Detection and optimization problems with applications in smart cities
MetadataShow full item record
This dissertation proposes solutions to a selected set of detection and optimization problems, whose applications are focused on transportation systems. The goal is to help build smarter and more efficient transportation systems, hence smarter cities. Problems with dynamics evolving in two different time-scales are considered: (1) In a fast time-scale, the dissertation considers the problem of detection, especially statistical anomaly detection in real-time. From a theoretical perspective and under Markovian assumptions, novel threshold estimators are derived for the widely used Hoeffding test. This results in a test with a much better ability to control false alarms while maintaining a high detection rate. From a practical perspective, the improved test is applied to detecting non-typical traffic jams in the Boston road network using real traffic data reported by the Waze smartphone navigation application. The detection results can alert the drivers to reroute so as to avoid the corresponding areas and provide the most urgent "targets" to the Transportation department and/or emergency services to intervene and remedy the underlying cause resulting in these jams, thus, improving transportation systems and contributing to the smart city agenda. (2) In a slower time-scale, the dissertation investigates a host of optimization problems, including estimation and adjustment of Origin-Destination (OD) demand, traffic assignment, recovery of travel cost functions, and joint recovery of travel cost functions and OD demand (joint problem). Integrating these problems leads to a data-driven predictive model which serves to diagnose/control/optimize the transportation network. To ensure good accuracy of the predictive model and increase its robustness and consistency, several novel formulations for the travel cost function recovery problem and the joint problem are proposed. A data-driven framework is proposed to evaluate the Price-of-Anarchy (PoA; a metric assessing the degree of congestion under selfish user-centric routing vs. socially-optimal system-centric routing). For the case where the PoA is larger than expected, three viable strategies are proposed to reduce it. To demonstrate the effectiveness and efficiency of the proposed approaches, case-studies are conducted on three benchmark transportation networks using synthetic data and an actual road network (from Eastern Massachusetts (EMA)) using real traffic data. Moreover, to facilitate research in the transportation community, the largest highway subnetwork of EMA has been released as a new benchmark network.