GPS Data Filtration Method for Drive Cycle Analysis Applications

ABSTRACT

Global Positioning System (GPS) data

acquisition devices

have proven useful tools for gathering real-world driving data

and statistics. The data collected by these devices provide

valuable information in studying driving habits and

conditions. When used jointly with vehicle simulation

software, the data are invaluable in analyzing vehicle fuel use

and performance, aiding in the design of more advanced and

efficient vehicle technologies. However, when employing

GPS data acquisition systems to capture vehicle drive-cycle

information, a number of errors often appear in the captured

raw data samples. Common sources of error in GPS data

include sudden signal loss, extraneous or outlying data points,

speed drifting, and signal white noise, all of which combine

to limit the quality of field data for use in downstream

applications. Unaddressed, these errors significantly impact

the reliability of source data and limit the effectiveness of

traditional drive cycle analysis approaches and vehicle

simulation software. Without reliable speed and time

information, the validity of derived metrics for drive cycles,

such as acceleration, power, and distance become

questionable. This study explores some of the common

sources of error present in collected raw GPS data and

presents a detailed filtering process designed to correct for

these issues. To illustrate the effectiveness of the proposed

filtration process across the range of vehicle vocations, test

data from both light- and medium/heavy-duty applications

are examined. Graphical comparisons of raw and filtered

cycles are presented, and statistical analyses performed to

determine the effects of the proposed filtration process on

raw data. Finally, the paper concludes with an evaluation of

the overall benefits of data filtration on raw GPS data and

presents potential areas for continued research.

INTRODUCTION

The cost-effective nature and ease of installation associated

with GPS data acquisition systems (DASs) have aided in

onboard global positioning system (GPS) data logging

rapidly becoming one of the more popular methods for

collecting real-world vehicle operating information [

1, 2, 3,

4]. The coupled vehicle speed-time data captured by these

devices are of particular

interest when performing vehicle

simulation and drive cycle analysis [

5, 6, 7, 8, 9]. However,

due to the sensitivity

of these downstream applications to

both the quality and integrity of GPS source data, the unique

operating behavior and errors inherently associated with GPS

data loggers have fostered a need for a novel filtration

process. The ideal filtration method for improving the quality

of raw GPS data is one that minimizes the effects of GPS data

logging errors such as sudden signal loss, data spiking, signal

white noise, and zero speed drift while maintaining the

integrity of the raw source data [

10, 11, 12]. This means

removing or replacing any

erroneous data points while

simultaneously minimizing the total amount of data altered.

The ideal filtration method must also be flexible enough for

application to GPS data sourced from different logging

devices. As such, any proposed filtration method must take

into account differences in data quality available through

different GPS DASs as well as the different vehicle vocations

or types in which the loggers may be mounted.

The GPS data filtration approach presented here consists of

seven distinct filters designed and arranged specifically to

meet these criteria. In an attempt to account for the common

errors associated with the analysis of GPS speed-time data, a

common GPS data processing method has been developed in

the Matlab environment that employs a series of linearly

progressing logic-based data filters [

13].

GPS Data Filtration Method for Drive Cycle

Analysis Applications

2012-01-0743

Published

04/16/2012

Adam Duran and Matthew Earleywine

National Renewable Energy Laboratory

doi:10.4271/2012-01-0743

NREL/CP-5400-53865. Posted with permission.

Presented at the SAE 2012 World Congress.

To evaluate the effectiveness of the proposed GPS data

filtration process,

National Renewable Energy Laboratory

(NREL) researchers analyzed over 1,700 individual drive

cycles sourced from GPS devices mounted onboard both

light- and medium/heavy-duty vehicles. The drive cycles

analyzed as part of this study contain coupled speed-time

information gathered using onboard GPS DASs, and contain

the driving information for either a single vehicle operating

day or shift. Statistics regarding the amount of data altered

during each filtration step were calculated, and the results of

the filtration process were analyzed. In addition to examining

the magnitude of data altered during the filtration process,

drive-cycle analysis calculations were performed to assess the

effects of filtration on the underlying cycle dynamics.

APPROACH

FILTER LOGIC

The proposed filtration process for analyzing GPS speed-time

data for drive-cycle applications consists of seven logic-based

filters arranged in order of increasing complexity. They are as

follows:

1. Remove duplicate records and data with negative values

or differential time steps

2. Replace outlying high/low speed values

3. Remove zero-speed signal drift when vehicle is stopped

4. Replace false zero-speed records

5. Amend gaps in data

6. Repair outlying acceleration/deceleration values

7. Denoise and condition final signal.

Removing Data with Duplicate/Negative Time

Records

As an

initial step in the filtration process, it is necessary to

remove any data points with duplicate time values and data

points that have negative or zero differential time values (i.e.,

time at point 2 is less than time at point 1, or time at point 2 is

equal to time at point 1). To remove these points, the filter

first calculates the differential time values for each of the data

points in the source set and then removes any with

differential time values less than or equal to zero.

This step must be performed first due to the influence of

differential time when calculating acceleration from speed

information. If duplicate time records exist, then by definition

of acceleration from basic kinematics, the acceleration at the

duplicate point would approach infinity. If erroneous time

information were not removed in this preliminary step, we

would be unable to perform accurate calculations based on

differential/integral time information, and the future filtration

steps that rely on this information would be impossible.

Replacing Outlying Speed Values

In the

second step in the filtration process, any erroneous data

points, such as single-sample high-speed data spikes and

negative speed signal dropouts, that are present in the data set

are removed and replaced with interpolated data. To achieve

this, each speed value in the sample set is processed through

the filter and compared individually to chosen high/low speed

limits. If the data point is found to lie outside the chosen

limits, the filter replaces the source sample data point with

speed information derived from a cubic spline interpolation

drawn from neighboring data [

14]. As an added step in the

interpolation

process, the filtered vehicle speed is limited on

the lower bound to zero miles per hour to facilitate vehicle

simulation and dynamometer testing. The application of user-

controlled boundary limits allows for flexibility during data

processing and improves the quality of the results when

vehicle speed limitations are known.

Figure 1, a 75-mph high-speed limit was used in

conjunction with

the outlying-speed filter to remove the

erroneous high-speed signal spike in the sample GPS vehicle

data. The limit was applied assuming that speeds in excess of

75 mph could be attributed to signal spikes and not normal

operation. For the case of unknown vehicles, selecting more

conservative high/low operating limits is advised to avoid

removal of accurate information. Employing controllable

limits in the outlying-speed filter provides a unique advantage

when filtering large numbers of drive cycles as filtration

limits can be adjusted for individual vehicle types.

Figure 1. Outlying speed filtration results for sample

GPS vehicle data.

Removing GPS Zero-Speed Drift

the

third step of the filtration process, an effect called

“zero speed drift” is removed. Zero speed drift appears when

a GPS DAS is running and the vehicle is stopped and idling

for a prolonged period of time. During extended-duration idle

events, GPS data loggers will often record a very small speed

value (0.1

or 0.2 mph) due to GPS satellite signal

reacquisition that occurs when a vehicle is stopped. To

remove the small fluctuations in vehicle speed recorded by

the GPS DAS during these periods, the zero-speed drift filter

examines the distance traveled during each microtrip in a

drive cycle and compares the value with a user-specified

limit. If the distance traveled over the course of the microtrip

is below the limit, the entire microtrip is replaced with zero-

speed data values.

As seen in

Figure 2, applying a 0.001-mile distance limit to

the sample

GPS data causes removal of the low-speed

microtrip. When performing traditional integration-based

distance calculations using coupled speed-time information,

removing the zero-speed drift data results in lower calculated

distances, which are more reflective of actual vehicle

operation. Similarly, when performing simulations using

plug-in hybrid-electric and electric vehicle (PHEV and EV)

models, removing these zero-speed drift segments results in

more accurate energy consumption calculations and improved

state-of-charge estimates.

Figure 2. Zero-speed drift filter results for sample GPS

vehicle data.

Replacing False Zero-Speed Records

The next

step in the filtration process, the false zero-speed

filter, removes single-point zero-speed data records that are

the result of temporary GPS signal dropout. Similar to the

outlying speed filter, the false-zero filter removes false zero-

speed records by examining the value of each individual

speed point in the data set in relation to its nearest neighbors.

If a given speed record value is zero and the neighboring

points on each side are both nonzero, the zero-speed point is

replaced with a point drawn from a cubic spline interpolation

of the entire remaining data set.

This simple filter, shown in

Figure 3, improves the continuity

the raw data by replacing significant outliers in the data

with data more representative of true vehicle operation. By

removing the dramatic changes in speed, energy calculations

are improved when analyzing the drive cycle; more

importantly, in vehicle simulation scenarios the model is

capable of accurately following the prescribed speed-time

trace.

Figure 3. False zero-speed filter results for sample GPS

vehicle data.

Amending Signal Gaps

this

step, the filtration algorithm attempts to correct for

gaps in the coupled speed-time GPS signal caused by urban

canyon effects and sudden signal loss. Employing a user-

supplied time gap limit, the filter examines the time stamp

information pulled from the input GPS data stream and

attempts to interpolate over any signal gaps that are shorter in

duration than the specified limit. If the signal gap duration is

shorter than the limit, the algorithm generates monotonically

increasing time signals based on the sampling rate of the

underlying source data. To generate the “new” speed data, the

same interpolated cubic spline curve fit used in previous

filtration steps is applied over the newly generated time

domain.

If the filter detects a time gap greater than the user-defined

limit, as shown in

Figure 4, the filter ramps the vehicle speed

down or up based on whether or not the data point is at the

leading or trailing boundary of the gap in the signal. The

ramp is constructing based on user-specified acceleration/

deceleration limits. Construction of the ramps is particularly

important when attempting to adapt field data for chassis

dynamometer and vehicle simulation applications. Fixing

time gaps is the only step in the filtration process that will

add data to the existing GPS data set (coupled time-speed

information). Along with testing/experimental applications,

the signal reconstruction provided by the signal gap filter is

also important

on an empirical level when examining the duty

cycle of the vehicle, especially when performing analysis on

real-time operation. Accurate knowledge of vehicle key

on/off time is integral in determining the potential for

opportunity charging events for PHEVs or EVs during daily

operation, as well as cool-down times for power electronics

and other thermal components.

Figure 4. Signal gap filter results for sample GPS vehicle

data.

Repairing Outlying Acceleration/Deceleration

Values

In the

last major step in the filtration process, an algorithm

attempts to fix outlying accelerations/decelerations that

remain in the raw GPS data set after all prior filtration steps.

While the outlying speed filter removed speed points that

existed outside the realm of velocity limits, this filter looks at

the derivatives with respect to time of the speed data to see if

the recorded data matches expected vehicle performance

levels. The filter loops through the data set and progressively

refines and smoothes the acceleration of outlying points.

While looping through the data set, if the acceleration/

deceleration of a data point is outside the user-supplied limits,

the point is replaced with a value that is calculated using the

interpolation method described above. However, if the

interpolated value still produces an acceleration outside of the

acceptable limits, the filter adjusts the new data point to fit

within 1% of the user-defined limits (+1% for acceleration,

−1% for deceleration). This filter loops through the speed/

time information until all acceleration/deceleration values fall

within the user-defined limits (

Figure 5).

Figure 5. Outlying acceleration filter results for sample

GPS vehicle data.

It is important to select limits for this filter that closely match

the performance expected of the vehicle. Applying heavy-

duty acceleration/deceleration limits to a light-duty vehicle

would result in large amounts of original data being removed;

similarly, applying light-duty limits to heavy-duty vehicle

data would remove too little data. In cases where vehicle and

power train information are known, a rough estimate of

appropriate acceleration/deceleration limits can be

determined. Otherwise, conservative limits based on vehicle

classification and weight are recommended.

Denoising and Conditioning the Signal

The final

step in the filtration process combines a localized

least squares polynomial smoothing filter [commonly known

as a Savitzky-Golay (SG) filter] [

15, 16, 17, 18] and a

binomial smoothing filter [19 - 20]. The combined SG-

binomial filter reduces the effect of signal white noise while

maintaining the overall profile of the drive cycle. The

advantage of the combined SG-binomial filter is that the SG

component acts as a complementary form of the acceleration/

deceleration filter by definition minimizing the effects of any

outlying artifacts remaining from prior filtration steps, while

the binomial component of the filter acts to remove any

underlying white noise found in the signal. As shown in

Figure 6, the filter is able to maintain the shape of the original

drive cycle while removing any remaining outliers and

background signal noise.

Figure 6. Denoising filter applied to time-speed GPS

vehicle data.

Removing noise from raw GPS data signals and providing

additional smoothing conditions are integral in preparing

source data for use in chassis dynamometer/vehicle

simulation applications. For example, removing existing

signal transients and background noise is especially important

when running long-duration chassis dynamometer tests.

Instead of having testing personnel attempt to follow a highly

variable speed-time trace due to signal noise components, the

testing personnel are able to maintain more natural driving

behavior and follow a smooth driving profile more in line

with typical driving behavior. For vehicle simulation

activities, removing the remaining signal noise avoids

complications that arise when simulating advanced vehicle

(PHEV/EV) control strategies. Noisy speed profiles generate

incorrect power demand information, resulting in inaccurate

state-of-charge and energy consumption calculations, directly

correlating to reduced simulated fuel economy values.

GPS DATA COLLECTION

To evaluate the effectiveness of the proposed filtration

process, the authors analyzed two unique sets of GPS speed-

time data. The light-duty data set was drawn from a

collection of unknown light-duty vehicles [21] instrumented

with GeoStats GeoLoggers as part of the year 2000 Southern

California Association of Governments post-census regional

travel survey, while the heavy-duty data set was drawn from

class 3 to class 8 heavy-duty vehicle data collected from 2008

to present [7, 22, 23]. The GPS data for the heavy-duty data

set was collected using a combination of GeoStats

GeoLoggers and Isaac Instruments DRU908 and DRU900

data acquisition devices. The light-duty data set contained

1,202 individual drive cycles, while the heavy-duty set

contained 588 individual drive cycles. The goal of selecting

two unique data sets, one light duty and the other heavy duty,

consisting of a mix of known and unknown vehicles, was to

evaluate the effectiveness of the filtration process across a

wide range of vehicle types and vocations.

FILTER INPUTS

Before evaluating

the effectiveness of the proposed filtration

process, acceptable inputs for each stage in the filtration

process were established. Conservative limits were selected

for both the light- and heavy-duty vehicle data sets, with the

focus on setting speed and acceleration limits at the bounds of

realistic operation. For the light-duty vehicle data set, a

vehicle speed range of up to 120 mph and maximum absolute

acceleration limits of approximately ±0.8G were selected.

These values were chosen based on the performance

characteristics of a 2010 Chevy Corvette, the maximum

performance values likely in a light-duty vehicle [

24]. For the

zero-speed drift

filter and the signal gap filter, standard

values were selected based on GPS signal reacquisition

technical specifications for a Garmin 18x GPS receiver [

25],

and an

assigned minimum trip distance of approximately 50

ft.

Table 1 summarizes the filtration inputs for the light-duty

vehicle data set.

Table 1. Light Duty Vehicle Data Set Filtration Limits

For the heavy-duty vehicle data set, the acceleration range

limits were selected based on a drive-cycle analysis of

standard heavy-duty chassis dynamometer test cycles. The

Rowan University Composite School Bus Cycle (RUCSBC)

[26] contained the greatest acceleration rate of the 32

standard cycles examined; therefore, the maximum

acceleration in the RUCSBC cycle was chosen as the limiting

range for the acceleration filter. The distance and time-gap

limits for the heavy-duty vehicle data set were chosen to

match those of the light-duty vehicle data set.

Table 2. Heavy Duty Data Set Filtration Limits

RESULTS

LIGHT DUTY

The filtration

results for the light-duty vehicle data set are

shown in

Table 3. One first observes that on average

approximately 20%

of the original data set is undergoing

some level of filtration. This would appear to be a high value;

however, when the two filtration steps that add/remove data

based on signal discontinuity are excluded, we see that, on

average, less than 1% of the collected data are undergoing

any filtration. These results support the goal of minimizing

the total amount of signal processing performed on raw data;

however, they raise questions about the quality of the source

GPS data. The approximately 20% addition/removal of data

that is occurring during the negative/duplicate time and signal

gaps filter suggest source data with poor continuity and

significant data acquisition issues.

Comparing the raw and filtered data for the light-duty vehicle

GPS drive-cycle samples, shown in

Table 4, we can see that

outlying speed

data in the raw data set generate dramatic

effects on the calculated mean maximum driving speed,

characteristic acceleration, and aerodynamic speed [

4]. While

only 0.004%

of data are adjusted on average as part of the

high/low speed filter, the outlying speeds in the raw data set

prior to this filtration step render global drive-cycle statistics

calculated from the raw data useless for direct comparison.

Due to the existence of these outliers, the average maximum

driving speed of the raw drive cycles is three orders of

magnitude higher than that of the filtered data.

Performing paired t-tests to determine the effects of the

filtration on the drive-cycle metrics, it was found that the

proposed filtration process produced statistically significant

changes in the mean values for each of the drive-cycle

metrics examined with the exception of characteristic

acceleration and aerodynamic speed. The large standard

deviations for the raw data set contribute to the uncertainty in

statistically determining the effects of the filtration on

characteristic acceleration and aerodynamic speed; however,

visually examining the data reveals the effects of filtration to

be nontrivial.

HEAVY DUTY

Examining the filtered results for the heavy-duty vehicle data

set, shown in

Table 5, one observes that compared to the

approximate 20%

data filtration for the light-duty vehicle

data, the heavy duty vehicle data examined in the study

showed less than 5% filtration on average. When the two

filtration steps that add/remove data are ignored, less than

0.5% of the raw data collected undergoes filtration on

average.

Performing drive-cycle analysis comparisons on both the raw

and filtered heavy-duty vehicle GPS drive-cycle samples as

shown in

Table 6, outlying data appear to contribute a

Table 3. Light-Duty Vehicle Filtration Results

Table 4. Light-Duty Drive Cycle Metrics

minimal effect on the quality of the raw heavy-duty vehicle

data when

compared to the light-duty vehicle data. On

average, the maximum and average driving speed of the cycle

are reduced, while the total speed of the cycle increases,

suggesting the repair of signal dropouts and spikes.

Performing paired t-tests to determine the effects of the

filtration on the drive-cycle metrics, it was found that the

proposed filtration process produced statistically significant

changes in the mean values for each of the drive-cycle

metrics examined. However, examining the differences in the

means, medians, and standard deviations of the metrics

examined, it appears that while the changes in mean are

statistically significant, overall the change in the data are

minimal on the whole.

CONCLUSION

Based on the results of this study, it can be concluded that the

proposed GPS data filtration adequately fulfills the

requirements proposed for an ideal GPS data filter. Excluding

removal of erroneous data and addition of lost signal

components, the proposed filtration process minimally affects

the source data while simultaneously generating significant

improvements in the quality of data required for calculation

of drive-cycle metrics used as part of drive-cycle analysis and

vehicle simulation applications.

Analyzing the filtered results from both the light- and heavy-

duty vehicle data sets, gaps in signal acquisition and

underlying data acquisition issues are a much greater

component of the filtration process than previously

anticipated. The loss of GPS signal is the greatest factor

contributing to the loss in data quality on a point-per-point

basis, accounting for at least an order of magnitude greater

influence in the filtration process than the other sources of

error. However, examining the duty-cycle analysis results

based on the filtered and raw data for both the light- and

heavy-duty vehicle data sets, it was also shown that removing

data spikes is the key step in producing accurate vehicle

speed/acceleration estimates.

Based on the results of this study coupled with increased

interest in the integration of elevation/road grade information

into engineering analyses, in the future it is recommended

that additional filtration methods be developed to repair

collected GPS elevation information. Combining an

additional dimension to the filtration process should

introduce additional logic-based filtration opportunities that

do not currently exist when examining the data in a single

dimension.

Table 5. Heavy Duty Filtration Results

Table 6. Heavy Duty Drive Cycle Metrics

REFERENCES

1. Ivani, Ž., “Data Collection and Development of New York

City Refuse Truck Duty Cycle.” SAE Technical Paper

2007-01-4118, 2007, doi:10.4271/2007-01-4118.

2. Casey, E.J., Smith, W.J., and Timoney, D.J.,

“Examination of Low-cost Systems for the Determination of

Kinematic Driving Cycles and Engine Operating Conditions

in Dublin, Ireland,” SAE Technical Paper 2009-01-2791,

2009, doi:10.4271/2009-01-2791.

3. Dembski, N., Rizzoni, G., Soliman, Fravert, J. et al.,

“Development of Refuse Vehicle Driving and Duty Cycles,”

SAE Technical Paper 2005-01-1165, 2005, doi:

10.4271/2005-01-1165.

4. O'Keefe, M.P., Simpson, A., Kelly, K.J., and Pedersen,

D.S., “Duty Cycle Characterization and Evaluation Towards

Heavy Hybrid Vehicle Applications,” SAE Technical Paper

2007-01-0302, 2007, doi:10.4271/2007-01-0302.

5. Frey, H.C., et al. “In-Use Measurement of the Activity,

Energy Use, and Emissions of a Plug-in Hybrid Electric

Vehicle,” Paper 2009-A-242-AWMA, Proceedings, 102nd

Annual Conference and Exhibition, Air & Waste

Management Association, Detroit, Michigan, June 16-19

2009.

6. Earleywine, M., et al. “Simulated Fuel Economy and

Performance of Advanced Hybrid Electric and Plug-In

Hybrid Electric Vehicles Using In-Use Travel Profiles.”

Vehicle Power and Propulsion Conference (VPPC), 2010

IEEE, pp 1-6, 1-3 September 2010.

7. Barnitt, R.A. and Gonder, J., “Drive Cycle Analysis,

Measurement of Emissions and Fuel Consumption of a

PHEV School Bus,” SAE Technical Paper

2011-01-0863,

2011, doi:10.4271/2011-01-0863.

8. Kulkarni, A., Sapre, R.R., and Sonchal, C.P., “GPS Based

Methodology for Drive Cycle Determination,” SAE

Technical Paper 2005-01-1060, 2005, doi:

10.4271/2005-01-1060.

9. Gonder, J., et al. “Using Global Positioning System Travel

Data to Assess Real-World Energy Use of Plug-In Hybrid

Electric Vehicles.” Transportation Research Record: Journal

of the Transportation Research Board, volume 2017, pp.

26-322007.

10. Witte, T.H., et al. “Accuracy of WAAS-enabled GPS for

the determination of position and speed over ground.”

Journal of Biomechanics, Volume 38, Issue 8, pp.

1717-1722, August 2005.

11. Arnold, L., et al. “Positional accuracy of the Wide Area

Augmentation System in consumer-grade GPS units.”

Computers & Geosciences, Volume 37, Issue 7, pp. 883-892,

July 2011.

12. Jackson, E., et al. “The Accuracy of GPS-Based

Acceleration for Vehicle Emissions Modeling.”

Transportation Research Board 86

Annual Meeting, Paper

No. 07-2202, 2007.

13. NREL Vehicle Drive Cycle Tool, User Guide. Copyright

Reserved.

14. Feng, G., “Data Smoothing by Cubic Spline Filters,”

Signal Processing, IEEE Transactions on, Volume 46, Issue

10, pp. 2790-2796, Oct 1998.

15. Savitzky, A., et al. “Smoothing and Differentiation of

Data by Simplified Least Squares Procedures.” Analytical

Chemistry, Volume 36, Issue 8, pp. 1627-1639, 1964.

16. Madden, H., “Comments on the Savitzky-Golay

Convolution Method for Least-Squares Fit Smoothing and

Differentiation of Digital Data.” Analytical Chemistry,

Volume 50, Issue 9, pp. 1383-1386, 1978.

17. Bromba, M., et al. “Application Hints for Savitzky-Golay

Digital Smoothing Filters.” Analytical Chemistry, Volume

53, Issue 11, pp. 1583-1576, 1981.

18. Gorry, A., et al. “General Least-Squares Smoothing and

Differentiation by the Convolution (Savitzky-Golay)

Method.” Analytical Chemistry, Volume 62, Issue 6, pp.

570-573, 1990.

19. Marchand, P., et al. “Binomial Smoothing Filter: A Way

to Avoid Some Pitfalls Of Least-Squares Polynomial

Smoothing.” Review of Scientific Instruments, Volume 54,

Issue 8, pp. 1034-1041, Aug 1983.

20. Aubury, M., et al. “Binomial Filters.” Journal of VLSI

Signal Processing, Volume 12, Issue 1, pp. 35-50, 1996.

21. NuStats, “Year 2000 Post-Census Regional Travel

Survey, Final Report of Survey Results.” Southern California

Association of Governments, 2003.

22. Barnitt, R. “FedEx Express Gasoline Hybrid Electric

Delivery Truck Evaluation: 12-Month Report.” Report

TP-5400-48896, 2011.

23. Lammert, M. “Twelve-Month Evaluation of UPS Diesel

Hybrid Electric Delivery Vans.” Report TP-540-44134, 2009.

24. Ponticel, P. “Focused on Fuel Economy.” Automotive

Engineering International, August 6 2010, pp. 28-30.

25. Garmin, “GPS 18x Technical Specifications.” October

2011. pp. 6

26. Hearn, J., Toback, A., Akers, J., Hesketh, R.P. et al.,

“Development of a New Composite School Bus Test cycle

and the Effect of Fuel Type on Mobile Emissions from Three

School Buses,” SAE Technical Paper

2005-01-1616, 2005,

doi:10.4271/2005-01-1616.

CONTACT INFORMATION

Adam Duran

is a research engineer working with the Center

for Transportation Technologies and Systems at the National

Renewable Energy Laboratory. Adam's work focuses

primarily in

the areas of drive cycle analysis and

characterization, custom drive cycle development, and

medium/heavy duty fleet evaluations. He may be reached at

[email protected].

Matthew Earleywine

is a research engineer working with the

Center for Transportation Technologies and Systems at the

National Renewable Energy Laboratory. His work focuses

primarily in the areas of advanced vehicle modeling,

simulation, & analysis, vehicle charging infrastructure, and

electric roadway implementation. He may be reached at

[email protected]

ACKNOWLEDGMENTS

The authors

wish to thank Lee Slezak, U.S. Department of

Energy Vehicle Technologies Program, as well as the

Department of Energy for its support in conducting this

project. In addition, the support of NREL staff Ben Rosen,

Jeff Gonder, and Kevin Walkowicz was invaluable to the

completion of this work.

DEFINITIONS/ABBREVIATIONS

DAS

data acquisition system

electric vehicle

gravitational constant (9.81 m/s

)

GPS

global positioning system

PHEV

plug-in hybrid vehicle

mph

mile per hour

NREL

National Renewable Energy Laboratory

RUCSBC

Rowan University Composite School Bus Cycle

second

Savitzky-Golay

The Engineering Meetings Board has approved this paper for publication. It has

successfully completed SAE's peer review process under the supervision of the session

organizer. This process requires a minimum of three (3) reviews by industry experts.

ISSN 0148-7191

Positions and opinions advanced in this paper are those of the author(s) and not

necessarily those of SAE. The author is solely responsible for the content of the paper.

SAE Customer Service:

Tel: 877-606-7323 (inside USA and Canada)

Tel: 724-776-4970 (outside USA)

Fax: 724-776-0790

Email: [email protected]

SAE Web Address: http://www.sae.org

Printed in USA