Agentic Trust Coordination for Federated Learning in Sustainable Industrial Networks

Summary

A paper (arXiv:2603.25334) by Paul Shepherd, Tasos Dagiuklas, Bugra Alkan, and Jonathan Rodriguez proposes a lightweight agentic trust coordination approach for federated learning (FL) in sustainable and resilient industrial networks. Industrial systems increasingly integrate sensing, communication, and computation across heterogeneous, resource-constrained devices. FL enables collaborative model training in these environments, but its reliability suffers from inconsistent client behavior, noisy sensing conditions, and faulty or adversarial updates. The proposed agentic trust control layer runs as a server-side control loop: it observes trust-related and system-level signals, interprets their evolution over time, and applies targeted trust adjustments when instability is detected. The design aims to improve robustness while remaining lightweight enough for resource-constrained industrial deployments, addressing trust management gaps that conventional FL aggregation schemes leave open.