Federated Learning-Based Network Slicing in 6G Systems for Enhanced Performance in Autonomous Vehicle Communication

To improve 6G connectivity for autonomous vehicles (AVs), this article suggests a network slicing system based on federated learning. AVs maintain privacy by using federated learning to train models locally on communication data, and then aggregate a global model without exchanging raw data. The system dynamically distributes network slices to satisfy real-time communication requirements, including interactions between vehicles and infrastructure V2I and V2V. Numerous simulations show significant gains, such as a $\mathbf{5. 2 ~ m s}$ latency reduction, 100.5 Mbps throughput increase, and 99.7% dependability improvement. With a score of 94.2%, privacy is protected while resource consumption is minimized to 85.4%. These findings show that, in comparison to more conventional approaches, the suggested architecture provides better performance and data safety while meeting the demanding specifications of 6G-enabled AV communication.