SwarmSync is my undergraduate capstone project — and the research behind it was accepted at ISCC 2026: 31st IEEE Symposium on Computers and Communications, one of the flagship IEEE conferences on networking and distributed systems. The work was developed in collaboration with Beatriz Cardoso de Oliveira and advised by Prof. Sarita Bruschi at USP/ICMC.

The Problem

Drone swarms operating in remote or hostile environments — forests, farmlands, disaster zones — can’t rely on stable network connections. Nodes move constantly, links drop, the network fragments. Traditional distributed systems assume some level of connectivity; FANETs (Flying Ad-Hoc Networks) offer none of those guarantees. The challenge was to keep a swarm’s shared state synchronized without a central coordinator, and without assuming messages will ever arrive in order — or at all.

What I Built

I designed and implemented a lightweight data-sharing layer for eventual consistency in UAV swarms, combining three core mechanisms:

• Delta-CRDTs (AWOR-Set) — conflict-free replicated data types that propagate only state deltas rather than full state, keeping bandwidth usage low on resource-constrained nodes
• Push-Gossip dissemination — a decentralized protocol where each node periodically pushes updates to a random subset of neighbors, spreading information across the swarm without any central broker
• Periodic anti-entropy — a background reconciliation process that catches missed updates and heals partitions after connectivity is restored

Together, these guarantee eventual convergence of the swarm’s shared state across all nodes, regardless of message loss, reordering, or temporary network splits.

Implementation & Evaluation

The sync layer was implemented in Go and evaluated in Mininet-WiFi, a network emulator that supports mobile wireless nodes. Experiments ran with up to 64 simulated drones under varying mobility and packet loss conditions, validating convergence behavior and measuring dissemination latency across realistic FANET topologies.

Applications

The system was designed with real-world use cases in mind: environmental monitoring, wildfire detection, and search-and-rescue operations — scenarios where autonomous drone swarms need to share observations reliably even when operating far from any infrastructure.