Abstract
Location-based services have been identified as a promising communication paradigm in highly mobile and dynamic vehicular networks. However, existing mobile ad hoc networking cannot be directly applied to vehicular networking due to differences in traffic conditions, mobility models and network topologies. On the other hand, hybrid architectures in vehicular networks, with ad hoc-based inter-vehicle and infrastructure-based vehicle-to-roadside communications, can facilitate robust and efficient communication services using geographical information. In this dissertation, we focus on the design and evaluation of location-based protocols and algorithms to improve scalability, efficiency, and resiliency in hybrid vehicular networks. We first provide a cross-layer self-localization algorithm for moving vehicles. A new ultra-wide band (UWB) coding method, based on an orthogonal variable spreading factor and time hopping, is proposed for minimum interference during ranging. Then, a UWB based non-metric multidimensional scaling derives accurate and robust self-localization results. In addition, we employ an online compressive sensing scheme to count and localize sparse roadside units (RSUs) for war-driving applications. Online war-driving records received signal strength (RSS) values at runtime, and can recover the number and location of RSUs immediately based on far fewer noisy RSS readings. After obtaining the location information of vehicles and RSUs, we address multiple channel scheduling in hybrid vehicular networks. We use the natural beauty of Latin squares to achieve fair and deterministic scheduling in micro-time scale for channel access and macro-time scale for channel assignment. A grid based scalable scheme is proposed to map Latin squares to grids for dynamic single-radio multi-channel scheduling. Another interference graph based scheme uses nodal location and social centrality to reflect the social behavior patterns related to access in vehicular networks, and then form adaptive clusters for multi-radio multi-channel scheduling. We also investigate several vehicular environments, and propose corresponding location- and environment-aware data dissemination solutions. We first present an efficient on-demand bounce routing method in vehicular tunnels. It applies a hybrid signal propagation model and location-based forwarding metric to choose the best data dissemination strategy. Then, we design a hybrid routing scheme for robust and reliable data dissemination in urban transportation environments, in which the choice of communication method is dependent upon geographical connectivity, by taking network coding based multicast routing in dense network and opportunistic routing using carry and forward method in sparse network. In addition, we propose an online learning based knowledge dissemination in unmanned aerial vehicle (UAV) swarms under delay/disruption-tolerant networking, where each UAV adaptively chooses broadcast probability by learning link status. A fractionated Cyber-Physical System framework, based on partial ordering for knowledge sharing and colored Petri net for work flow, is implemented to achieve distributed knowledge management in UAV swarms. Our extensive simulation and real testbed results show the robustness and efficiency of location-based services in vehicular networks with hybrid architectures.
