Digital Twin for UAV-RIS Assisted Vehicular Communication Systems

Abstract

This paper investigates the issue of resource allocation for unmanned aerial vehicle and reconfigurable intelligent surface (UAV-RIS) assisted vehicular communication systems. To adapt the high dynamics of vehicular networks, we conceive a digital twin-based system over RIS-embedded environment towards environmental-aware communications. Specifically, a digital twin system can leverage data-driven models to predict the large-scale fading of future stages, while RIS is capable of controlling the propagation environments in real time, which can be utilized to mitigate prediction errors imposed by the small-scale fading. Using the capabilities of "prediction" and "reconfiguration", we expect to comprehensively foresee the dynamic changes in vehicular networks. In particular, the above-mentioned issue is formulated as a multi-slot total power consumption minimization problem under the quality of service (QoS) and energy constraints. Considering the finite battery energy of the UAV and the circuit power of the RIS, the transmit power of the UAV and the number of active reflecting elements (REs) are jointly scheduled for a finite time horizon. To tackle this mixed integer non-linear programming (MINLP) problem, we transform the original model into a discrete-time dynamic system. According to whether the dynamics of radio environments are predictable or not, the optimal offline and online policies are derived by using the deterministic and stochastic dynamic programming algorithms, respectively. To reduce the computation complexity, we further propose a novel online policy based on the idea of double-strategy selection. Finally, numerical results demonstrate that the proposed online policy exhibits near-optimal performances and outperforms other benchmarks in terms of transmission failure probability and effective power consumption.