The traffic safety and efficiency applications made possible by vehicular communications have the potential to improve the lives of millions of people who, every day, use automobiles as their primary means of transportation. To be well connected and fully functional, these networks of cars require a minimum number of active nodes, which often may not happen due to a lack of radio-equipped vehicles on the road. These same networks can also be overwhelmed with traffic and signaling in the presence of too many cars, requiring careful coordination between all nodes to ensure proper operation. One way to overcome both these problems is to supplement vehicle-to-vehicle (V2V) communications with vehicle-to-infrastructure (V2I) systems by deploying Roadside Units (RSUs) along the road to support the network of moving cars. RSUs are infrastructure nodes that can supplement sparse networks in low-density scenarios, and help coordinate and move data in denser networks. RSUs have an associated cost, however, and so their numbers need to be minimized while still maintaining a significant improvement to the vehicular network. he work presented in this thesis quantities the benefits of Roadside Unit deployments and proposes innovative approaches that can reduce and even eliminate the need for RSUs altogether.he first part of the thesis focuses on highway networks: first, an analytical model is developed to analyze communication delay in scenarios with sparse bi-directional traffic, considering both disconnected and connected RSUs.hen, a study on connectivity and message dissemination in these networks reveals how significant benefits of RSUs are only achieved when the deployed RSUs are interconnected. Extensive simulation work paired with sets of experimental measurements validate both model and study. Supplementing the work on sparse highway networks, an infrastructure-less approach is then proposed, consisting of two methods to improve communication delays in these scenarios: decelerate disconnected vehicles as they receive safety messages, and boost the same vehicles’ radio transmit power, to shorten the time to restore connectivity. Both techniques are modeled analytically, and data from a simulation study validate the models and show significant improvements in the connectivity of sparse highway networks with this infrastructure-less approach. he second part of the thesis sets its sights on urban vehicular networks. High costs associated with RSUs prevent their deployment at scale, and therefore finding alternative solutions to this longstanding problem is very important. A novel, low-cost self-organizing network approach to leveraging parked cars as RSUs in urban areas is proposed here, enabling parked cars to create coverage maps based on received signal strength and to decide whether to become RSUs from that knowledge. Initial simulation work reveals significant benefits to emergency message broadcasting delay in sparse scenarios and shows the ability of the self-organizing approach in providing robust and widespread coverage to dense urban areas, using only a small fraction of the cars parked in a city. he parking behaviors of individual drivers are then studied, by analyzing and gathering statistics on travel survey data from various metropolitan areas. Daily and hourly analytical models of parking events are provided, along with important derivations.he statistical data show that parking events can be classified into two major groups based on the time a car spends parked, and that these patterns vary substantially throughout the day while being markedly similar across different cities. he last part of the thesis focuses on self-organization for parked car RSUs. Novel mechanisms for self-organization are introduced that are innovative in their ability to keep the network of parked cars under continuous optimization, in their multicriteria decision process, and in their control of each car’s battery usage, rotating roadside unit roles between vehicles as required.he first comprehensive study of the performance of such approaches is presented, via realistic modeling of mobility, parking, and communication, thorough simulations, and an experimental verification of concepts that are key to self-organization. his analysis leads to strong evidence that parked cars can serve as an alternative to fixed roadside units, and organize to form networks to support smarter transportation and mobility.