This chapter explores two critical aspects of federated learning: communication optimization and fairness perception. It introduces communication-efficient techniques, such as knowledge distillation, parameter compression, edge computing, and client clustering, to address resource constraints and enhance model convergence. For fairness, methods like reweighting parameter aggregation, regularization, multi-task learning, and differential privacy are discussed to mitigate biases and promote equity. The chapter highlights how these strategies improve efficiency and fairness in federated learning for IoE scenarios and outlines challenges for future research, including dynamic data heterogeneity and balancing fairness with performance.