UC Berkeley

Transportation systems face increasing challenges in managing traffic congestion and improving mobility. Recent advancements in sensor networks and cyber-physical systems have led to an exponential increase in data collection, often presenting issues such as ambiguity, inconsistency, inaccuracy, and problematic data formats. This dissertation investigates leveraging data from multiple sources to develop traffic simulations. The developed simulations help understand dynamic navigation strategies in urban environments, focusing on how information-aware routing impacts congestion dynamics. This work explores the influence of dynamic routing algorithms on overall traffic performance and congestion levels. By employing the SIR model and average marginal regret, the research quantifies the effects of navigation apps on traffic patterns, highlighting the potential for increased congestion and its implications. Moreover, two case studies are presented, capturing the effects of traffic light coordination on routing and using speed limits as a control parameter to improve traffic performance. On another note, sensor data can be incomplete due to malfunctions. This research demonstrates how machine learning techniques can accurately fill in the missing data. Beyond macroscopic traffic analysis, this dissertation examines the individual behavior of drivers, which can lead to `phantom congestion'. With the growing interest in automated vehicles (AVs), a chapter evaluates the impact of microscopic traffic behavior on longitudinal AV control policies in a ring road setting. The investigation of stop-and-go waves in closed-circuit ring road traffic reveals that improvements are possible using specific AV controllers, which could be affected by their distribution in mixed traffic settings. Additionally, this dissertation presents a chapter on designing bus routes using location-based services data. By understanding routing behavior and integrating existing transit data with shortest path algorithms and machine learning techniques, we can design new bus routes and potentially enhance existing ones. Overall, through advanced traffic simulation techniques and data-driven analyses, this research demonstrates that a balanced mix of app and non-app users can improve traffic conditions, AVs can mitigate stop-and-go waves, and innovative bus route design can enhance urban transit systems.