GTG-CoL: A New Decentralized Federated Learning Based on Consensus for Dynamic Networks

Miguel Rebollo, Jaime Andres Rincon, Luis Hernández, Francisco Enguix Andrés, Carlos Carrascosa

Abstract: One of the main lines of research in distributed learning in the last years is the one related to Federated Learning (FL). In this work, a decentralized Federated Learning algorithm based on consensus (CoL) is applied to Wireless Ad-hoc Networks (WANET), where the agents communicate with other agents to share their learning model as they are available to the range of the wireless connection. When deploying a set of agents is very important to study previous to the deployment if all the agents in the WANET will be reachable. The paper proposes to study it by generating a simulation close to the real world using a framework that allows the easy development and modification of simulations based on Unity and SPADE agents. A fruit orchard with autonomous tractors is presented as a case study.