Game Theory Based Opportunistic Computation Offloading in Cloud-Enabled IoV

Minghui Liwang, Jiexiang Wang, Zhibin Gao, Xiaojiang Du, Mohsen GuizaniIEEE Access 2019 · DOI

Introduction

As the Internet of Vehicles (IoV) evolves, modern automobiles are transforming into mobile computing hubs. Advanced applications—such as Augmented Reality (AR) dashboards, autonomous navigation, and real-time safety analytics—demand significant computational resources. However, individual vehicles often suffer from a "resource crunch," where on-board hardware is insufficient for peak demands, and fixed infrastructure like Road Side Units (RSUs) may suffer from latency or coverage gaps. This paper proposes a game-theoretic framework for "opportunistic computation offloading," allowing vehicles with heavy tasks to "hitchhike" on the idle CPU cycles of nearby peers.

System Architecture and Problem Formulation

The researchers envision a 5G cloud-enabled IoV environment where vehicles are categorized into two roles: Task Vehicular Terminals (Task VTs), which require external processing power, and Service Vehicular Terminals (Service VTs), which offer their idle resources for a fee.

System architecture of the 5G cloud-enabled IoV, illustrating V2V and V2I communication links.

The challenge is inherently dynamic. High mobility means communication windows are brief, and offloading decisions must account for relative speeds and contact durations. Beyond the technical constraints, there is an economic hurdle: Service VTs require incentives (profits) to expend their battery and CPU resources for others.

The Stackelberg Game Framework

To address the conflicting interests of Task and Service VTs, the authors employ a two-player Stackelberg game. In this model, the Service VT acts as the Leader, setting a unit price ($p$) per bit of data processed. The Task VT acts as the Follower, observing the price and determining the optimal offloading rate ($\lambda$)—the fraction of the task to be processed remotely versus locally.

The interaction model between a Task VT and a Service VT, defining the offloading rate and pricing strategy.

The study explores two information scenarios:

  1. Mobile Offloading Complete Information Game (MOCIG): The Task VT knows the Service VT's internal costs, allowing for a direct mathematical optimization of resource assignment.
  2. Mobile Offloading Incomplete Information Game (MOIIG): A more realistic scenario where the Task VT only knows the quoted price. Here, the authors derive a Stackelberg Equilibrium (SE) to find a stable point where both parties maximize their respective utilities.

Simulation and Key Insights

Using Monte-Carlo simulations in MATLAB, the authors modeled a four-lane highway with varying densities of Service VTs. Several key insights emerged:

  • Parallel Processing Efficiency: The model demonstrates that offloading is most effective when tasks are split. By processing data locally and remotely in parallel, the total duration is determined by the slower of the two processes, significantly outperforming local-only computing.
  • Market Dynamics: As the density of Service VTs increases, competition naturally drives down the price of computing services, benefiting the Task VTs.
  • Equilibrium Stability: The game-theoretic approach ensures that offloading is sustainable. Unlike simple heuristics like First-Come-First-Serve (FCFS), the Stackelberg Equilibrium guarantees that Service VTs remain profitable, which is essential for a functional real-world ecosystem.

Conclusion and Future Directions

The proposed framework successfully reduces task latency while maintaining economic fairness. However, the authors note that future work should address more complex real-world variables, such as non-linear vehicle motion, signal interference in dense traffic, and multi-player cooperative games where tasks are distributed across multiple neighboring vehicles simultaneously.

Reference: Liwang, M., Wang, J., Gao, Z., Du, X., & Guizani, M. (2019). Game Theory Based Opportunistic Computation Offloading in Cloud-Enabled IoV. IEEE Access, 7, 32551-32561.

Summary updated

BibTeX
@article{liwang2019game,
  author = {Minghui Liwang and Jiexiang Wang and Zhibin Gao and Xiaojiang Du and Mohsen Guizani},
  title = {{Game Theory Based Opportunistic Computation Offloading in Cloud-Enabled IoV}},
  journal = {IEEE Access},
  year = {2019},
  doi = {10.1109/access.2019.2897617},
}
Figure