Smart Traffic Operation: from Human-Driven Cars to Mixed Vehicle Autonomy

The goal of my research is to enhance urban mobility by developing reliable and efficient traffic control and management strategies. As cities grow everywhere, and urban roadways become overburdened, the need for the development of such strategies becomes more evident. With the prevalence of smart sensing devices, such as smart phones and smart intersections, cities are becoming smart. Moreover, with the emergence of new and inevitable technologies, such as autonomous and connected vehicles, electric vehicles, and mobility on demand systems, smart cities are rapidly evolving. As we experience the arrival of such technologies, there is an opportunity to reclaim urban mobility. However, a blind utilization of these technologies may deflect us from reaching this goal. In this dissertation, we study the efficient operation of smart cities via management strategies that can guarantee overall societal benefits both in the cities of today and future.We focus on two natural instances of this agenda. In the first part, we tackle some of the existing challenges in the smart operation of traffic networks which are solely shared by human-driven cars. If all vehicles are human-driven, there is room for improving the efficiency of traffic networks by appropriate coordination and control of traffic signal lights. For these networks, we develop signal control algorithms that are capable of minimizing the number of stop-and-go movements, encoding fairness among vehicular arrivals, and are robust to the knowledge of system parameters. In the second part, we analyze fundamentals of traffic networks with mixed vehicle autonomy, where both human-driven and autonomous cars coexist on roadways. We study the mobility implications of selfish autonomy, i.e. autonomous cars that are not concerned about their overall impact and simply attempt to optimize their own travel benefits. Having shown the negative consequences that the increased deployment of selfish autonomy may have, we develop a pricing mechanism which can guarantee the overall societal-scale efficiency of traffic networks with mixed vehicle autonomy. Finally, we show how autonomy can act altruistically, i.e. by taking into account the decision making process of humans, autonomous cars can potentially plan for their actions in the favor of the overall good