Abstract:

The integration of Edge AI into 6G networks marks a technological shift for enabling ultra-low latency, high adaptability, and intelligent selection-making within complex future networks. Such technologies respond in real-time to extreme command challenges—autonomous systems, remote surgeries, industrial automation, etc. Still, there is no existing edge or cloud-based AI solution that meets the extreme adaptability and ultra low latency requirements of 6G. This paper proposes a solution to these challenges using Distributed Intelligent Post-Edge Learning and Inference for 6G Ecosystems (DIPLAI-6G), a new hybrid AI system in which the intelligence is distributed at the Near-Device, Post-Edge, and Core Intelligence levels. It employs a federated learning approach for decentralized training, knowledge distillation for model compression, and dynamic task allocation for Inter-Network Layer Silo LASER (Latency-Aware Orchestration Algorithm) scheduling. The framework enables responsive, privacy-compliant, and optimal decision processes in diverse 6G scenarios. Simulation results demonstrate that DIPLAI-6G significantly reduces end-to-end latency, achieves higher accuracy, and improves reliability in mobile and dynamic network environments when compared to traditional edge and cloud-based alternatives. The proposed approach results in a robust, intelligent, and flexible framework that meets the 6G application demands with tolerable latency requirements.