Effects of Connected Automated Vehicles on Traffic Flow

Abstract

In this dissertation we provide a comprehensive framework for evaluating the merits of wireless vehicle-to-vehicle (V2V) communication on traffic. In particular we focus on mixed traffic scenarios that will dominate highways in the next several decades. Such mixed traffic primarily contains conventional human driven vehicles, however also includes connected human-driven vehicles, automated vehicles, and connected automated vehicles. Connected human-driven vehicles are human-operated vehicles that are able to send and receive messages using V2V. Automated vehicles rely on an internal computer (rather than a human) to process information from sensors such as cameras or radars to control their motion. Finally connected automated vehicles are automated vehicles that use information received from V2V communication in addition to sensory information for controlling their motion.

Our framework is based on developing a prototype connected automated vehicle and investigating its effects on traffic patterns amongst human driven vehicles. We first establish an experimental procedure and criteria for tuning the connected automated vehicle's controller to follow a connected human-driven vehicle at a desired distance. We then showcase an experimental configuration that allows us to observe traffic patterns in a three-car connected vehicle network, where our connected automated vehicle interacts with two connected human-driven vehicles. These experiments demonstrate the effectiveness of connected automated vehicles using beyond-line-of-sight information in promoting smooth traffic flow in a mixed traffic environment.

To investigate the effects of connected automated vehicles for large networks, we first focus on simple car-following models without communication, actuation or human reaction delay. For these models we are able to analytically characterize the traffic patterns occurring in human driven traffic at various densities, as well as show that connected automated vehicles can indeed mitigate congestion and promote stable uniform flow of traffic. By exploiting the cyclic symmetry of the governing equations, we rigorously show that the results hold for arbitrarily large connected vehicle networks, and also that the feedback to long-range information in connected automated vehicles should be carefully chosen to ensure the benefit to traffic flow.

Lastly we use simulations to investigate large connected vehicle networks, where delays, nonlinearities, wireless communication delay, and eclectic driving dynamics are considered. We use these simulations to demonstrate that indeed the information from beyond-line-of-sight is the key feature that allows the connected automated vehicles to bring significant benefits to traffic. Finally, we conduct penetration studies to quantify the extent to which connected automated vehicles may benefit traffic at partial penetrations, and discuss the implications of this study on the current competing wireless V2X communication technologies.