Automatic Calibration of the Fundamental Diagram and Empirical Observations on Capacity

We present a method for automated, empirical calibration of freeway traffic flow characteristics. The method uses 5-min flow and density values for a section of freeway and rapidly and reliably estimates key parameters such as free flow speed, capacity, critical density, congestion wave speed and jam density, which are key inputs to many macroscopic traffic simulation models. The method consists of data filtering, capacity identification, and approximate quantile regression steps. The method was used to calibrate a cell transmission model of Interstate-880 in San Francisco Bay Area, California, a 40-mile long urban freeway with lots of recurrent and nonrecurrent congestion and with dozens of loop detector stations. The calibrated model reproduced the observed traffic congestion behavior within 9% error for performance measures VMT (vehicle miles traveled), VHT (vehicle hours traveled) and total flow. Also, the empirical results suggest that capacity, defined as the maximum observed 5-minute flow rate over several days, differs from breakdown flow, defined as the flow that is observed just before the freeway section becomes congested.