With the remarkable increase in the population and number of vehicles, traffic has become a severe problem in most metropolitan areas. Traffic congestion has imposed tight constraints on economic growth, national security, and mobility of riders and goods. The open-mode integrated transportation system (OMITS) has been designed to improve the traffic condition of roadways by increasing the ridership of vehicles and optimizing transportation modes through smart services integrating emerging information communication technologies, big data management, social networking, and transportation management. Even a modest reduction in the number of vehicles on roadways will lead to a considerable cost savings in terms of time and money. Additionally the reduction in traffic jams will lead to a significant decrease in both gasoline consumption and greenhouse gas emissions. As a result, novel transportation management is critical to reduce vehicle mileage in the peak time of the road network. The OMITS was proposed to enhance transportation services in respect to the following three aspects: optimization of the transportation modes by multimodal traveling assignment, dynamic routing and ridesharing service with advanced traveler information systems, and interactive user interface for social networking and traveling information.

Therefore, the OMITS encompasses a broad range of advanced transportation research topics, say dynamic trip- match, transportation-mode optimization, traffic prediction, dynamic routing, and social network- based carpooling. This dissertation will focus on a kernel part of the OMITS, namely traffic simulation and prediction based on data containing the distribution of vehicles and the road network configuration. A microscopic traffic simulation framework has been developed to take into account various traffic phenomena, such as traffic jams resulting from bottlenecking, incidents, and traffic flow shock waves. Four fundamental contributions of the present study are summarized as follows: Firstly, an accurate and robust vehicle trajectory data collection method based on image data of unmanned aerial vehicle (UAV) has been presented, which can be used to rapidly and accurately acquire the real-time traffic conditions of the region of interest. Historically, a lack in the availability of trajectory data has posed a significant obstacle to the enhancement of microscopic simulation models.

To overcome this obstacle, a UAV based vehicle trajectory data collection algorithm has been developed. This method extracts vehicle trajectory data from the UAV’s video at different altitudes with different view scopes. Compared with traditional methods, the present data collection algorithm incorporates many unique features to customize the vehicle and traffic flow, through which vehicle detection and tracking system accuracy can be considerably increased. Secondly, an open mechanics-based acceleration model has been presented to simulate the longitudinal motion of vehicles, in which five general factors—namely the subject vehicle’s speed and acceleration sensitivity, safety consideration, relative speed sensitivity and gap reducing desire—have been identified to describe drivers’ preferences and the interactions between vehicles. Inspired by the similarity between vehicle interactions and particle interactions, a mechanical system with force elements has been introduced to quantify the vehicle’s acceleration.

Accordingly, each of the aforementioned five factors are assumed to function as an individual trigger to alter each vehicle’s speed. Based on Newton’s second law of motion, the subject vehicle’s longitudinal behavior can be simulated by the present open mechanics-based acceleration model. By introducing feeling gap, multilane acceleration behavior is included in the presented model. The simulation results fit realistic conditions for the traffic flow and the road capacity very well, where traffic shockwaves can be observed for a certain range of the traffic density. This model can be extended to more general scenarios if other factors can be recognized and introduced into the modeling framework. Thirdly, a driver decision-based lane change execution model has been developed to describe a vehicle’s lane change execution process, which includes two steps, i.e. driver’s lane selection and lane change execution. Currently, most lane change models focus on the driver’s lane selection, and overlook the driver’s behavior during a process of lane change execution which plays a significant role in the simulation of traffic flow characteristics.

In this model, a lane change execution is analyzed as a driver’s decision-making process, which consists of desire point setting, priority decision-making, corresponding actions and achievement of consensus analysis. Compared with the traditional lane change execution models, the present model describes a realistic lane change process, and it provides more accurate and detailed simulation results in the microscopic traffic simulation. Based on the presented open mechanics-based acceleration model and the driver decision- based lane change execution model, a reverse lane change model has further been developed to simulate some complex traffic situations such as reverse lane change process at a two-way-two- lane road section where one lane is blocked by a traffic incident. Based on this reverse lane change model, information on the average waiting time and road capability can be obtained. The simulation results show that the present model is able to reflect real driver behavior and the corresponding traffic phenomenon during a reverse lane change process Through a homogenization process of the microscopic vehicle motion, we can obtain the macroscopic traffic flow of the roadway network within certain time and spatial ranges, which will be integrated into the OMITS system for traffic prediction.

The validation of the models through future OMITS operations will also enable them to be high fidelity models in future driverless technologies and autonomous vehicles.