Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference - May 19, 2009 Presented by: Jaesup Lee, Virginia Department of Transportation Dean Munn, The Corradino Group Outline • Introduction • Traffic Data used in basic analysis and model estimation • Data Development and Definition – Free Flow Speed – Traffic Flow (Uninterrupted vs. Interrupted) – Link Capacity • Various Curve Fittings by Functional Class • Findings and Further Research Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Introduction Project Goals • Use empirical data obtained from Virginia facilities to evaluate speed-flow relationships • Test various volume-delay functional forms for each facility type and determine which provides the best performance • Calibrate volume delay function parameters for each facility type • Outcome should be suitable for implementation in Virginia urban travel models Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Introduction Fundamental Elements of Volume-Delay Estimation • Converting spot speeds to space-mean speed • Characteristics of free-flow • Identifying boundary between uninterrupted and interrupted flow • Using knowledge of this boundary to estimate the maximum sustainable flow rate (Capacity) • Use empirical observations to fit VDF curve parameters Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Traffic Monitoring System (TSM) Data • 5,848 locations from 17,400 detector locations available • Three locations per classification selected Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Traffic Data • Data records are a summary of each 15 minute period • Speed bins are in 5 mph increments • Data records are organized by lane and vehicle class LINKID 010148 010148 010148 010148 010148 010148 010148 010148 010148 010148 010148 COUNTERNUM LANE 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 1 1 1 2 2 STARTDATE VEHICLEB_1 VEHICLEB_2 VEHICLEB_3 INTERVAL SPEEDBINLA BIN01 9/1/2007 6 1 1 15 65 9/1/2007 6 1 2 15 65 9/1/2007 6 1 3 15 65 9/1/2007 6 1 1 15 65 9/1/2007 6 1 2 15 65 9/1/2007 6 1 3 15 65 9/1/2007 6 1 1 15 65 9/1/2007 6 1 2 15 65 9/1/2007 6 1 3 15 65 9/1/2007 6 1 1 15 65 9/1/2007 6 1 2 15 65 BIN02 0 0 0 0 0 0 0 0 0 0 0 BIN03 0 0 0 0 0 0 0 0 0 0 0 • CUBE/Voyager script simplifies data • Spot speeds are converted to Space Mean Speeds Filed TIME_PER TOT_AUTOS TOT_TRUCK TOT_VEH LANES FRATE BLANK1 AUTO1 TRUCK1 BOTH1 FDEN1 BLANK2 AUTO2 TRUCK2 BOTH2 FDEN2 Description Time Period Total number of Autos Total number of Trucks Total Vehicles Maximum number of lanes Flow rate Separator column Harmonic Space mean speed for autos Harmonic Space mean speed for trucks Harmonic Space mean speed for all vehicles Flow Density for all vehicles with harmonic space mean speed Separator column Space mean speed for autos ( as per the formula from paper) Space mean speed for trucks ( as per the formula from paper) Space mean speed for all vehicles ( as per the formula from paper) Flow Density for all vehicles for paper space mean speed Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 BIN04 0 0 0 0 0 0 0 0 0 0 0 BIN05 0 0 0 0 0 0 0 0 0 0 0 BIN06 0 0 0 0 0 0 0 0 0 0 0 BIN07 0 0 0 0 0 0 0 0 0 0 0 BIN 1 0 2 0 0 0 0 0 1 0 0 Processed Data Speed vs Density Speed by Time of Day 600 Lane1 Lane2 Lane3 500 400 300 200 100 0 0 5 10 15 20 25 10 15 20 25 Speed vs Flow 80 70 60 50 40 30 Lane1 Lane2 Lane3 20 10 0 Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 5 Estimating Free Flow Speed 1. HCM recommends using mean value for low volume conditions 2. Standard practice also includes using 85th Percentile speed 73.7 mph 71.1 mph HCM Method Low Volume Conditions Observed Speed Range Standard Min Max Mean 55.5 75.2 71.1 Alternative Method All Flow Conditions Deviation 85th Percentile 1.8 73.0 Observed Speed Range Min 5.3 Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Max Mean 76.0 71.1 Standard Deviation 85th Percentile 6.5 73.7 Defining Interrupted Flow 1. Plots of flow vs density and speed vs density show two flow states 2. Others have defined the transition point as the maximum flow or the density at maximum speed, but this is not representative of typical conditions 3. Statistical techniques can define the transition between the two states 3000 2500 Flow (Veh/Hr/Ln) 2000 1500 Interrupted Flow Interrupted Flow 1000 500 0 0.00 20.00 40.00 60.00 80.00 100.00 120.00 Density (Veh/Mi/Ln) Flow vs. Density Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Speed vs. Density Defining Interrupted Flow We define interrupted flow as: Any speed below the threshold where there is 0.0001% probability that it is the same as freeflow. Speed Distribution 4000 Observed 3500 Freeflow Normal Distribution Frequency 3000 2500 For our Rural Freeway example: The computed threshold is 62.58 MPH 2000 1500 1000 500 0 50 55 60 65 Observed Speed Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 70 75 80 Defining Interrupted Flow Rural Freeway example, with flow states identified 80.0 70.0 60.0 Speed 50.0 Uninterrupted 40.0 Interrupted 30.0 20.0 10.0 0.0 0.00 20.00 40.00 60.00 80.00 Density Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 100.00 120.00 Interrupted Flow Histogram – Percent with Interrupted Flow vs. Flow Density 100.00% 90.00% Probability Interrupted Flow Flow withofInterrupted Percent 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 2. 50 7. 50 12 .5 0 17 .5 0 22 .5 0 27 .5 0 32 .5 0 37 .5 0 42 .5 0 47 .5 0 52 .5 0 57 .5 0 62 .5 0 67 .5 0 72 .5 0 77 .5 0 82 .5 0 87 .5 0 92 .5 0 97 .5 0 10 2. 50 0.00% Flow Density Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Estimating Capacity 1. Our data shows a classic logistic distribution 2. We estimated parameters (using density as the only variable) to create a probability function that best fits the data PI = 1/[1 + e(b D+ b )] , where D = Density (veh/mi) 1 0 3. Capacity corresponds to flow density with a 50% probability of being interrupted Probability Function 120.00% 120.00% Observed Frequency 80.00% 80.00% 60.00% 60.00% 40.00% 40.00% 20.00% 20.00% 0.00% 0.00% Flow Density Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Freqency 100.00% Predicted by Probability Function 2. 50 7. 5 12 0 .5 17 0 .5 22 0 .5 27 0 .5 32 0 .5 37 0 .5 42 0 .5 47 0 .5 52 0 .5 57 0 .5 62 0 .5 67 0 .5 72 0 .5 77 0 .5 82 0 .5 87 0 .5 92 0 .5 97 0 10 .50 2. 50 Probability 100.00% Capacity Estimates 0.50% Probability of Interrupted Flow 50.0% Probability of Interrupted Flow 99.5% Probability of Interrupted Flow 3000 Mean Flow (Observed) 2500 2000 1500 1000 500 0 0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 Flow Density This example gives a 39.7 pc/mi Density Threshold or a 2384 pc/hr Max Flow Rate Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 90.00 100.00 Fitting Volume Delay Functions Volume-Delay Functions - Using the computed capacity, the following volume delay functions were estimated based on speeds during uninterrupted flow BPR: Conical: Akcelik: R = R0[1 + a(V/C)^b] R=R0[2 + sqrt(a2(1-V/C) + b2) –a(1-V/C) – b] R=R0+D0+0.25T[(V/C-1)+sqrt{(V/C-1)2+(16J(V/C)L2)/T2}] Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Fitting Volume Delay Functions • Curve Fitting - non-linear regression • Goodness of Fit – – – R-squared Root Mean Square Error Non-Parametric tests e.g. Chi-Square • Other Criteria - suitability for model applications Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Fitting Volume Delay Functions BPR Conical Akcelik Urban Interstate Alpha 0.24 34.5 1.11E-04 J Beta R-squared 3.82 0.947 1.0 0.954 1.0 0.957 T RMSE 6.017% 9.489% 8.787% 80.0 Observed Traffic 70.0 BPR Function Conical Function 60.0 Akcelik Function Speed 50.0 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions BPR Conical Akcelik Rural Interstate Alpha 0.15 78.9 4.89E-05 J Beta R-squared 4.18 0.790 1.0 0.790 1.0 0.812 T RMSE 4.491% 8.381% 7.376% 80.0 Observed Traffic 70.0 BPR Function Conical Function 60.0 Akcelik Function Speed 50.0 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions BPR Conical Akcelik Urban Expressway Alpha 0.18 49.1 8.41E-05 J Beta R-squared 4.63 0.666 1.0 0.625 1.0 0.646 T RMSE 4.162% 6.430% 5.770% 80.0 Observed Traffic 70.0 BPR Function Conical Function 60.0 Akcelik Function Speed 50.0 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions BPR Conical Akcelik Rural Principal Arterial Alpha 0.69 3.1 1.63E-03 J Beta R-squared 1.68 0.880 1.2 0.862 1.0 0.780 T RMSE 5.674% 6.442% 10.362% 60.0 Observed Traffic BPR Function 50.0 Conical Function Akcelik Function Speed 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions Urban Other Principal Arterials BPR Conical Akcelik Alpha 0.24 19.0 2.42E-04 J Beta R-squared 2.86 0.874 1.0 0.770 1.0 0.784 T RMSE 4.602% 14.797% 13.860% 60.0 Observed Traffic BPR Function 50.0 Conical Function Akcelik Function Speed 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions BPR Conical Akcelik Rural Minor Arterial Alpha 0.14 19.2 2.30E-04 J Beta R-squared 1.51 0.248 1.0 0.243 1.0 0.245 T RMSE 4.752% 5.348% 5.282% 70.0 Observed Traffic 60.0 BPR Function Conical Function Speed 50.0 Akcelik Function 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions BPR Conical Akcelik Urban Minor Arterial Alpha 0.16 57.7 1.00E-04 J Beta R-squared 4.53 0.475 1.0 0.521 1.0 0.545 T RMSE 4.270% 8.560% 6.880% 60.0 Observed Traffic BPR Function 50.0 Conical Function Akcelik Function Speed 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions BPR Conical Akcelik Rural Collector Alpha 0.35 7.0 8.28E-04 J Beta R-squared 1.94 0.172 1.1 0.200 1.0 0.203 T RMSE 9.346% 9.186% 9.196% 60.0 Observed Traffic BPR Function 50.0 Conical Function Akcelik Function Speed 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions Alpha 0.21 28.1 1.79E-04 J BPR Conical Akcelik Urban Collector Beta R-squared 3.06 0.566 1.0 0.549 1.0 0.569 T RMSE 5.345% 8.234% 7.419% 70.0 Observed Traffic 60.0 BPR Function Conical Function Speed 50.0 Akcelik Function 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions BPR Conical Akcelik Rural Local Alpha 0.26 5.3 1.15E-03 J Beta R-squared 1.00 0.050 1.1 0.074 1.0 0.077 T RMSE 10.611% 11.155% 11.316% 80.0 Observed Traffic 70.0 BPR Function Conical Function 60.0 Akcelik Function Speed 50.0 40.0 30.0 20.0 10.0 0.0 0.00 0.50 1.00 1.50 Density Ratio Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 2.00 2.50 Fitting Volume Delay Functions Summary of calibrated inputs to VDF fitting process Freeflow Speed Miles per Hour Functional Classification Rural Interstate Rural Other Principal Arterial Rural Minor Arterial Rural Major Collector Rural Local Urban Interstate Urban Freeways and Expressways Urban Other Principal Arterials Urban Minor Arterial Urban Collector Low 70.3 49.0 41.6 34.1 28.4 63.6 46.7 45.2 39.6 40.6 Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 High 73.7 51.5 57.8 47.8 45.7 66.0 65.4 52.9 51.9 51.2 VDF Inputs Capacity Flow Rate Flow Density Hourly PCE/Lane Hourly PCE/Mile Low Low 1747 980 922 450 368 1442 810 888 752 669 High 2384 1130 1188 589 377 1710 1981 1291 1181 853 28.3 23.2 19.1 17.0 11.9 29.5 14.2 25.1 18.9 17.4 High 39.9 41.3 35.9 25.5 19.3 32.7 36.6 29.7 28.4 28.0 Fitting Volume Delay Functions Summarized results from VDF fitting process BPR Functional Classification Rural Interstate Rural Other Principal Arterial Rural Minor Arterial Rural Major Collector Rural Local Urban Interstate Urban Freeways and Expressways Urban Other Principal Arterials Urban Minor Arterial Urban Collector Alpha Low High 0.15 0.25 0.11 0.35 0.26 0.16 0.10 0.24 0.16 0.08 Volume Delay Functions Conical Beta Alpha Low High Low High 0.15 0.69 0.14 0.37 0.41 0.24 0.42 0.28 0.29 0.35 Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 4.2 1.7 1.5 1.4 1.0 3.3 1.9 1.9 2.0 0.4 5.0 3.2 1.9 1.9 1.9 3.8 8.7 2.9 4.5 3.5 78.9 3.1 19.2 5.3 5.3 29.2 6.6 14.8 15.6 12.7 178.9 29.4 23.9 7.0 6.0 57.6 89.0 19.0 57.7 28.1 Akcelik J Low 2.5E-05 1.8E-04 2.3E-04 8.3E-04 1.2E-03 7.5E-05 5.2E-05 2.4E-04 1.0E-04 1.8E-04 High 4.9E-05 1.6E-03 2.7E-04 1.4E-03 1.4E-03 1.4E-04 7.1E-04 3.6E-04 4.0E-04 5.2E-04 Initial Findings • Standard VDF functions are all capable of performing adequately across road classes • For a given road class, VDF parameters fitted for one location, seem to be transferable to other locations • Goodness of fit measures do not strongly differentiate between functions • The Akcelik function, with its more rigorous theoretical underpinnings, seems to work very well Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Next Steps • Additional facility types • Check model transferability to other facilities • Compare HCM capacity, planning capacity, and empirical capacity • Continue to automate analysis process • Test functions in urban models (assignment convergence, average travel speeds) • New VDF functional forms and calibrated parameters will become part VDOT modeling standards Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009 Q&A Thank you ! Contact points Jaesup Lee: jaesup.lee@VDOT.Virginia.gov Dean Munn: dmunn@corradino.com Jeremy Raw : jeremy.raw@VDOT.Virginia.gov Investigation of Speed-Flow Relations and Estimation of Volume Delay Functions for Travel Demand Models in Virginia TRB Planning Applications Conference, May 19, 2009