Federated learning for edge computing is an emerging research area that focuses on enabling distributed, privacy-preserving model training directly on edge devices, such as smartphones, IoT sensors, and autonomous systems, without transmitting raw data to centralized servers. This paradigm addresses challenges in data privacy, communication efficiency, limited computation, and heterogeneous data distributions across devices. Research explores optimization techniques for resource-constrained environments, personalized federated learning to handle non-i.i.d. data, model compression, and communication reduction strategies, as well as privacy-preserving methods like differential privacy and secure aggregation. Integration with deep learning architectures, including CNNs, RNNs, and transformers, enables real-time analytics and inference for computer vision, natural language processing, and IoT applications. Recent studies also investigate hybrid edge-cloud frameworks, task offloading strategies, and robust federated learning against adversarial attacks, establishing federated learning as a key enabler for intelligent, decentralized, and privacy-aware edge computing systems.