  • Ada Diaconescu, Louisa Jabe Di Felice
  • Emails: ada.diaconescu@telecom-paris.fr
  • Bureaux: 4D49

Nombre d'étudiant par instance du projet:

  • Minimum: 2
  • Maximum: 5

Nombre d'instances du projet :

1

Sigles des UE couvertes et/ou Mots-clés :

Imperative programming : procedural or Object-Oriented (Java) ; graphical user interfaces ; git

Logo/Favicon

project image

Description du projet :

This project aims to study how different architectures of multi-scale feedback systems impact the ability of multiple agents to reach consensus ; as well as to analyze the qualitative properties of the consensus process – e.g. performance, resilience, scalability. This is part of a larger research project that studies how various architectures of decentralized social organizations impact the efficacy and efficiency of their knowledge acquisition and decision-making processes (e.g. centralized, hierarchic or peer-to-peer organizations, with various degrees of top-down authority and control).

The study will rely on simulations of an agent-based model (to be developed – e.g. using NetLogo or Processing). In brief, a number of agents start with individuals opinions (e.g. random values) and coordinate via a multi-scale structure (i.e. hierarchy) so as to converge to a common opinion (e.g. average value). The initial version will enable agents to converge to a shared opinion (average, max, median, etc). A second version will take into account constraints on the range of opinions that each agent can take ; as well as preferences that agents have within those ranges. Further versions will also include : preferences of higher-levels agents (i.e. placed higher in the hierarchy) ; overall system goals ; and/or various configuration parameters to resolve conflicts between system goals and individual agent constraints and preferences.

Inspiration can be drawn from the literature on pertinent topics (e.g. hierarchical decision-making and consensus-formation, multi-scale feedback systems, decentralized social organizations).

Objectifs du projet :

The project’s objectives include :

  • Developing an initial agent-based model that can reach consensus ; and studying the qualitative properties of various system topologies (e.g. number of agents, number of hierarchical levels, number of children per parent node)
  • Developing the second, more sophisticated model, including agent constraints and preferences ; and evaluating it (as above)
  • Developing further model extensions, as in the project descriptions, or as discussed with the uspervisor ; and evaluating them (as above).
  • Drawing general conclusions about the impact of multi-scale system architecture on the functional and qualitative properties of consensus-reaching multi-agent systems.

Références bibliographiques:

  • Ada Diaconescu, Louisa Jane Di Felice, and Patricia Mellodge, "Multi-scale feedbacks for large-scale coordination in self-* systems", 13th IEEE International Conference on Self-Adaptive and Self-Organizing Systems (SASO 2019), Umea, Sweden, 16-20 June, 2019 (preview) – 6 pages
  • Ada Diaconescu, Louisa Jane Di Felice, and Patricia Mellodge, Exogenous Coordination in Multi-Scale Systems: How Information Flows and Timing Affect System Properties, Future Generation Computer Systems, Vol 114, January 2021, pp 403-426, (preview) – 23 pages
  • more to come, depending on the focus of each student team