Topology Matters: Evaluating Multi-Agent Organizations for Resilient Flood Detection

Flood detection networks often fail when fixed infrastructure such as gateways or cellular towers is damaged. Multi-agent systems (MAS) operating over ad hoc peer-to-peer networks offer a resilient alternative by enabling sensors to self-organize, reroute data, and forward alerts cooperatively. This work simulates and compares two MAS organizational models for IoT-based flood monitoring: Flat MAS and Federated MAS. Flat MAS rely on peer-to-peer communication, allowing sensors to exchange acknowledgments and reroute messages through neighbors during connectivity loss. Federated MAS use selected gateways to coordinate clusters and maintain inter-gateway failover links, improving system continuity. Results show that Flat MAS respond faster under partial failure, while Federated MAS achieve higher delivery success and recover more effectively from gateway outages.