Delay-Aware Task Offloading for Heterogeneous VLC-RF-based Vehicular Fog Computing

TL;DR

This work tackles delay-sensitive computation in dense vehicular networks by proposing a heterogeneous VLC-RF vehicular fog computing (VFC) architecture that enables dynamic multi-partition task offloading to idle vehicles. It formulates an average delay minimization problem that couples offloading and computing resource allocation, then transforms the non-convex problem into a second-order cone program using auxiliary variables and a sqrt-based substitution. The core contribution is the residual-based majorization-minimization (RBMM) algorithm, which iteratively solves convex surrogates enhanced by residual terms to approach a KKT point; the method achieves notable reductions in task processing delay and outperforms state-of-the-art baselines. Simulation results demonstrate about 15% average TPD reduction over VLC-only or RF-only systems and illustrate the algorithm’s convergence and robustness to varying bandwidth and computation intensity. The approach holds practical significance for enabling low-latency services in future autonomous, connected transportation systems.

Abstract

Vehicular fog computing (VFC) is a promising paradigm for reducing the computation burden of vehicles, thus supporting delay-sensitive services in next-generation transportation networks. However, traditional VFC schemes rely on radio frequency (RF) communications, which limits their adaptability for dense vehicular environments. In this paper, a heterogeneous visible light communication (VLC)-RF architecture is designed for VFC systems to facilitate efficient task offloading. Specifically, computing tasks are dynamically partitioned and offloaded to idle vehicles via both VLC and RF links, thereby fully exploiting the interference resilience of VLC and the coverage advantage of RF. To minimize the average task processing delay (TPD), an optimization problem of task offloading and computing resource allocation is formulated, and then solved by the developed residual-based majorization-minimization (RBMM) algorithm. Simulation results confirm that the heterogeneous VLC-RF architecture with the proposed algorithm achieves a 15% average TPD reduction compared to VFC systems relying solely on VLC or RF.

Figures (4)

• Figure 1: The model of the heterogeneous VLC-RF-based VFC system.
• Figure 2: Various components of TPD during RBMM algorithm convergence.
• Figure 3: Comparison of multiple task offloading methods under various computing resources.
• Figure 4: Dual impact of subchannel bandwidth and task computation intensity on the heterogeneous VFC system.