Swarm-Based Federated Learning for Scalable Optimization of Edge and Cloud Resource Allocation

The fast development of edge and cloud ecosystems has created a need to work out how to use and protect resources properly. Optimization methods that concentrate key functions on a single place usually have latency, privacy and scalability problems. The paper presents a new Swarm-Based Federated Learning (SBFL) approach that connects swarm algorithms with federated learning to help with resource allocation in various edge-cloud settings. According to the model, edge nodes team up to train resource estimators locally, without showing their data and a Particle Swarm Optimization (PSO) or Ant Colony Optimization (ACO) algorithm continuously controls the global model aggregation and parameter updates. With federated learning combined with swarm intelligence, tasks in the cloud are allocated efficiently, scaling occurs instantly and all computing, storage and bandwidth resources are used well. Several experiments carried out in edge-cloud environments confirm that SBFL is better than baseline approaches in terms of how fast it converges, how much resources it needs, how well it scales and its overall energy cost. It suggests a system that is flexible and intelligent which can play a huge role in IoT, 5G/6G networks and real-time artificial intelligence in the future.