Adaptive AI Model Partitioning over 5G Networks

Mobile devices increasingly rely on deep neural networks (DNNs) for complex inference tasks, but running entire models locally drains the device battery quickly. Offloading computation entirely to cloud or edge servers reduces processing load at devices but poses privacy risks and can incur high network bandwidth consumption and long delays. Split computing (SC) mitigates these challenges by partitioning DNNs between user equipment (UE) and edge servers. However, 5G wireless channels are time-varying and a fixed splitting scheme can lead to sub-optimal solutions. This paper addresses the limitations of fixed model partitioning in privacy-focused image processing and explores trade-offs in key performance metrics, including end-to-end (E2E) latency, energy consumption, and privacy, by developing an adaptive ML partitioning scheme based on real-time AI-powered throughput estimation. Evaluation in multiple scenarios demonstrates significant performance gains of our scheme.