Simulating the effectiveness of wave dissipation by FollowerStopper autonomous vehicles

Abstract The traffic jam phenomenon known as stop-and-go waves is commonplace on many roadways and is unavoidable due to the imperfect nature of human driving behaviour. Previous research by Stern et al. (2018) has demonstrated experimentally that a single Autonomous Vehicle (AV) can be harnessed to dissipate stop-and-go waves produced by 20 other passenger vehicles driving in a cycle. However, the experiment was conducted in an idealistic situation of a single-lane ring road with one AV; meaning no lane changing was possible and multiple AV interaction was left untested. To address these limitations, the AV driving behaviour used to achieve this, known as FollowerStopper (FS), was modelled for use in the Aimsun traffic simulation software and validated before further numerical experiments were conducted. The microsimulation scenarios were designed to observe how multiple AVs driven by the FS controller (referred to as FS-AVs in this paper) affect their traffic wave dissipation capability. Both single and double-lane ring roads experimental designs were used with the latter capturing the untested effect of lane changes. FS-AV was found to be less effective than originally documented in both cases. Multiple FS-AVs included within one lane were still able to improve traffic flow and dissipate the stop-and-go waves, although it is sensitive to the combination of different factors and any deviation away from the ideal condition is likely to produce worse-off traffic operations in terms of traffic flow. For double-lanes, human-driven vehicles (HDVs) would change lanes at heightened rates to occupy the larger gap the FS-AV needs for its dissipating strategy, causing it to further pullback and set off a chain reaction to upstream traffic. Stop-and-go waves had not been dissipated and decreased traffic performance in terms of flow, speed and delay time resulted. Our results also showed that the wave dampening effect of the FS-AV does not translate between the ring road and its linear equivalent. On the other hand, our preliminary simulation results using a real-life freeway model did not suggest multiple FS-AVs worsen traffic conditions. The contradicting results might be due to different traffic conditions such as vehicle density. Further research is required to investigate what factors could adversely affect FS-AV’s ability to dampen shockwaves and explore possible improvement to its driving algorithm.