Browsing by Author "Hu, Jeffrey"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Towards Implementation of Max-Pressure Signal Timing on Minnesota Roads
    (Minnesota Department of Transportation, 2022-12) Barman, Simanta; Levin, Michael W.; Robbennolt, Jake; Hu, Jeffrey; Odell, Michael; Kang, Di
    Max-pressure control is a new adaptive method for signal timing that is mathematically proven to achieve maximum throughput for the entire city road network. This throughput guarantee is nevertheless achieved by a decentralized control algorithm that depends only on local traffic information and is easy to compute. These mathematical properties suggest high potential for use in Minnesota, but the method?s performance in practice is not well-known. Furthermore, it lacks some practical constraints on signal timing that could cause confusion to drivers. This project conducted methodological improvements and simulation experiments on a calibrated model of 7 intersections in Hennepin County. We modified the theory behind max-pressure control to model first-in-first-out behaviors on lanes shared by multiple turning movements, and to force max-pressure control to follow a signal cycle. After making these significant methodological improvements, we proved that the maximum throughput properties still hold. Then, we calibrated SUMO (Simulation of Urban MObility) microsimulation models of 2 Hennepin County corridors with 7 intersections using signal timing data and 15-minute observed counts, and compared different versions of max-pressure control with existing actuated-coordinated signals. We varied the maximum cycle length and the time step (signal phases can only change once per time step). The performance depended on the control parameters. Overall, for most intersections and demand periods, we were able to find max-pressure control settings that significantly improved over current signal timings. Large reductions in delay (sometimes over 50%) suggested that max-pressure signal timing both achieved higher throughput during peak demand and was more responsive to queues.