Exploring the use of probe vehicle data for system state estimation and traffic signal control

The ability to observe traffic conditions in real-time is critical for the majority of Intelligent Transportation Systems (ITS) which are designed and deployed to address the growing traffic congestion problems on surface transportation networks. Technology has evolved to a point where thousands of vehicles equipped with tracking, computing, and communication technologies---so-called probe vehicles---can be utilized as mobile sensors to collect real-time traffic data in large-scale transportation networks as an alternative option to the traditional vehicle detectors such as inductive loops and radar that are placed at selected points along the roadway. Probe vehicles can be tracked anonymously and can report data on their locations, speeds, travel times and arrival times as they perform their regular trips. Probe data can then be used to estimate and monitor traffic conditions, which is a major component in traffic control and management.;In this research, new models for estimating the system state or key parameters (e.g., queue lengths, flow rate) from probe data are developed. Since not all vehicles in the traffic stream may be instrumented, reliable and robust prediction methods are developed to determine the state of the system from the probe data. The tradeoff between the percentage of probe vehicles in the traffic stream and the accuracy of the estimated parameters is investigated. Also, a new innovative signal control method is developed to optimize traffic operations at a signalized intersection based on probe data, i.e., queue lengths. This method uses the queue length information to adjust the maximum green times in every cycle. A state of the art microsimulation platform in VISSIM is developed to test, validate, and evaluate the models formulated in this research. The benefits of using the queue length information on signal operations are assessed in VISSIM for different scenarios under different assumptions (e.g., different traffic demand levels) while representing the vehicular movements realistically. For a simple isolated intersection, it is found that the proposed signal control method that uses queue length information performs significantly better than the typical fully actuated control logic.;In addition, new methods are developed to test the optimality of the signal timing from probe vehicle data for offline applications. Traffic signals in the current state-of-the-practice are generally optimized based on the data collected for a fixed period of time (e.g., several hours of data collected on a weekday peak period). If actual volumes deviate from the data collected previously then the signal timing plans may no longer be optimal. The proposed methods use probe information to estimate flow rates so that possible deviations from the assumed flow rates can be evaluated. There are several types of basic information that can be obtained from probe vehicles: location of probes in the queue in relation to the stop bar, time instance they join the back of the queue, and total count (number of probes). A methodology based on statistical decision theory is employed to determine what information element(s) to use and how to use them (i.e., which decision function to use) such that the flow rate (at a particular time of the day) is estimated with the least error or maximum accuracy. Numerical experiments are conducted to evaluate selected decision functions under different traffic demand scenarios.