Energy-Efficient Dynamic Retransmission Timeouts with Enhanced Stability for Constrained Application Protocol-Based Internet of Things Networks via Edge Intelligence-Assisted Cross-Layer Congestion Control

The co-existence of event-driven critical traffic and time-driven periodic traffic inevitably exacerbates cross-layer network congestion in resource-constrained edge environments. Although hybrid protocol architectures integrating the Constrained Application Protocol (CoAP) at the edge and Quick UDP Internet Connections (QUIC) in the core network have emerged, existing gateways manage these protocols independently, failing to provide an organic congestion control mechanism. To overcome these limitations, this paper proposes an ultra-lightweight Edge Intelligence (EI)-assisted end-to-end (E2E) CoAP-QUIC cross-layer congestion control framework powered by Proximal Policy Optimization (PPO). The proposed scheme introduces an ultra-lightweight traffic classification mechanism that instantly distinguishes traffic classes by parsing the existing two-bit type field in the CoAP header, effectively bypassing the payload inspection overhead. On the basis of this, the PPO agent shapes its reward function in real time, actively shifting optimization weights between delay reduction and throughput optimization. This dual-action control directly mitigates congestion by dynamically tuning the QUIC congestion window and CoAP back-off timers to prevent edge buffer saturation. Extensive simulations using Network Simulator 3 (NS-3) demonstrate that the proposed framework significantly outperforms state-of-the-art baselines, bounding end-to-end latency for critical traffic under 100 ms and improving overall energy efficiency by 21.5% while achieving a 98.2% packet delivery ratio.