While modern roundabout has been proven superior to signalized intersection in safety, delay, and capacity under homogeneous traffic, its merits in heterogeneous traffic are neither proven nor analyzed. Both unique characteristics and models to reproduce behaviors in simulation has not been discussed comprehensively, especially in the high proportion of two-wheeler (TW) condition. In this background, the study aims to answer the question of how to reproduce unique characteristics of TW at roundabout in simulation. In order to achieve the answer, the study has gone through the below procedures step by step.
Firstly, chapter 1 presents background, research needs, research gap, research question, objectives and scope. In chapter 2, the literature review section determined the two constitutions of mixed traffic that are the performance rule and the appearance of the small-sized vehicle. Traffic in Ho Chi Minh city, Vietnam, satisfies both two conditions and has a uniqueness that only one type of small-sized vehicle as well as non-lane-based vehicle, TW, also named as motorcycle, and its dominance in traffic proportion. Thus, roundabouts in the city are selected as case studies. Moreover, this section also goes through the concepts and techniques related to model development, collective behavior, two-player game theory, agent-based modeling.
Secondly, the surveyed videos are recorded by unmanned areial vehicle (UAV), DJI Phantom 4 Pro, and the trajectory data is extracted by using semi-automated software. The accuracy of extracted data is examined with under 3.4% error. From extracted data, the macroscopic and microscopic characteristics of TWs are analyzed. Chapter 3 highlights three points that are the exponential relationship between turning angle rate and speed, the small critical gap of TW, 1.25 seconds, and the oval shape of the following space.
Chapter 4 presents the model development in detail and its components. Based on the collective behavior and game theory, the TW’s interaction model is built at the microscopic level, including regular movement model, conflict-solving model, and collective behavior model. The implementation of the model in traffic simulator is detailed in chapter 5. The simulator is built based on the multi-agent programmable modeling environment, Netlogo. The parameters are calibrated using half of the collected data.
Chapter 6 presented a series of results and indicators for validating the developed simulator. The remaining collected data is using for validation. Totally nine indicators are used for both macroscopic and microscopic validation. The total turning angle and low-speed duration are uniquely proposed in this study for validation. In total, the developed simulator is validated as good in representing both macroscopic and microscopic characteristics. The developed simulator is superior to the popular commercial software for heterogeneous traffic, PTV VISSIM, in simulating heterogeneous traffic at the roundabout.
Finally, chapter 7 concluded the study’s achievements, contributions, and limitations. Concerning research interest, the study proposed the novel TW’s interaction model at the individual level. The model potentially applies and expends in the future to improve the accuracy of the future microscopic simulations. In addition, the developed agent-based simulator also has practical contributions. It could be a useful tool for measuring the current roundabout performance under different schemes, or for policy test as well as new geometric design test.