Transitioning to autonomous driving: Mixed vehicle autonomy levels on freeways

Abstract

The increasing presence of vehicle automation is transforming freeways into environments of mixed traffic, where vehicles of varying autonomy levels interact. Before all vehicles become fully autonomous, a transition will be made that causes a high mix of those autonomy levels. Therefore, this thesis researches the impact of different levels of vehicle automation on traffic performance and safety on a multi-lane freeway with an on-ramp. Microscopic simulation is utilised to explore how varying levels of vehicle automation, while taking human driving factors into account, affect traffic flow, speed, density and dangerous car-following interactions.
Currently, the majority of vehicles are defined as level 0 vehicles. This does not mean that these vehicles have no automated features at all but the Advanced Driver-Assistance Systems (ADAS) only provide temporary support such as an emergency brake. This is different for level 1 vehicles where the car-following driving tasks are automated and for level 2 vehicles both the car-following and lane-changing tasks are automated to support the driver. Level 3 vehicles are conditionally autonomous where all driving tasks are automated. The study aims to fill the knowledge gap in understanding the impact of these mixed traffic conditions on overall traffic dynamics.
To simulate the varying levels of automation, the study utilizes OpenTrafficSim (OTS), a microscopic traffic simulation software that incorporates a mental model to realistically represent human driving behaviour. This allows the simulation to account for human factors such as reaction time, perception, cognitive workload, and distractions, which are crucial in differentiating human drivers from automated vehicles. Four automation levels (0, 1, 2, and 3) defined by the Society of Automotive Engineers are modelled for specific driving characteristics within the freeway environment. Model parameters are adjusted for each level based on literature findings and practical considerations.
Simulation results indicate that the introduction of level 1 and level 2 vehicles, characterised by larger headway values, can negatively impact traffic performance but also result in less dangerous car-following behaviour. The increased headway leads to disruptions in traffic flow and an earlier onset of congestion. However, as the penetration rate of level 3 vehicles increases, traffic conditions significantly improve, with higher mean speeds, reduced travel times, and increased traffic flow observed. These findings highlight the potential benefits of higher levels of automation in enhancing traffic performance and safety.
The study also examines the impact of driver distraction on traffic performance and safety. By simulating both in-vehicle and roadside distractions, the research demonstrates that higher cognitive workloads can lead to more disruptive driving behaviour. As automation levels increase, the negative effects of distraction are mitigated.
Overall, this research provides valuable insights into the complexities of mixed traffic with varying automation levels. It demonstrates that while the transition phase may present challenges, higher levels of vehicle automation can significantly improve both traffic performance and safety on multi-lane freeways. Special emphasis is given to accurately simulating human driver behaviour and suggestions are made for future research, including the need for a dual-perception framework for more accurate modelling of level 1 vehicles and further investigation into the impact of different distraction types.