Game-theoretic joint optimization of spectrum sharing and task offloading for digital twin-assisted IoV

Abstract

The Internet of Vehicles (IoV) enables advanced applications such as autonomous driving, but it also demands significantly more computing and communication resources. Current research on task offloading struggles to address key challenges, including limited spectrum resources, real-time decision-making in highly dynamic topologies, and interference coordination. To overcome these issues, this paper first constructs a joint system model that integrates spectrum sharing with task offloading, clarifying multi-dimensional optimization objectives and constraints. We then design a deep learning-based channel communication time prediction mechanism, which uses historical data and a neural network to accurately predict channel availability. Finally, we propose a distributed task offloading algorithm based on non-cooperative game theory. In this approach, task vehicles act as independent game players that autonomously optimize their multi-dimensional strategies using local information. Through limited information exchange, the system converges to a Nash equilibrium. Simulation results show that the proposed method outperforms existing baseline approaches in convergence speed, latency reduction, and prediction accuracy, demonstrating its suitability for highly dynamic vehicular networks.