🪄 Workshop on Analysis of Network Dynamics 🪄

About
WAND is a half-day workshop dedicated to the analysis of network dynamics, namely the study of dynamics both on and of networks. Network dynamics refer to a set of simple protocols, built upon local and elementary rules, for solving fundamental computational problems.
The purpose of the workshop is to bring together researchers interested in studying network dynamics and to facilitate discussions about the strategies and techniques utilized for their analysis and design.

The workshop will take place in Rome on November 13th, 2026 in the afternoon (13.45 - 17.05), it is co-located with the conference DISC 2026 and aims at the general DISC audience. 

Program
The workshop will take place at Aula DIET at Sapienza Engineering Faculty, at the Chiostro di San Pietro in Vincoli in Rome.
Address: Via Eudossiana, 18, 00184, Roma RM, Italy. Here is a link to the map.

• 13:45 - 14:40 Avery Miller : Packet Forwarding Strategies Under Bounded Adversarial Injections

• 14:45 - 15:40 Niccolò D'Archivio: Minimalist Mechanisms for Distributed Plurality Consensus

• Coffee break

• 16:00 - 16:55 Christian Scheideler: Getting Closer to the Programmable Matter Vision - Extensions of the Amoebot Model

Avery Miller (University of Manitoba)

Packet Forwarding Strategies Under Bounded Adversarial Injections

Consider a network of n nodes, each possessing a fixed-size "buffer" that can store a finite number of packets. Time proceeds in discrete rounds, and in each round, new packets are injected into the network: each injected packet starts in the buffer of some node and has a specified route and "destination" node. The injections are arbitrary but bounded with respect to two parameters ρ and σ, where ρ roughly corresponds to the average rate of injection over time, and σ roughly corresponds to an additional burst of injected packets (more formally, over any interval of t consecutive rounds, the number of injected packets whose route requires any particular edge is bounded above by ρ·t + σ).

At the start of each round, each node chooses, for each of its incident edges, whether or not to forward one packet from its buffer along the packet's intended route. Each node makes its decision according to a simple "forwarding rule". If a packet eventually arrives at its destination, the packet is considered "delivered" and is removed from the network. However, any packet that is forwarded to or injected at a node with a full buffer is considered "dropped" and lost forever.

A recent line of research has considered various forwarding rules, and for each rule, has analyzed the buffer size needed at each node to ensure that every packet is delivered to its destination (i.e., no packets are ever dropped) under worst-case adversarial injections. The goal of this talk is to provide a survey of these results, along with related impossibility results and separations between centralized vs. local forwarding rules, and to discuss directions for future work.

Niccolò D'Archivio (INRIA, Sophia Antipolis)

Minimalist Mechanisms for Distributed Plurality Consensus

Plurality consensus is a fundamental distributed problem in which a population of agents, each initially holding one of several opinions, aims to agree on the initially most frequent one. How limited can agents’ capabilities be while still allowing them to achieve plurality consensus?

A standard approach is the h-Majority protocol: each agent samples h random neighbors and adopts the most frequent opinion in the sample. Although simple to describe, this protocol asks agents to compare opinion frequencies. Can an even simpler mechanism suffice?

In this talk, we explore this question through DéjàVu, a protocol in which an agent queries random neighbors until it encounters an opinion for the second time, then adopts that opinion. The protocol replaces frequency comparisons with repetition detection. We show that repetition detection suffices to solve plurality consensus without increasing communication cost.

Christian Scheideler (Universität Paderborn)

Getting Closer to the Programmable Matter Vision - Extensions of the Amoebot Model

The amoebot model is one of the models that have been proposed for the algorithmic study of programmable matter. Its basic form was shown to be useful for many problems relevant for programmable matter like shape formation and coating. However, it also has a number of deficits like slow information propagation and slow shape transformation due to the restriction to local communication and movements. Several extensions already improved the situation like the ability to set up circuits and allowing joint expansions and contractions. Still, various issues cannot be addressed well so far, including fault-tolerance, energy dissemination, the 3D case, and changing the mechanical properties of the structure. In my presentation, I will mostly focus on appropriate extensions for the latter two issues. A natural approach to extend the amoebot model to the 3D case is to use the face-centered cubic (FCC) lattice. However, designing algorithms for that setting is challenging. Therefore, we are currently exploring other directions in which the amoebots are able to flexibly change their bond lengths and the spatial orientation of the bonds. As I will demonstrate, this allows the amoebots to transform 2D structures into 3D structures and opens up the exciting world of changing the mechanical properties of the structure, from rigid to different modes of flexibility.

Workshop Chair
Robin Vacus (CNRS, LIP6, Sorbonne Université)

Steering Committee
Emilio Cruciani (European University of Rome)
Francesco d'Amore (Gran Sasso Science Institute)
Isabella Ziccardi (IRIF, Paris)

Other Editions
For informations about the other editions of WAND, visit the WAND website.