Video Semantics based Resource Allocation Algorithm for Spectrum Multiplexing Scenarios in Vehicular Networks

Abstract

Due to the time-varying scenarios and multiple requirements in vehicular networks, it is difficult to guarantee the accuracy of video semantic understanding within the scarce spectrum resources. Existing resource allocation algorithms such as quality of service (QoS) based and quality of experience (QoE) based algorithms, respectively prone to optimize network performance and user experience, are no longer applicable to semantic understanding tasks. Furthermore, a recently proposed semantics based algorithm exclusively considers vehicle to infrastructure (V2I) video transmission tasks, regardless of the practical vehicular networks scenarios. To tackle the challenges above, we propose a video semantics based resource allocation model under the vehicle moving and spectrum multiplexing scenarios. On the ground of the time-dependent environment and diverse needs, multi-agent reinforcement learning, which peculiarly owns sequential decision and reward mechanism, is employed to obtain the optimal resource allocation scheme. Simulation results show the effectiveness of our proposed algorithm and its better performance than the traditional and existing semantics based algorithms.