Announcing the Final Examination of Fatima Afifah for the degree of Doctor of Philosophy
Smart mobility, enabled by advanced sensing, communication, vehicle, and emerging mobility technologies, has transformed transportation systems. Real-time information shared by public and private entities plays a pivotal role in smart mobility, which facilitates informed decision-making, including effective mode choice, dynamic vehicle control, optimized travel routing, and strategic vehicle relocation. While more information is believed to benefit individual decision makers, it is crucial to acknowledge that the effects of information on transportation network performance are
contingent; more information may not always benefit the safety and mobility of the whole system.
The goal of this dissertation is to investigate the effects of information shared by public and private transportation entities on system-level performance. The challenges are primarily due to the lack of a unified modeling framework to endogenously reflect the decentralized multi-agent interaction involved in the interconnected transportation networks and the resulting computational complexities arising from non-convexity and high dimensionality. To address these
challenges, this dissertation proposes novel modeling frameworks and computational solutions for three cutting-edge smart mobility applications. First, to examine the impact of en-route information on a transportation network, we propose a novel two-stage stochastic traffic equilibrium model to characterize the equilibrium traffic patterns considering adaptive routing behavior when locational en-route traffic information is provided through infrastructure-to-vehicles (I2V) communications. This model is formulated as a convex stochastic optimization problem so that efficient stochastic programming algorithms can be directly leveraged to achieve scalability. Second, to achieve optimal control over real-time variable speed limits information sharing and evaluate its impact on the network, we propose a twin-delayed deep deterministic policy gradient model, which converges more reliably than state-of-the-art deep reinforcement learning models. We investigate the transferability of the control algorithm and conduct comparative analyses of different traffic control strategies and spatial distributions of variable speed limit control (VSLC) deployment. Third, to assess the impacts of information provided by private ride-sourcing companies on transportation network congestion, we propose a Stackelberg framework for spatial pricing of ride-sourcing services considering traffic congestion and convex
reformulation strategies under mild conditions. We perform numerical experiments on transportation networks of varying scales and with diverse transportation network company (TNC) objectives, aiming to derive policy insights regarding the implications of spatial pricing information on transportation systems.
Committee in Charge: Zhaomiao Guo (Chair), Naveen Eluru, Shaurya Agarwal, Qifeng Li
Read More