Office of Operations
21st Century Operations Using 21st Century Technologies

Traffic Analysis Toolbox Volume VI:
Definition, Interpretation, and Calculation of
Traffic Analysis Tools Measures of Effectiveness

7.0 Practical Application Results

The purpose of the practical application task is to demonstrate through case studies the validity of the approach(s) recommended in the previous chapter. The intent is to use existing real-world data to demonstrate how transportation professionals should generate and interpret the basic system MOEs recommended by this study.

7.1 Selection of Case Studies

The original hypothesis of this research was that a set of procedures could be developed to convert the MOEs reported for typical microscopic and macroscopic traffic analysis tools into a consistent set of MOEs comparable to the Highway Capacity Manual. The investigation into how various tools compute the standard MOEs discovered that there is a wide range of methods used to compute and average the standard MOEs, and these methods diverge at the microscopic level. The reports produced by typical traffic analysis tools do not provide sufficient microscopic information to enable the analyst to convert the MOEs produced by one tool into the equivalent MOEs produced by another tool.

Consequently, the recommended approach for achieving consistent MOE results across various analytical tools is to compute the MOEs directly from vehicle trajectory data produced by the tool, rather than trying to correct aggregate MOE results reported by each microscopic or macroscopic tool.

One advantage of this approach is that vehicle trajectory data is the lowest common denominator shared by all microscopic and mesoscopic tools. A single procedure for computing MOEs directly from vehicle trajectory data is applicable to all micro and mesoscopic tools. The procedures identified in this report can be automated today through the use of commonly available analytical tools (ACCESS and EXCEL for example) or in the future through procedures built into each tool to process vehicle trajectory data.

Since none of the macroscopic tools and few of the microscopic tools produce vehicle trajectory data readily accessible to the average user, these cases studies use vehicle trajectory data measured directly in the field. The FHWA NGSIM program has developed 3 datasets of tenth-second vehicle trajectory data. Two of these vehicle trajectory data sets were selected for use in the case studies:

  1. The I-80 Freeway – Emeryville, California; and
  2. Lankershim Boulevard – Los Angeles, California.

7.2 The I-80 Freeway Case Study

The I-80 data set consists of tenth-second vehicle trajectories for a 1,650-foot-long section of the I-80 eastbound freeway between Powell Street and Ashby Avenue in Emeryville, California. The section is a 5-lane plus HOV lane weaving section of freeway with a one lane on ramp merge point for Powell Street about 420 feet into the section. The grade is nominally 0 percent in this section.

Vehicle trajectory data was gathered for three 15-minute periods on the afternoon of April 13, 2005. The 5:15 p.m. through 5:30 p.m. period was selected for the case study as likely to have the most congestion.

Further details on the data set can be found in: Cambridge Systematics, NGSIM I-80 Data Analysis (5:15 p.m. to 5:30 p.m.) Summary Report, December 2005, available on the NGSIM web site (http://ngsim-community.org/). Table 33 defines the data fields in this data set.

Table 33. I-80 Data Dictionary

Field

Name

Definition

Used in Case Study

1

Vehicle ID

Preliminary ID number assigned to each vehicle trajectory during first pass recording trajectories. Several ID numbers dropped during manual reconnection of broken trajectories.

Yes

2

Frame ID

Frame ID Number.

No

3

Total Frames

Number of frames vehicle is visible.

No

4

Global Time

Number of milliseconds after Midnight, January 1, 1970 Greenwich Standard Time.

Yes

5

Local X

Lateral (X) coordinate of the front center of the vehicle with respect to the left-most edge of the section in the direction of travel.

Yes

6

Local Y

Longitudinal (Y) coordinate of the front center of the vehicle with respect to the entry edge of the section in the direction of travel.

Yes

7

Global X

CA State Plane NAD83 coordinate.

No

8

Global Y

CA State Plane NAD83 coordinate.

No

9

Vehicle Length

Feet.

No

10

Vehicle Width

Feet.

No

11

Vehicle Class

1 = motorcycle, 2 = auto, 3 = truck.

Yes

12

Vehicle Velocity

Feet per second.

Yes

13

Vehicle Acceleration

Feet per second squared.

Yes

14

Lane Identification

Lane 1 = farthest left.
Lane 6 = farthest right.
Lane 7 = on-ramp.
Lane 9 = shoulder on the right-side

No
[Although not used in this particular test, it would be handy to have the lane identification in the MOE dataset, so that the difference between HOV and non-HOV lane operation could be evaluated.]

15-18

 

Additional data on leading vehicle, following vehicle, and headway

No

Source: Cambridge Systematics – NGSIM Data Dictionary.

The MOE results for the I-80 data set are shown in Table 34.

Table 34. I-80 Decision-Maker MOEs

Decision-Maker MOEs

Value

Interpretation

Travel Time Index (Trip Time/Ideal Trip Time)

18.69

Severely congested. Quite a bit higher than HCM TTI at capacity. See table below.

Throughput (vph)

6,844

Operating at about 50 percent of capacity for 6-lane freeway.

Freeway MOEs

No value

No value.

Maximum Extent Breakdowns (%/dir.miles)

100%

Severely congested.

Duration of Breakdowns (%/Time)

100%

Severely congested.

Surface Street MOEs

No value

No value.

Maximum Street Segment Overflows

Not applicable

No value.

Maximum Segment Overflows (%/segs)

Not applicable

No value.

Duration Segment Overflows (%/time)

Not applicable

No value.

Maximum Turn Bay Overflows

Not applicable

No value.

Maximum Turn Bay Overflows (%/bays)

Not applicable

No value.

Duration Turn Bay Overflows (%/time)

Not applicable

No value.

A single ACCESS query of the I-80 vehicle trajectory database provided most of the decision-maker MOEs. This query identified the first millisecond and the last millisecond when the vehicle was present anywhere on the system for those milliseconds within the analysis period. The starting and end "Y" position for each vehicle also was identified in the same query.

The query was output to a spreadsheet, which then computed the VMT and VHT for each vehicle. Each vehicle and the VMT and VHT it generated was separated into five categories:

  1. Vehicles already present at the start of the analysis period that were able to exit the system before the end of the analysis period.
  2. Vehicles already present at the start of the analysis period that were still within the system at the end of the analysis period.
  3. Vehicles still present on the system at the end of the analysis period, but which were not present at the start.
  4. Vehicles completely denied entry to the system during the entire analysis period. (The NGSIM datasets do not include these vehicles so they had to be estimated for the purpose of the case study.)
  5. Vehicles successfully entering and exiting the system during the analysis period.

Travel Time Index (Trip Time/Ideal Trip Time)

Computation Notes – The travel time index was computed by taking the ratio of the sum of the VHT traveled for all 5 vehicle categories to the ideal free-flow VHT. The ideal free-flow VHT was computed by dividing the sum of the VMT accumulated by all 5 vehicle categories by the posted speed limit of 65 mph.

The NGSIM dataset does not include vehicles unable to enter the system, so the vehicle hours accumulated by vehicles denied entry to the system had to be estimated. Since the study section had a mean density of close to 100 vehicles per mile per lane over the entire analysis period, it was assumed that the queue extended beyond the start of the study segment for the entire analysis period. Based on personal knowledge of the freeway section, it was estimated that the queue of vehicles waiting to enter the study section extended back about 1 mile and persisted in steady state throughout the entire analysis period. The VHT of delay for vehicles temporarily delayed from entering the study section was therefore estimated at 100 vehicles/mile/lane x 4 lanes x 1 mile x 0.25 hrs = 100 VHT (see Table 35).

Table 35. Adjustment to TTI Computation for Delayed Entry to System (I-80)

Vehicle Group

Component

Value

V1 (vehicles present in system at start of period and successfully exiting before end of period)

No value

138

V2 (vehicles present in system at start of period but unable to exit before end of period)

No value

0

V3 (vehicles entering system during period but unable to exit before end of period)

No value

79

V4 (vehicles denied entry for entire period)

No value

0

V5 (vehicles successfully entering and exiting during the period)

No value

1,573

V (total vehicles)

No value

1,790

Vehicle Miles Traveled 1 (vehicles at start and successfully exiting)

No value

28.25

Vehicle Miles Traveled 2 (vehicles at start unable to exit)

No value

0.00

Vehicle Miles Traveled 3 (vehicles at end, but not at start)

No value

17.90

Vehicle Miles Traveled 4 (vehicles denied entry)

No value

0.00

Vehicle Miles Traveled 5 (vehicles successfully entering and exiting)

No value

461.33

Vehicle Miles Traveled (total vehicles)

No value

507.47

Vehicle Hours Traveled 1 (vehicles at start and successfully exiting)

No value

2.30

Vehicle Hours Traveled 2 (vehicles at start unable to exit)

No value

0.00

Vehicle Hours Traveled 3 (vehicles at end, but not at start)

No value

7.27

Vehicle Hours Traveled 3A (accumulated on Network)

2.49

No value

Vehicle Hours Traveled 3B (accumulated waiting to enter Network) (manual adjustment)

4.78

No value

Vehicle Hours Traveled 4 (accumulated while vehicle denied entry)

No value

0

Vehicle Hours Traveled 5 (vehicles successfully entering and exiting)

No value

136.38

Vehicle Hours Traveled 5A (accumulated on Network)

41.29

No value

Vehicle Hours Traveled 5B (accumulated waiting to enter Network) (manual adjustment)

95.09

No value

Vehicle Hours Traveled (total vehicles)

No value

145.94

Mean Speed (miles per hour) 1 (vehicles at start and successfully exiting)

No value

12.30

Mean Speed (miles per hour) 2 (vehicles at start unable to exit)

No value

0.00

Mean Speed (miles per hour) 3 (vehicles at end, but not at start)

No value

2.46

Mean Speed (miles per hour) 4 (while vehicles denied entry)

No value

0.00

Mean Speed (miles per hour) 5 (vehicles successfully entering and exiting)

No value

3.38

Mean Speed (miles per hour) (total vehicles)

No value

3.48

Note: TTI = FFS/MPH where FFS equals free flow speed and MPH equals adjusted mean speed for all vehicles.

The vehicle hours accumulated for vehicles entering the system during the analysis period (Vehicle categories 3 and 5) were split into two subcategories. The first subcategory represents the VHT accumulated within the study section. The second subcategory represents the estimated VHT accumulated by the same vehicles waiting their turn to enter the study section during the analysis period.

The final TTI is then the ratio of the posted speed limit to the final adjusted mean speed for all vehicles taking into account the delays to vehicles queued waiting to enter the system during the analysis period.

Interpretation – The value of 18.69 for the TTI is very high, quite a bit higher than the metropolitan areawide TTIs reported by the Texas transportation Institute, which in 2005 ranged from 1.05 (Anchorage, Alaska) to 1.75 (Los Angeles, CA) (Source: The 2005 Urban Mobility Report by David Schrank and Tim Lomax of the Texas Transportation Institute, May 2005, http://mobility.tamu.edu).

The TTI also is quite a bit higher than the TTI s at capacity flows computed using the Highway Capacity Manual method for freeways.

Throughput (vph)

Computation Notes – The system throughput is the total number of vehicles able to exit the system during the analysis period, including vehicles present anywhere in the system at the start of the analysis period. The throughput is normalized to an hourly rate by dividing the number of vehicles by the number of hours in the analysis period.

The throughput was the sum of vehicle categories 1 and 5, all vehicles able to exit the system during the analysis period.

Interpretation – Throughput is a solely a function of demand until demand starts approaching capacity. Then it is a function of solely capacity. A throughput of 6,844 vph is equivalent to 1,141 vph/lane for a 6-lane freeway, which is less than 60 percent of the potential capacity of this freeway. The throughput is constrained by a downstream bottleneck. Throughput is useful for comparing alternative improvements.

Freeway MOEs – Maximum Extent Breakdowns (%/dir.miles)

An ACCESS query grouping PCEs present on the segment for each millisecond of the analysis period was exported to an EXCEL spreadsheet. Within the spreadsheet the PCUs per lane-mile were compared to the Highway Capacity Manual density thresholds for LOS "F" on freeways (43 pcu/lane-mile).

The mean density the full length of the analysis period and across all lanes and for the full length of the study section exceeded the LOS "F" density of 43 pcus/lane/mile. Thus the study section reached 100 percent LOS "F."

Freeway MOEs – Duration of Breakdowns (%/Time)

The total milliseconds where PCUs per lane-mile were greater than the Highway Capacity Manual density thresholds for LOS "F" on freeways (43 pcu/lane-mile) was divided by the total milliseconds in the analysis period to obtain the percent duration. Since the duration was 100 percent of the analysis period, the study section is considered severely congested.

Surface Street MOEs – Street Segment Overflows

The surface street MOEs are not applicable to the I-80 data set.

7.3 The Lankershim Case Study Data Set

The Lankershim data set consists of tenth-second vehicle trajectories for a 1,600-foot-long bidirectional section of Lankershim Boulevard between Valley Heart Drive and the U.S. 101 freeway in California. The section includes 4 closely spaced signalized intersections. The arterial has three to four through lanes in each direction with left and right turn bays many of the intersections. The posted speed limit is 35 mph.

Adjacent land uses include: Bus Terminal, Metro station, 36 story office building, Metro station and office building parking lots and garages, and Universal Studios parking lot.

Vehicle trajectory data was available for two 15-minute periods on the morning of June 16, 2005. The 8:30 a.m. to 8:45 a.m. period was selected for the case study as likely to have the most congestion.

Further details on the data set can be found in: Cambridge Systematics, NGSIM I-80 Data Analysis (8:30 a.m. to 8:45 a.m.) Summary Report, March 2006, available on the NGSIM web site (http://ngsim-community.org/). A data dictionary listing the contents of this dataset is provided in Table 36.

Table 36. Lankershim Data Dictionary

Field

Name

Definition

Used in Case Study

1

Vehicle ID

Preliminary ID number assigned to each vehicle trajectory during first pass recording trajectories. Several ID numbers dropped during manual reconnection of broken trajectories.

Yes

2

Frame ID

Frame ID Number.

No

3

Total Frames

Number of frames vehicle is visible.

No

4

Global Time

Number of milliseconds after Midnight, January 1, 1970 Greenwich Standard Time.

Yes

5

Local X

Lateral (X) coordinate of the front center of the vehicle – perpendicular to the median of the Lankershim Boulevard. Vehicles traveling on the east side of the median have positive Local X values, while those traveling on the west side of the median have negative Local X values.

Yes

6

Local Y

Longitudinal (Y) coordinate of the front center of the vehicle along the median of the Lankershim Boulevard. The start point is at the southern boundary of the study area.

Yes

7

Global X

CA State Plane NAD83 coordinate.

No

8

Global Y

CA State Plane NAD83 coordinate.

No

9

Vehicle Length

Feet.

No

10

Vehicle Width

Feet.

No

11

Vehicle Class

1 = motorcycle, 2 = auto, 3 = truck.

Yes

12

Vehicle Velocity

Instantaneous feet per second.

Yes

13

Vehicle Acceleration

Instantaneous feet per second squared.

Yes

14

Lane Identification

Current lane position of vehicle. Lane numbering is incremented from the left-most lane, except for locations where left-turn or right-turn bays exist. Left-turn bays are numbered starting from 11 and are incremented from the left-most left-turn bay. Right-turn bays are numbered starting from 31 and are incremented from the left-most right-turn bay. The left-turn bay mid-block between intersections 3 and 4 is numbered 101.

Yes

15

Origin Zone

No value

No

16

Destination

No value

No

17

Intersection

From south to north, intersection numbers.

Yes

18

Section

From south to north, section number.

Yes

19

Direction

Moving direction of the vehicle. 1 – eastbound (EB), 2 – northbound (NB), 3 – westbound (WB), 4 – southbound (SB).

Yes

20

Movement

Movement of the vehicle. 1 – through (TH), 2 – left-turn (LT), 3 – right-turn (RT).

No

21-24

Movement

Additional data on leading vehicle, following vehicle, and headway.

No

Source: Cambridge Systematics – NGSIM Data Dictionary.

The MOE results for the Lankershim data set are shown in Table 37.

Table 37. Lankershim Decision-Maker MOEs

Decision-Maker MOEs

Value

Interpretation

Travel Time Index (Trip Time/Ideal Trip Time)

2.58

Better than typical for an uncoordinated arterial operating at capacity.

Throughput (vph)

4,212

Useful for comparing the productivity of alternative improvements.

Freeway MOEs

No value

No value

Maximum Extent Breakdowns (%/dir.miles)

Not applicable

Not applicable

Duration of Breakdowns (%/Time)

Not applicable

Not applicable

Surface Street MOEs

No value

No value

Maximum Street Segment Overflows

2

Two segments to investigate.

Maximum Segment Overflows (%/segs)

33%

Significant fraction of system under stress.

Duration Segment Overflows (%/time)

20%

Problems endure longer than peak 15 minutes of hour.

Maximum Turn Bay Overflows

2

Two problem spots to investigate.

Maximum Turn Bay Overflows (%/bays)

40%

Significant fraction of turn bays have problems.

Duration Turn Bay Overflows (%/time)

12%

Greater than design standard of 5 percent implied by 95 Percentile Turn Bay Length design standard.

Decision-Maker MOE Results

A single ACCESS query of the Lankershim vehicle trajectory database provided most of the decision-maker MOEs. This query identified the first millisecond and the last millisecond when the vehicle was present anywhere on the system for those milliseconds within the analysis period. The starting and end "Y" position for each vehicle also was identified in the same query.

The query was output to a spreadsheet, which then computed the VMT and VHT for each vehicle. Each vehicle and the VMT and VHT it generated was separated into five categories:

  1. Vehicles already present at the start of the analysis period that were able to exit the system before the end of the analysis period.
  2. Vehicles already present at the start of the analysis period that were still within the system at the end of the analysis period.
  3. Vehicles still present on the system at the end of the analysis period, but which were not present at the start.
  4. Vehicles completely denied entry to the system during the entire analysis period. (The NGSIM datasets do not include these vehicles so they had to be estimated for the purpose of the case study.)
  5. Vehicles successfully entering and exiting the system during the analysis period.

Since the NGSIM dataset does not include vehicles unable to enter the system, the vehicle hours accumulated by vehicles denied entry to the system had to be estimated. Since the queues in the system did not fill up the street segments very often during the analysis period (less than 12 percent of the time), the vehicle hours of delay for denied entry vehicles was assumed to be zero. No adjustment was necessary.

Travel Time Index (Trip Time/Ideal Trip Time)

Computation Notes – The travel time index was computed by taking the ratio of the sum of the VHT traveled for all 5 vehicle categories to the ideal free-flow VHT. No adjustment to the TTI was necessary for vehicles delayed from entering the system during the analysis period, because queues did not extend beyond the physical limits of the study area. The ideal free-flow VHT was computed by dividing the sum of the VMT accumulated by all 5 vehicle categories by the posted speed limit of 35 mph.

Interpretation – The value of 2.58 for the TTI is quite a bit higher than the metropolitan areawide TTIs reported by the Texas transportation Institute, which in 2005 ranged from 1.05 (Anchorage, Alaska) to 1.75 (Los Angeles, California) (Source: The 2005 Urban Mobility Report by David Schrank and Tim Lomax of the Texas Transportation Institute, May 2005, http://mobility.tamu.edu).

According to the Highway Capacity Manual method the TTI for an urban arterial would range from 1.9 to 2.1 under free-flow uncoordinated conditions and from 2.9 to 3.3 under capacity flow uncoordinated conditions for a Class II/Class III arterial. Since the computed TTI of 2.58 falls below the values at capacity for uncoordinated arterials, the arterial traffic operation is considered to be generally good.

Throughput (vph)

Computation Notes – The system throughput is the total number of vehicles able to exit the system during the analysis period, including vehicles present anywhere in the system at the start of the analysis period. The throughput is normalized to an hourly rate by dividing the number of vehicles by the number of hours in the analysis period.

The throughput was the sum of vehicle categories 1 and 5, all vehicles able to exit the system during the analysis period.

Interpretation – Throughput is solely a function of demand until demand approaches capacity. Then it is a function of solely capacity. A throughput of 4,212 vph is equivalent to 702 vph/lane for a 6-lane arterial, which is slightly below the planning capacity for a signalized intersection (planning capacity would be 750 vph/ln = 1,500 vphgl * 0.50 g/c). Throughput is useful for comparing alternative improvements.

Freeway MOEs – Maximum Extent Breakdowns (%/dir.miles)

This MOE is not applicable to the Lankershim data set.

Freeway MOEs – Duration of Breakdowns (%/Time)

This MOE is not applicable to the Lankershim data set.

Surface Street MOEs – Street Segment Overflows

Computation Notes – A specialized set of segment by segment, direction by direction set of ACCESS queries were required to identify street segments where the queues of vehicles had filled up one or more through lanes on the street.

First a query was conducted to identify the maximum and minimum "Y" coordinates for each vehicle on each direction of each street segment. The results were exported to an EXCEL spreadsheet where a pivot table was used to compute the average maximum and average minimum "Y" coordinates for each direction of each street segment. The average was used because of slight differences in the starting and end "Y" coordinates for each lane of each segment.

The average maximum "Y" was subtracted from average minimum "Y" for each direction of each segment to obtain the mean segment length by direction. Since vehicles that cannot fit within the segment are never recorded, the target queue length for determining if a segment is full was set at 25 feet less than the actual length of the segment. If the queue extends to within 25 feet of the back of the segment, the segment is considered to be full, for all practical purposes.

However, the "Y" coordinates in the data set represent the position of the front of the vehicle, not the back, so the target length of the segment (for determining when the queue filled up the segment) was reduced by a further 25 feet. Thus the target "Y" coordinate for determining if the vehicle queue in any lane had filled up the segment was the "Y" coordinate of the back of the link minus 50 feet.

A separate query was then conducted in ACCESS for each direction of each segment. For each millisecond of the analysis period, if the "Y" coordinate of any vehicle exceeded the maximum "Y" coordinate minus 50 feet for that direction of that segment, and the vehicle was traveling at that moment at a speed of less than 7 fps, with an acceleration of less than +2 fpss, then one or more lanes of that direction of the segment was considered to be "blocked" with queuing vehicles (even if not other queuing vehicle is present in the lane in front of the queued vehicle).

The number of segments "blocked" at any millisecond of the analysis period was tallied to obtain the maximum number of blocked directional segments for the analysis period.

The duration of blocked directional segments within the system during the analysis period was obtained by summing up the number of millisecond snapshots where at least one directional segment was blocked.

Interpretation – The percentage of street segments subject to risk of overflow is relatively high (33 percent), with at least one segment subject to risk of overflow 20 percent of the analysis period. Since standard practice is to design turn bays so that they have less than a 5 percent risk of overflowing during the peak hour, the 20 percent overflow for an individual street segment would be undesirable for a new design.

Surface Street MOEs – Turn Bay Overflows

Computation Notes – Given the tediousness of the approach used to arrive at the segment blockages, an alternative approach was tried to identify turn bay overflows.

An ACCESS query was launched to report all instances of queued vehicles (speeds less than 7 fps and acceleration less than +2 fpss) within all turn bays of the system. The maximum and minimum "Y" coordinate was reported for the queued vehicles each millisecond within the turn bay. The query result was exported to a spreadsheet.

The difference between maximum and minimum "Y" coordinates for each turn bay for each millisecond was compared to the nominal storage length for the turn bay reported in the NGSIM data collection report, less 25 feet (50 feet was not subtracted from the storage length because several turn bays are only 70 feet long. Subtracting 50 feet from the storage length would have resulted in these turn bays being reported as at risk of overflow every time a single vehicle entered the bay).

Note that this method was easier to implement with an ACCESS query, but it only gets the maximum number of vehicles in queue in the lane at any one time, not the maximum back of queue. Consequently, the method described above for identifying street segment backups is preferable because of its superior accuracy at identifying queue overflow problems.

A pivot table was used to tally up the number of milliseconds during the analysis period that each turn bay was at risk of overflowing.

The maximum duration of turn bay overflows was estimated to be the longest duration of overflows for the turn bays. The maximum number of turn bays at risk of overflow at any one time during the analysis period was estimated by visual inspection of the turn bay query results. (It would have been more accurate to plot the turn bay overflows by millisecond through the analysis period and identify the maximum impacted at any one time and the total number of milliseconds of the analysis period when any of the turn bays was at risk of overflowing, however; resource and time limitations caused the author to select these two shortcuts for approximating the maximum duration and extent of turn bay overflows.)

Interpretation – The maximum number of turn bays at risk of overflowing at any one time is quite high (40 percent). Also, the percent of time when one or more turn bays is at risk of overflowing (12 percent) is higher than 5 percent risk that standard practice would call for in the design of new turn bays.

7.4 Conclusions from Case Studies

The above practical applications have demonstrated how vehicle trajectory data can be post-processed to yield key basic measures of effectiveness useful for transportation operations analysis and decision-makers. People more familiar with database processing than the author can no doubt find betters ways to process the trajectory data than were used in these examples. However, the practical applications have illustrated the feasibility of using off-the-shelf database programs to generate MOEs from vehicle trajectory data.