ALTERNATIVE TRAFFIC ASSIGNMENT METHODS FRAMEWORK REPORT

ALTERNATIVE TRAFFICASSIGNMENT METHODSFRAMEWORK REPORTOregon Department of TransportationTransportation Planning Analysis UnitMay 2018

TECHNICAL REPORT #1ALTERNATIVE TRAFFIC ASSIGNMENT METHODS FRAMEWORK REPORTPrepared by:DKS Associates, Inc.Study Working Group:Alex Bettinardi – ODOT TPAUBrian Dunn – ODOT TPAUPeter Schuytema ‐ ODOT TPAUDoug Norval ‐ ODOT TPAUChristi McDaniel‐Wilson – ODOT TPAUJin Ren Jinxiang ‐ ODOT TPAUChi Mai – ODOT Region 1Ray Jackson – Mid‐Willamette Council of GovernmentsOregon Metro – Peter BosaTyler Deke – Bend Metropolitan Planning OrganizationCopyright @2018 by the Oregon Department of Transportation. Permission is given toquote and reproduce parts of this document if credit is given to the source. This projectwas funded in part by the Federal Highway Administration, U.S. Department ofTransportation.For more information, contact:Alex BettinardiTransportation Modeling Program ManagerOregon Department of Transportation555 13th Street NESalem, Oregon 97301‐4178Telephone: 503.986.4104E‐mail: alexander.o.bettinardi@odot.state.or.us

iContents1.0EXECUTIVE SUMMARY . 11.1ASSIGNMENT METHODS . 11.2EVALUATION OF ASSIGNMENT METHODS. 21.3ASSIGNMENT‐RELATED TOPICS . 4Static, Macroscopic Assignment Topics . 4Dynamic, Macroscopic Assignment Topics. 6Dynamic, Microscopic/Mesoscopic Assignment Topics . 6General Assignment Topics .72.0INTRODUCTION. 93.0ASSIGNMENT METHODS .113.1ASSIGNMENT MODEL COMPONENTS . 11Route Choice Models .11Network Flow Models .123.2CATEGORIES OF ASSIGNMENT METHODS . 13Static, Macroscopic Assignment . 14Dynamic, Macroscopic Assignment. 15Dynamic, Microscopic/Mesoscopic Assignment . 164.0EVALUATION OF ASSIGNMENT METHODS .184.1ASSIGNMENT METHOD OBJECTIVES AND EVALUATION CRITERIA . 184.2ASSIGNMENT METHOD EVALUATION RESULTS . 215.0ASSIGNMENT‐RELATED TOPICS .245.1STATIC, MACROSCOPIC ASSIGNMENT TOPICS . 24Volume Delay Functions .24Generalized Cost Assignment . 33Use of Demand Adjustment Procedures. 345.2DYNAMIC, MACROSCOPIC ASSIGNMENT TOPICS. 36Calibration of DTA Models .36Level of Network Disaggregation . 37

iiLevel of Time Resolution.375.3DYNAMIC, MICROSCOPIC/MESOSCOPIC ASSIGNMENT TOPICS . 38Calibration of Microsimulation Models . 38Size of the Modeling Problem . 405.4GENERAL ASSIGNMENT TOPICS . 42Development of Multi‐Resolution Modeling Networks . 42Consideration of Travel Time Reliability in Traffic Assignment . 44Cost Effectiveness of Dynamic vs. Static Assignment Methods . 456.0CONCLUSIONS .47GLOSSARY .50APPENDIX A .52APPENDIX B .62

iiiList of FiguresFigure 1Acyclic and Cyclic Networks . 15Figure 2Alternative VDFs . 25Figure 3Florida Speed Change vs. Congestion Level. 30

ivList of TablesTable 1Equilibrium Assignment Software Packages . 17Table 2Assignment Method Objectives and Evaluation Criteria . 18Table 3Assignment Method Scoring Summary . 22Table 4Estimated Florida Free‐Flow Speeds and Capacities . 28Table 5Estimate Florida VDF Coefficients . 29Table 6General Applicability of Microsimulation Approaches . 40Table A‐1 Assignment Method Evaluation ResultsTable B‐1 Summary of Assignment Methods Evaluation for Southern Oregon ABMTable B‐2 Results of Assignment Method Evaluation for Southern Oregon ABM

SECTION 1. EXECUTIVE SUMMARY11.0 Executive SummaryThe purpose of the Alternative Traffic Assignment Methods Study was to investigate alternativetraffic assignment methods that can be used for various applications in ODOT TPAU’s transportmodels and those of its OMSC partner agencies. The investigation was limited to trafficassignment only, and did not include transit assignment or other model components.The study consisted of the following main steps: Step 1 – Identify Alternative Assignment Methods Step 2 – Evaluate Assignment Methods Step 3 – Investigate Assignment‐Related TopicsInput was obtained from the study working group on the assignment method objectives, thealternative methods to be evaluated and the evaluation criteria to be used, and assignment‐related topics. The group consisted of staff from ODOT TPAU, ODOT Region 1, Oregon Metro,the Mid‐Willamette Valley COG, and the Bend MPO.1.1Assignment MethodsThe primary components of traffic assignment models are a route choice model and a networkflow model. A route choice model determines the trip‐maker’s path selection between originand destination zones. These models can be classified according to the method of route choice(deterministic or stochastic), the treatment of time (static or dynamic), and the level of vehicleaggregation (microscopic, macroscopic, and mesoscopic). Network flow models define howlinks and nodes perform under congested conditions. Static network flow models assume thatroute flows propagate instantaneously through the network. Dynamic flow models move trafficthrough the network in time slices or distinctive time periods. Microscopic flow modelsconsider vehicles separately and simulate specific characteristics of driver behavior.Mesoscopic flow models are hybrids of macroscopic and microscopic models.Several of the more frequently found combinations of these model dimensions are:

SECTION 1. EXECUTIVE SUMMARY Static, macroscopic Dynamic, macroscopic Dynamic, microscopic or mesoscopic2All of these are equilibrium assignment methods in which demand is distributed according toWardrop's first principle which states, “Every road user selects his route in such a way that thetravel time on all alternative routes is the same, and that switching to a different route wouldincrease personal travel time."1.2Evaluation of Assignment MethodsThe assignment method objectives defined by the study working group describe the desiredproperties and capabilities of the assignment methods as well as the types of applications themethods are to be used for. Evaluation criteria were defined that would allow the objectives tobe reflected in the comparison of the alternative methods. The criteria were applied to thethree assignment methods listed above: Method 1 – Static, macroscopic assignment Method 2 – Dynamic, macroscopic assignment Method 3 – Dynamic, microscopic/mesoscopic assignmentRatings of “low”, “medium”, or “high” were assigned for each criterion.The evaluation was not intended to rank the methods to identify a “best” method, but rather toto establish a general framework for considering the advantages and disadvantages of themethods. If there is interest in determining which method may be the most appropriate for aparticular urban area, travel demand forecasting model, or model application, the methods canbe quantitatively ranked by using a combination of weights and numeric scores for the criteriato develop a weighted total score for each method. A summary of the evaluation results isshown below.

SECTION 1. EXECUTIVE SUMMARY3Assignment Method Rating SummaryRatingMethod 1Static, MacroMethod 2Dynamic, MacroMethod 3Dynamic, umHighTotal No. of CriteriaReading across the diagonal of the table, it can be seen that overall, Method 1 received thehighest number of “low” ratings, Method 2 had the highest number of “medium” ratings, andMethod 3 received the largest number of “high” ratings. The main reasons that Methods 2 and3 rated relatively well compared to Method 1 are: Better accuracy of link traffic flows; Better representation of travel times/speeds; More realistic estimation of intersection delay; Higher levels of temporal resolution; More realistic representation of traffic operations characteristics; Ability to represent peak spreading; More accurate estimation of the effects of capacity improvements; Ability to reflect the effects of small‐scale improvements, such as TSMO‐typeimprovements; and Better ability to support project selection.These advantages are primarily related to the more realistic representation of networkresponse to congestion (in terms of delay), the higher level of temporal resolution and, in thecase of Method 3, the higher network resolution. The higher overall rating for Method 3compared Method 2 is also largely accounted for by the greater advantages of Method 3 inthese areas.

SECTION 1. EXECUTIVE SUMMARY4There are several criteria, however, for which Methods 2 and 3 were rated lower than Method1. This is related to the following disadvantages of Methods 2 and 3: Larger implementation and maintenance time requirements; Larger data collection time requirement; Less intuitive understanding of the assignment method processes; Greater difficulty in interpreting the cause‐effect relationships in the assignmentoutputs; Higher level of staff expertise for application and maintenance; and Lower degree of assignment convergence.All of these disadvantages, except the last one, are due to the greater complexity of Methods 2and 3. The lower degree of assignment convergence is related to the constrained physicalcapacities used in the these methods compared to the continuous VDFs used in Method 1,which result in the spillover of traffic to adjacent time periods if capacity is exceeded. Also withMethods 2 and 3, fractional vehicles are used in the assignment which reduces the level ofconvergence possible compared to Method 1, which uses whole vehicles.1.3Assignment‐Related TopicsSpecific topics related to the implementation and use of the assignment methods wereidentified for investigation by the study working group. The focus of the investigation was onstatic assignment topics because this will likely be the main method used by ODOT and itspartner agencies within the near‐term. Brief highlights for each topic are summarized below.Static, Macroscopic Assignment TopicsAlternative Forms of Volume Delay Functions (VDFs) ‐ The most commonly used VDFs are theBureau of Public Roads (BPR) function, Davidson’s delay model, the Akcelik function, and theconical delay model.

SECTION 1. EXECUTIVE SUMMARY5Incorporation of Node‐Based Delay in VDFs – Potential advantages of VDFs that include anintersection delay component are more accurate estimation of link traffic flows and traveltimes/speeds, more accurate estimation of intersection delay, and better representation of theeffects of capacity improvements.Calibration of VDFs – In a study conducted by Florida A&M University ‐ Florida State University,traffic volume and speed data from the Florida Department of Transportation’s trafficmonitoring stations and statewide transportation engineering warehouse were used tocalibrate four types of VDFs across four facility types and three area types.Representation of Truck Volumes in VDFs ‐ The travel time functions of trucks and cars are notidentical, and furthermore depend on not only traffic volume, but traffic composition as well.Thus, there is a need for travel time functions that consider both the volume and proportion oftrucks in the traffic stream.Representation of Network Capacity ‐ The treatment of network capacity varies according toassignment method used. Capacities can be more realistically represented with the dynamic,macroscopic method, providing more accurate and detailed information about capacityimprovements compared to the static, macroscopic method. Capacities are an output, not aninput, of dynamic, microscopic/mesoscopic models.Generalized Cost AssignmentGeneralized cost assignment attempts to more realistically represent the traveler’s pathdecision‐making process by including other factors in addition to travel time. Since thegeneralized cost for a specific link must be expressed as a single value, all of the factors notmeasured in monetary units must be converted to a constant dollar amount.