Trusted Task Offloading in Vehicular Edge Computing Networks: A Reinforcement Learning Based Solution

Mobile edge computing (MEC) has emerged as a promising approach to address the time-sensitive requirements of mobile Internet of Vehicles (IoVs) systems. Unfortunately, the current deployment density of roadside units (RSUs) is relatively sparse, and the direct V2I communication coverage is limited, making it impossible to meet the communication and computing requirements of all vehicles. There is an urgent need for V2V communication to assist V2I communication, which can achieve a wider coverage of RSUs, a diversified selection of task processing locations, and even load balancing between RSUs. However, V2V communication also faces a series of challenges. On the one hand, due to the sparsity, time-varying, and high-speed mobility of vehicle nodes in IoVs, the selection of collaborative communication paths becomes more difficult. On the other hand, there are inevitably malicious vehicles in IoVs, and how to achieve efficient task processing while ensuring privacy and driving safety is also a problem worth studying. Existing research generally optimized the delay of direct V2I task offloading, ignoring the necessity of V2V-assisted communication and the presence of malicious communication nodes. To address the above challenges, we present a vehicular edge computing network structure with multiple communication modes, including V2V, V2I, etc, and use a recommended trust model to analyze the trust degree between the nodes in IoVs. Then, we discuss the issue of trusted task offloading for IoVs and propose a Deep Deterministic Policy Gradient (DDPG) scheme. The numerical results indicate that our proposed strategy outperforms current methods in terms of task offload latency and credibility.