An Advanced Intelligent Route Guidance System in Urban Traffic Network with Mixed Autonomous Vehicles and Human-Driven Vehicles
Bokui Chen,Kahing Lam,Jun Zhou,Lixin Miao
DOI: https://doi.org/10.2139/ssrn.4059161
2022-01-01
SSRN Electronic Journal
Abstract:In recent years, with the advance in both processing power and algorithmic design, there have been numerous studies on both the improvement of route guidance systems on traffic performance and the introduction of autonomous vehicles(AV), into a traditional traffic system filled with human-driven vehicles(HV). This paper aims to combine these two field of studies and investigates their combined effect on the future intelligent traffic system. To such end, in this paper, we conduct and study the microscopic simulation of different route guidance strategies under local and global algorithm in a mixed AV/HV traffic flow. It for the first time, such study of route guidance strategies in mixed traffic flow is performed in a complex traffic scenario such as Manhattan Network, and based on the cellular automation model, we propose a new collision-free AV speed update model. In addition, taking into consideration of the sensory capabilities of AVs and the advent 5G communication networks, we design an efficient intersection passing model to describe AVs' speed state upon reaching the intersection. From the simulation result, we have come to the following important conclusions: (1) Congestion coefficient based global algorithm yields superior improvement on the traffic flow. And given the local congestion coefficient based routing algorithm is applied, (2) with more AVs in the traffic network, the average flow at both free-flow state and saturation state increases significantly. (3) Higher AV/HV ratio does not guarantee traffic improvement across the whole range of vehicle density. In particular, the critical density of saturation state, meta state, and deadlock state shrink when there are more AVs run in the traffic. (4) Given different AV/HV ratios, there exist different respective optimal sensing ranges or foresight k, above which little improvement on traffic flow could be observed. (5) Taking a step further, we also investigate how different information collection periods for local congestion coefficient based routing algorithm affect the average flow. We found that real-time traffic information collection is not necessary as collecting information every one-hundred seconds could yield nearly the same traffic performance like that in real-time.