Use Big Data and Modeling Tools To Decipher Traffic Patterns: Case Studies in Virginia

Agenda▪ Need for O-D Data

▪ O-D Data Collection – A Primer

▪ Big Data and StreetLight Data

▪ Use Modeling Tool for O-D Estimation

▪ Applications in Transportation Planning

▪ Lessons Learned and Future Research

2

Performance Metrics Why of Interest?

Point-to-point traffic flow patterns

Corridor studiesInform design decisions

Zonal activities Travel shed analysis

Select link analysisDetermine origins and destinations

3

Need for O-D Data

Field-Collected

Data

Bluetooth, License Plate Recognition

Customized approach, expensive

limited samples

Regional Model

Less validation on routes and peak

hours

Good for developing growth rates

Big DataStreetLight,

AirSage, TomTom

Covers longer time periods, large data

set, analytics

O-D Data Collection

4

StreetLightData

AirSage BluetoothLicense Plate

Survey

Type of Devices

Navigation GPS

Cell Phone Bluetooth,wifi

ALPR technology

SpatialPrecision

Great Less Great Great

WeaknessPotential data bias

Less precision for routes, cost

Potential data bias, cost

Most costly

StrengthLarge data set, flexible, low cost

Large data setStrong sample rate

Large sample rate (accurate representation)

Best Uses

O-Ds for zonesand routes,select link analysis

Zonal activities (O-Ds for travel sheds)

O-D routes, speed

O-D routes5

O-D Data Comparison

Use Navigational GPS (INRIX) and Location-Based Services Data

Process geospatial data

points with contextual info

1Normalize the

trips and aggregate the

trips into analytics

2

Users apply analytics to

infer key trends

3

6

StreetLight Data

Integrate Big Data with Modeling Tools▪ Validation, monitoring, and prediction

▪ Regional travel demand models• O-D and speed validation

• External and through traffic estimation

• Freight analysis

▪ Traffic simulation and mesoscopic models• Route level O-Ds

• Travel times and speeds

7

Overview of Modeling Process

Regional Model

Subarea or Corridor Model

Corridor or Intersection Simulation

TransCADCube

VISUM

VISSIMCORSIM

SimTraffic

8

VISUMDynameqDTALite

VISUM ODME Process

▪ Matrix correction: adjusting a demand matrix so that assignment results match target conditions

• Balanced Link Volumes

• Trip Length Distribution

Demand Matrix

O-D Data Source• StreetLight Data

AssignmentTFlowFuzzy

O-D Routes

Export to VISSIM

VISUM

No

t C

on

verg

edConverged

9ODME: Origin-Destination Matrix Estimation

Case Studies of O-D Data Applications in Planning

▪ Freeway Corridor O-D Estimation (I-95 Corridor, Chesterfield County and Richmond)

▪ Traffic Flow Patterns at a Weaving Area (STARS Route 7, Loudoun County)

▪ Understand Traffic Patterns (Fairfax County)

10

Case 1: I-95 Corridor

▪ I-95 in Chesterfield County and Richmond, VA

▪ 13 miles, 9 interchanges, 55 ramps, and 23 intersections (including 2 IMRs)

▪ Purpose and need• Understand existing traffic

operations

• Use model for existing and future studies

11

StreetLight Data▪ Users establish desired origin and

destination zones to evaluate travel patterns

Origin Zones Destination Zones

12

StreetLight Data - Outputs▪ Downloadable OD tables for personal (cars)

and commercial (trucks) trip index

▪ Data example• 13-mile segment of I-95; 7 interchanges

• 2015 data; average weekday (M-Th)

How do we manipulate this data to get O-Ds for peak hour trips?

20,444data points!

13

StreetLight Data Processing

▪ Pivot tableSum of Streetlight

Trip Frequency VISUM Destination Zone

VISUM Origin Zone 12 51 52 53 54 57 58 91 93 95 97 99 102 103 105 107 109 114 115 116 119 122 126 127 129 Grand Total

1 2 2 917 2 2 925

2 39 4 6 2 6 6 11 2 9 4 146 4 6 245

3 533 2 152 8 11 2 555 54 4 4 4 6 2 45 2 13 400 34 116 1947

4 2 9 221 2 2 2 238

7 9 41 2 2 201 9 13 277

8 69 2 2 6 8 9 6 43 43 11 308 28 11 546

41 602 2 8 150 212 240 2 137 8 885 114 2 56 6 109 9 101 2062 272 186 5163

94 2 2 2 154 160

96 2 2 77 81

98 4 4 2 38 4 15 2 69

100 9 6 6 17 9 2 4 11 4 2 8 111 103 4 296

101 9 28 707 2 4 49 2 2294 2 2 452 156 2 107 2 206 2 8 722 26 9 4791

104 109 2 4 2 117

106 171 2 2 2 2 2 181

108 4 11 21 8 195 6 6 2 9 6 62 2 143 4 182 6 667

110 4 4 36 8 9 4 11 13 4 2 107 6 4 212

111 84 358 2 2 135 1765 2 184 4 2 4 154 370 972 9 212 2 66 1563 90 15 5995

113 1090 9 13 60 148 2 39 77 36 4 2 15 13 9 1517

117 2 2 4 2 9 306 8 24 357

118 56 2 53 26 129 2 9 8 21 8 6 9 2 4 335

120 6 2 11 4 41 2 8 4 8 6 15 15 4 126

121 4 6 2 548 8 15 583

125 576 148 328 4 2 4 216 2009 2 45 133 163 113 2 398 8 660 206 454 5471

128 8 2 9 2 58 2 4 2 11 4 6 2 154 264

130 6 11 47 19 154 2 111 38 75 2 30 1128 6 1629

Grand Total 3299 324 1665 162 226 480 492 8217 160 266 452 158 1904 555 259 162 388 1293 516 802 734 230 7790 797 861 32192

Destination ZonesOriginZones

Trip Index of Each O-D Pair

14

StreetLight Data – Overall ProcessCreate trip index O-D matrix based on raw StreetLight data

Review matrix for feasibility and reasonableness (remove unlikely O-D pairs)

Convert to percentage O-D table (% of each origin to each destination)

Combine with target volumes at origins

Import seeding table into VISUM

Complete process separately for car and truck results

15

O-D Matrix Reasonableness

▪ Traffic volume targets

▪ Local travel patterns

▪ Trip length distributions

Trip Length (miles)

Trip

%

Seeding Matrix Distribution

Final Matrix Distribution16

I-95 Through Traffic

13-14 miles

ChippenhamParkway to North

and South7-8 miles

Between Interchanges

and Route 288 to South1-2 miles

Case 2: Route 7 Weaving –Loudoun County and Leesburg▪ Heavy traffic congestion at weave area

▪ >2 mile queue on northbound Dulles Greenway (Route 267) in PM

17

Reasonableness Check

SL: 21%Counts: 10%

SL: 79%Counts: 90%

18

Lessons Learned

▪ Capture rate• ALPR: 40% to 60%

• Bluetooth: 8%

• StreetLight Data: to be determined

▪ Data results more consistent between ALPR and Bluetooth, but comparable among the three methods

▪ Evaluation underway by VDOT/VRTC

▪ A solid understanding of weaving patterns helps identify issues

19

Case 3: Study Traffic Patterns in Fairfax County▪ Cut-through traffic

▪ Origin-Midpoint-Destination (O-M-D) from StreetLight Data

▪ Visualize select link analysis

20

OriginMidpoint

Destination

StreetLight Midpoint Filter Zone

Midpoint Filters

▪ Allows for evaluation of traffic movements through specific intersections or segments• Directional

• O-M-D for each feasible combination

21

Integrate StreetLight Data to VISUM

22

▪ Select link analysis example with percentage distribution from midpoint zone of interest

23

Fine-Tune Visualization in GIS

Summary and Future Research▪ Understand the strengths and

weakness/limitations of each data type• Reasonableness check

• Defendable process (qualitative vs. quantitative)

▪ At scoping meeting, establish data needs• Field data

• Supplemented by big data

• Approvable assumptions

▪ Decipher the results

▪ Big data is evolving

24

Questions?

Contact Us

MARK HERMAN, EITmark.herman@kimley-horn.com

(804) 672-4716

JIAXIN TONG, P.E.jiaxin.tong@kimley-horn.com

(703) 870-3601

25