An agent-based urban simulation to assess the system-wide, spatial, and modal impacts of the deployment of an automated vehicle corridor with mixed traffic: a case study in New York City

Abstract

Self-driving car companies aim to offer hands-off journeys with Automated Vehicle (AV) technology, but public skepticism persists on safety aspects of AVs. A proposed solution involves creating dedicated AV corridors with traffic control and interruptions. Realistically, full penetration of AVs is unlikely to occur in the near future while mixed traffic of AVs with human driven vehicles is more likely. This study applies this concept to a hypothetical scenario in NYC, proposing an AV corridor from the Staten Island Expressway to the Brooklyn Battery Tunnel. The goal is to assess the network-wide impacts of a limited AV deployment from a multi-modal perspective. The NYC transportation system is modeled using MATSim, an open-source agent-based modeling platform. Although each traveler is treated as an agent making multi-modal trip making decisions, it does not model microscopic behavior of cars. Thus, the modeling approach for AVs has to be a macroscopic one, where average headways are reduced on the AV (with mixed traffic) corridor portion of the overall network based on a link capacity relationship adopted from a study on Advanced Cruise-Assist Highway Systems [1]. Two reduced headway scenarios are simulated in this MATSim NYC model, 1.1 and .85 second headways. Results indicate that Scenario B (1.1-second headway) is preferable over Scenario A (0.85 second headway), mitigating potential congestion on connectors, especially on Victory Boulevard in Staten Island. This study suggests implementing Scenario B for a smoother corridor-wide introduction of AVs to avoid congestion on connectors of the AV corridor to the rest of the NYC network.