Highway traffic jams can be caused by external factors, such as physical bottlenecks and severe weather conditions, as well as internal disturbances induced by underlying driving behaviors of vehicles participating in the traffic. For example, conflicts of uncoordinated traffic can cause disturbances and thus impede the traffic throughput. As development of vehicle automation advanced in recent years, introducing vehicle automation technology in traffic became a potential solution to improve traffic flow. Vehicle automation technology can enable vehicles to coordinate with other vehicles and to respond faster and smoother to traffic situations, and thus reduce traffic jams.In this dissertation, we consider vehicle automation in three different traffic scenarios. Firstly, we study traffic jams at highway merging sections. The bottleneck at the intersection is because of accumulated conflicts of mainline traffic and on-ramp traffic. We propose methods based on vehicle-to-infrastructure and vehicle-to-vehicle communication to coordinate mainline traffic and on-ramp traffic. The effectiveness of proposed methods is evaluated in micro-simulation under different penetration rates of automated vehicles.
Secondly, we study the phantom traffic jams, which are stop-and-go waves that spontaneously emerge without apparent external factors. Experiments on closed ring roads have demonstrated that the stop-and-go waves can be induced purely by underlying dynamics of driving behaviors in the traffic. The emergence of stop-and-go waves on normal highways can be modeled with string unstable car-following models and the emergence of stop-and-go waves on closed-ring roads can be seen as an unstable dynamical system. We prove that for traffic of non-identical car-following models, the ring road stability is necessary for string stable car-following models; the ring road stability is also sufficient for string stability if traffic is formed of identical car following models. Because the ring road stability is essential for string stability, the ring road can be used to verify whether an automated vehicle control has potential to dissipate stop-and-go waves on normal highways. Because the ring road is a closed field, it is easier to control and easier to observe than an open-ended highway. Therefore, it would be an ideal field to benchmark performance of automated vehicle control. Ten automated vehicle controllers are benchmarked under different penetration rates and different distributions.
Lastly, a car-following control system for trucks is developed. The controller is composed of an upper controller and a lower controller. An adaptive parameter estimation is integrated with the lower controller to deal with unknown parameters. The design of the upper controller considers actuator uncertainties and the string stability condition. To achieve better car following performance in practice, we use vehicle-to-vehicle communication to enhance the car following control system. Experimental results showed that the enhanced system enables the truck to follow the leading vehicle well while the leading vehicle speed is varying.