Research on Distributed AI Algorithms Based on Federated Learning in Edge Computing Environments

With the rapid development of big data and artificial intelligence technologies, edge computing has become an important paradigm for supporting distributed AI applications. However, traditional centralized machine learning frameworks face challenges such as privacy leakage, high communication overhead, and limited scalability in edge scenarios. This paper investigates the design and optimization of federated learning algorithms for distributed AI in edge computing environments. First, the theoretical foundations and key technologies of federated learning and edge computing are discussed. Then, key challenges including non-IID data, network latency, and node heterogeneity are analyzed. To address these issues, communication-efficient strategies such as model compression and local update frequency control are proposed. Simulation-based experiments demonstrate that the proposed methods can significantly reduce communication costs while maintaining high model accuracy. Finally, a smart traffic management case study illustrates the practical applicability of the approach. This research provides a reference for developing privacy-preserving, efficient, and robust distributed AI systems in future edge computing applications.