A Lyapunov-Guided Task Offloading Approach for Backscatter Communication Assisted Edge Computing

Huaming Wu, Junhui Du, Huijun Tang, Pengfei Jiao, Ruidong Li

Abstract: Optimizing multi-user Mobile Edge Computing (MEC) networks, particularly those with time-varying wireless channels and Backscatter communication (BackCom) models, presents a significant challenge. To address this, we propose an online offloading algorithm that maximizes system data throughput while ensuring stable long-term average energy consumption. This optimization problem is formulated as a multi-stage stochastic Mixed Integer Non-Linear Programming (MINLP) problem, where both binary offloading decisions and resource allocation across multiple time slots are jointly optimized. To handle the complexity introduced by coupled decision-making at different time slots, we introduce LyDRL, a novel approach that combines Lyapunov optimization with Deep Reinforcement Learning (DRL). LyDRL is further enhanced with a threshold quantization method, which significantly reduces computational time and is well-suited for real-time implementation, particularly in environments where channel fading is rapid and unpredictable. Simulation results show that LyDRL reduces runtime by nearly 50% compared to state-of-the-art approaches, confirming its effectiveness and efficiency in dynamic MEC networks.

External IDs:dblp:conf/infocom/WuDTJL25