Operational performance of a congested corridor with lanes dedicated to autonomous vehicle traffic

This paper considers the operational performance impact of autonomous vehicles (AV) on a multi-lane freeway corridor with separate lanes dedicated to AV and non-AV traffic. Autonomous vehicle behavior is modeled at the macroscopic level by modifying the fundamental diagram relating hourly traffic flow and vehicle density, a step that is justified by adjusting a parameter from Newell's car-following model at the microscopic level and transforming back to a macroscopic representation. The model is applied to the I-95 corridor between Washington, DC and Baltimore, MD during the PM peak period, where the impact of introducing a managed AV-only lane is assessed at varying penetration rates of autonomous vehicles. The results show that the overall corridor performance metrics improve with increasing penetration rates up to 30%, 40% or 50% (depending on the underlying assumptions that govern AV behavior), after which the performance deteriorates drastically. Implications of the results are discussed in light of the per-lane and aggregated metrics, and future directions for research are proposed.