We studied several important management problems in modern transportation systems based on proper use of various types of information in different ways. We focused on two main research questions: 1) How to design smart information schemes for decentralizing better (e.g., more efficient, stable and sustainable) flow patterns on transportation networks (Chapter 1 & 2); and 2) How to utilize information and system data fully and efficiently for better (e.g., closer to optimal and more cost-effective) centralized decision making (Chapter 3 & 4). Specifically, we have explored the following four dimensions.

1) Strategic information scheme design for enforcing optimal flow on traffic networks with minimal tolls. We explored how disclosing flexible new information can help reduce the toll intensity needed for decentralizing a Nash equilibrium on a general traffic network that minimizes certain system-level cost. We formulated the Minimal Toll Information Design Problem (MTIDP) and designed efficient algorithms for finding near-optimal solutions to the problem. Numerical examples are used to reveal insights of MTIDP and validate the effectiveness of the proposed solution algorithms.

2) Remedy of the negative effect of inaccurate travel time estimate on dynamic routing using additional endogenous information feedback. We proposed to provide en-route real-time traffic-sensitive pollution information to drivers for suppressing traffic oscillations caused by delay in travel time reporting. Theoretical analysis (based on a novel queueing model), numerical examples, and simulation experiments for simple traffic networks are adopted to demonstrate the potential traffic stabilizing benefit of this new information.

3) Utilization of system data and demand forecast for controlling complex human-centered infrastructure systems. We used multi-access managed lanes systems as an illustrative application. With the available measurement of traffic condition and demand forecast, we developed a hybrid model predictive control based dynamic pricing algorithm using origin-destination specific tolls. Through proper formulation of system models and practical constraints, the proposed control model can be implemented efficiently in real-time.

4) Value of information and optimal learning in solving large scale network optimization problems with uncertainty. We looked at the challenging second-best network pricing problem (SNPP) with stochastic demand. We designed Bayesian learning model for the problem and tailored linear belief-based Knowledge Gradient sampling policy to SNPP. Experiment on a benchmark network with more than a million candidate solutions showed superior performance of our approach to the benchmark heuristic. We have proposed novel methodology and generated new insights in each dimension, concrete examples are involved. Our goal is to provide useful references, practical solutions and new thinking for existent nontrivial problems and emerging challenges in traffic management under this information age and unprecedented demand for efficient and sustainable urban mobility.