Semi-Persistent Resource Allocation Based on Traffic Prediction for Vehicular Communications

Abstract

In cellular vehicular communications, high density and mobility of vehicles require frequent resource allocation, which can cause network congestion and large signalling and processing delay. To overcome this problem, we propose a novel semi-persistent resource allocation scheme based on a two-tier heterogeneous network architecture. The architecture includes a central macro base station (MBS) and multiple roadside units (RSU). In the proposed semi-persistent scheme, the MBS pre-allocates persistent resource to RSUs based on predicted traffic, and then allocates dynamic resource upon real-time requests from RSUs while vehicles simultaneously communicate using the pre-allocated resource. A simple Space-Time k-Nearest Neighbour (ST-kNN) method is developed for short-term traffic prediction, and a geometric water-filling algorithm is developed for minimizing the relative latency. Simulation results validate the effectiveness of the proposed semi-persistent scheme in comparison with two benchmark schemes.