Resource allocation for vehicular communications

Abstract

This thesis aims to develop efficient and effective resource allocation schemes to meet the diverse quality-of-service requirements of vehicular communications while taking into account the strong dynamics in vehicular environments. Specifically, we study the spectrum and power allocation problem in device-to-device (D2D)-enabled vehicular networks. We design low-complexity algorithms to maximize the capacity of vehicle-to-infrastructure (V2I) links while guaranteeing the reliability of each vehicle-to-vehicle (V2V) link, evaluated in terms of outage probabilities, using only slowly varying large-scale fading information or delayed rapidly varying small-scale fading information from periodic feedback. To further improve spectrum utilization, we investigate the case where each V2I link shares spectrum with multiple V2V links and exploit graph theoretic tools to develop high performance approximation algorithms to support flexible spectrum sharing in vehicular communications. For ease of (semi-)distributed resource management, we exploit recent results in multi-agent reinforcement learning to develop a learning-based resource allocation algorithm for vehicular agents. Resource sharing decisions are made based on a mix of slowly-varying global large-scale channel information and fast-varying local observations. The four proposed schemes, including both centralized and semi-distributed designs with varying complexity-performance tradeoffs, constitute a comprehensive study of the resource allocation problem in vehicular communications.