Integration of Autonomous Vehicles with Adaptive Signal Control to Enhance Mobility
This report summarizes the results of a one-year project aimed at exploiting vehicle-toinfrastructure (V2I) communication to enhance the effectiveness of real-time adaptive traffic signal control systems. As originally formulated, the project’s goal was to explore the potential of using the sensing capabilities of connected autonomous vehicles (CAVs) to detect other vehicles in close proximity and use this information to “virtually increase” the level of penetration of connected vehicles in the traffic network, and enhance the predictive accuracy of real-time traffic signal control. However, following initial discussions with project partners Rapid Flow Technologies Inc., provider of the surtrac adaptive traffic signal control system , and Argo AI, an autonomous vehicle technology company, the project focus was shifted to a problem of more immediate and pragmatic importance: a field demonstration and analysis of the ability to further optimize traffic signal control performance through vehicle-toinfrastructure (V2I) communication of real-time CAV route information. In [Hawkes 2016], a mechanism was proposed for incorporating this information into surtrac to reduce uncertainty and generate more accurate predictions of vehicle flows through a controlled traffic network. A benefits analysis of this mechanism, conducted using a microscopic traffic simulation of various traffic networks, showed that network delay was substantially reduced for those vehicles willing to share their routes, and moreover, there was little adverse effect (and even some benefit) to those vehicles not sharing route information.
Building on these ideas, this project has aimed at demonstrating and evaluating V2I route sharing performance in the field and further validating these claims. A cloud-based mechanism for sharing Argo vehicle routes with surtrac in real-time was developed, and appropriate extensions to the current commercial implementation of surtrac were made to factor this additional information into traffic signal control decisions. A pilot test experiment was then designed and carried out using Argo AI test vehicles within the Pittsburgh surtrac deployment. Comparative vehicle delay data were collected along the various routes driven, both with and without route sharing enabled, and results corroborate the benefits predicted by the earlier simulation analysis. Argo vehicles experienced an average reduction in delay of 20% when they shared their routes, and additional analysis of overall surtrac network delay indicated essentially no change in travel time performance to other vehicles. These results are important in that they show that directly benefiting a subset of vehicles does not have to be a zero-sum game, and enable a new, more sustainable model for upgrading urban infrastructure and improving urban mobility through voluntary tolling at the intersection.
Read the full study here.