The Effects of Varying Penetration Rates of L4-L5 Autonomous Vehicles on Fuel Efficiency and Mobility of Traffic Networks 2020-01-0137
With the current drive of automotive and technology companies towards producing vehicles with higher levels of autonomy, it is inevitable that there will be an increasing number of SAE level L4-L5 autonomous vehicles (AVs) on roadways in the near future. The effect of this gradually increasing penetration of AVs on mobility, viewed as traffic congestion or traffic flow efficiency in this paper, and fuel efficiency improvement for the individual AV and for the whole road network with a mixed traffic of AVs and non-AVs is currently not well known. Microscopic traffic simulators that simulate realistic traffic flow are crucial in studying, understanding and evaluating the possible effects of having a higher number of autonomous vehicles (AVs) in traffic under realistic mixed traffic conditions including both autonomous and non-autonomous vehicles. In this paper, L4-L5 AVs with varying penetration rates in total traffic flow were simulated using the microscopic traffic simulator Vissim on urban, mixed and freeway roadways to study the effect of penetration rate on fuel consumption and efficiency of traffic flow. The roadways used in these simulations were replicas of real roadways in and around Columbus, Ohio, including two AV shuttle routes in operation. The road-specific information regarding each roadway, such as the number of traffic lights and positions, number of STOP signs and positions, and speed limits, were gathered using OpenStreetMap with SUMO. In simulating L4-L5 AVs, the All-Knowing CoEXist AV model and a vehicle with Wiedemann 74 driver model were used to represent AV and non-AV driving, respectively. Then, the driving behaviors, such as headway time and car following, desired acceleration and deceleration profiles of AVs, and the non-AVs’ car following and lane change models were modified. The effect of having varying penetration rates of L4-L5 AVs were evaluated using criteria such as average fuel consumption, existence of queues and their average/maximum length, total number of vehicles in the simulation, average delay experienced by all vehicles, total number of stops experienced by all vehicles, and total emission of CO, NOx and volatile organic compounds (VOC) from the vehicles in the simulation. The results showed that while increasing penetration rates of L4-L5 AVs generally improved overall fuel efficiency and mobility of the traffic network, there were also cases when the opposite trend was observed.