Abstract

Experience of daily commuters shows that stationary traffic patterns can be observed during peak periods in urban freeway networks. Such stationary states play an important role in various traffic flow studies. Conceptually, studies on the impact of capacity drop and design of traffic control strategies have been built on the assumption of stationarity. Mathematically, the existence and stability of stationary states in general road networks have been proved. Empirically, near-stationary states have been utilized for calibration of fundamental diagrams and investigation of traffic features at freeway bottlenecks. Therefore, an imperative need for real-world near-stationary data has been realized to better understand, investigate, and explore such above studies. However, there lacks an efficient method to identify near-stationary states.

To fill the gap, in this research, an automated method has been developed to efficiently identify near-stationary states from large amounts of inductive loop-detector data. The method consists of four steps: first, a data pre-processing technique is performed to select healthy datasets, fill in missing values, and normalize vehicle counts and occupancies; second, a PELT changepoint detection method is adopted to detect changes in means and partition time series into candidate intervals; third, informative characteristics of each candidate, including duration and gap, are defined and calculated; finally, near-stationary states are selected from candidates through duration and gap criteria.

A game theory approach is further designed to directly calibrate parameters of the above method. First, a multi-objective optimization problem is formulated to consider the quantity and quality of near-stationary states as the objective functions. Then the problem is converted into a non-cooperative game with at least one Nash equilibrium. To solve the game and obtain a unique solution, an alternated hill-climbing search algorithm is developed.

Furthermore, two calibration schemes for multi-lane and multi-class fundamental diagrams are respectively designed by utilizing near-stationary states. Such multi-commodity fundamental diagrams possess unifiable and non-FIFO properties and can capture interaction among different commodities. Calibration and validation results show that both the calibrated unifiable multi-lane and multi-class fundamental diagrams are well-fitted, physically meaningful, and have robust performance on the estimation of commodity flow-rates.

Abstract

Disasters, specifically earthquakes, result in worldwide catastrophic losses annually. The first seventy-two hours are the most critical and so any reduction in response time is a much-needed contribution. This is especially true in cases where parts of the communication infrastructure are severely damaged. Traditional disaster relief logistics models tend to rely on the assumption that information flow is continuous throughout the system following the onset of a natural disaster. A new integrated framework for disaster relief logistics that optimizes the movement of critical information along with physical movements is proposed in order to alleviate post-disaster conditions in a more accurate and timely manner. The framework consists of an information network and a transportation network with interrelationships. The framework was applied to the Irvine Golden Triangle Network and the Knoxville Network for up to three different cases. The DYNASMART-P simulation program performance was compared against the Time Dependent Network Simplex paths approach combined with the information updating feedback loop. The average total travel times of vehicles travelling to the trauma center in the study areas were compared in order to quantify the improvements of the integrated solution framework. The results show a significant reduction of average total travel times for vehicles transporting injured patients to the trauma center.

Abstract

The purpose of this study is to develop a traffic estimation framework which combines different data sources to better reconstruct the traffic states on the freeways. The framework combines both traffic parameter and state estimation in the same work flow, which resolves the inconsistency issue of most existing traffic state estimation methods.

To examine the quality of the traffic sensor data, the study starts with proposing the network sensor health problem (NSHP). The optimal set of sensors is selected from all sensors such that the violation of flow conservation is minimized. The health index for individual detector is then calculated based on the solutions. We also developed a tailored greedy search algorithm to find the solutions effectively. The proposed method is tested using the loop detector data from PeMS on a stretch of the SR-91 freeway. We compared the results with PeMS health status and found considerable level of consistency.

Two different traffic state estimation methods are proposed based on the data availability and traffic states. The LoopReid method is derived from the Newell’s simplified kinematic wave model by assuming the whole road segment is fully congested. We formulate a least square optimization problem to find the initial states and traffic parameters based on the first-in-first-out principle and the congested part of the Newell’s model. While developing the LoopCT method, we derived a counterpart of the Newell’s kinematic wave model in the Lagrangian coordinates under Eulerian boundary conditions. This model also leads to a new method to estimate vehicle trajectories within a road segment. We formulate a least square optimization problem in initial states and traffic parameters which works for mixed traffic states. The two estimation methods turned out to be highly related and the LoopCT method degenerates to the LoopReid method when the traffic is fully congested. The two methods are validated using two datasets from the NGSIM project. Both methods achieved considerable level of accuracy at reconstructing the traffic states and parameters.

Abstract

Strategies, models, and algorithms facilitating such models are explored to provide transportation network managers and planners with more flexibility under uncertainty. Network design problems with non-stationary stochastic OD demand are formulated as real option investment problems and dynamic programming solution methodologies are used to obtain the value of flexibility to defer and re-design a network. The design premium is shown to reflect the opportunity cost of committing to a “preferred alternative” in transportation planning. Both network option and link option design problems are proposed with solution algorithms and tested on the classical Sioux Falls, SD network. Results indicate that allowing individual links to be deferred can have significant option value. A resource relocation model using non-stationary stochastic variables as chance constraints is proposed. The model is applied to air tanker relocation for initial attack of wildfires in California, and results show that the flexibility to switch locations with non-stationary stochastic variables providing 3-day or 7-day forecasts is more cost-effective than relocations without forecasting. Due to the computational costs of these more complex network models, a faster converging heuristic based on radial basis functions is evaluated for continuous network design problems for the Anaheim, CA network with a 31-dimensional decision variable. The algorithm is further modified and then proven to converge for multi-objective problems. Compared to other popular multi-objective solution algorithms in the literature such as the genetic algorithm, the proposed multi-objective radial basis function algorithm is shown to be most effective. The algorithm is applied to a flexible robust toll pricing problem, where toll pricing is proposed as a strategy to manage network robustness over multiple regimes of link capacity uncertainty. A link degradation simulation model is proposed that uses multivariate Bernoulli random variables to simulate correlated link failures. The solution to a multi-objective mean-variance toll pricing problem is obtained for the Sioux Falls network under low and high probability seasons, showing that the flexibility to adapt the Pareto set of toll solutions to changes in regime – e.g. hurricane seasons, security threat levels, etc – can increase value in terms of an epsilon indicator.