DATA-DRIVEN OPTIMIZATION-BASED TRANSPORTATION MODELS FOR CONNECTED AND AUTOMATED VEHICLES

This dissertation proposes a data-driven optimization-based framework to model traffic dynamics under uncertainty including travel demand, transportation network (i.e. route choice), and connected and automated driving dynamics (on freeways and arterials) using connected vehicle data though Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), and Vehicle-to-Network (V2N) communications. To model travel demand dynamics, this dissertation proposes Distributionally Robust Stochastic Optimization (DRSO) using V2N data. A challenge of modeling travel demand dynamics directly using the real-world V2N data is the incomplete and inaccurate trajectory records from the raw data due to technical and privacy issues. Through the proposed DRSO models, this dissertation offline reconstructs the missing choices of activity locations, durations, and paths using the partially observed trajectories from a real-world connected vehicle dataset. This dataset contains around 2,800 connected vehicles over two separate months in Southeast Michigan from the Safety Pilot Model Deployment (SPMD) project. For modeling route choice dynamics, this dissertation develops a Conditional Value-at-Risk (CVaR) based DRSO model for the route choice problem under the impacts of travel time uncertainties and travelers’ risk attitudes using vehicle trajectory data from the SPMD dataset. The proposed CVaR-DRSO model offline estimates route choices under uncertainties by using a data-driven uncertainty set. For modeling connected and automated driving, this dissertation develops DRSO-based Model Predictive Control (MPC) models with Distributionally Robust Chance Constraints (DRCC). For the connected and automated driving on freeways (i.e. uninterrupted flow facilities), a DRSO-DRCC based MPC model is proposed to improve the stability, robustness, and safety for the online longitudinal cooperative automated driving of a platoon of Connected and Automated Vehicles (CAVs) under uncertain traffic conditions by using real-time V2V data. For the energy efficient connected and automated driving on arterials (i.e. interrupted flow facilities), a DRSO-DRCC based MPC model is developed to improve the safety, energy, efficiency, driving comfort, and robustness of the automated driving on signalized arterials under traffic uncertainties by using real-time V2I and V2V data. This dissertation provides a comprehensive data-driven optimization-based framework to model the traffic dynamics using the connected vehicle data and improve the connected and automated driving control under uncertainty based on CAV technologies