UAV-assisted collaborative sensing task offloading and resource allocation in IoV

Unmanned aerial vehicles (UAVs), leveraging the advantages of flexible deployment and a wide field of view, can effectively maintain inter-vehicle collaborative sensing. However, such UAV-assisted sensing still faces practical challenges such as complex communication environments and limited resources. This paper focuses on task offloading and resource allocation for UAV-assisted collaborative sensing in IoV. By constructing a system model and deeply analyzing the vehicle motion, communication, and computation processes, a resource allocation scheme based on deep reinforcement learning (DRL) is proposed. This scheme leverages DRL's strong global exploration, adaptability, and nonlinear modeling capabilities to jointly optimize wireless communication and computing resources in each time slot, aiming to improve the completion rate of vehicle-UAV collaborative sensing tasks. The Actor-Critic structure in the proposed DRL scheme mitigates the temporal difference overestimation problem via multiple Critic networks, target networks, and experience replay buffers, and optimizes the Actor network using deterministic policy gradients—effectively enhancing the stability of policy updates and learning efficiency. Simulation experiments verify and analyze the proposed scheme under different parameter settings, identifying optimal parameters such as a batch size of 128 and a learning rate of 0.001. The comparison results demonstrate that the proposed scheme significantly outperforms other comparative schemes in indicators such as reward values and task completion rates. When the number of vehicles increases from 10 to 22, it achieves an average improvement of 21%, 135%, and 276% in task completion rate compared to the three baseline schemes, respectively. This fully verifies its performance advantages in complex scenarios with dynamic changes in the number of vehicles. Open source code link: https://github.com/renruixing/Open-Source-Code-Section