1. 16th TRB National Transportation Planning Applications Conference
APPLICATION OF HIGH-RESOLUTION DATA TO DETERMINE THE IMPACT OF LOS ON DRIVING BEHAVIOR ON ARTERIALS Nabaruna Karmakar Celeste Chavis, Ph.D. Mansoureh Jeihani, Ph.D. Zohreh Rashidi Moghaddam Behzad Aghdashi, Ph.D.
May 17, 2017

2. Outline Introduction and Motivation Data collection Methodology
Preliminary Results and Observations
Conclusions and Future work
5/17/2017

3. Introduction and Motivation
5/17/2017

4. Introduction and Motivation
Typically, “user delay costs” is the key factor in characterizing benefits of treatment projects on arterials.
However, improvements in LOS may have impact on driver aggressiveness and, as such, bring safety costs into decision making.
The main goal of this project is to assess the impact of LOS on driver behavior on arterials.
Conventional freeway analysis has often times focused on mobility improvements. Not much study has been done on analysis of impacts of LOS on arterial streets.
This is an ongoing project between ITRE and Morgan State University, funded through UMD-NTC.
5/17/2017

5. Introduction and Motivation
Active Traffic Management (ATM)
Regional Planning
Strategic Development Plans
Traffic Condition (LOS = A, B, C, D, E, F)
Emerged Travel Time and Delays
Driver Behavior (Aggressiveness, etc.)
User Delay Cost
Fuel Cost and Emissions; Safety Concerns
Economic Competitiveness
Traffic management and planning activities are focused on improving mobility performance measures, aimed at improving LOS. The efficacy of these activities are usually gauged using Travel Time delay. Worsening LOS can also give rise to safety concerns on arterials, resulting from driver frustration and aggressiveness.
5/17/2017

6. Data Collection
5/17/2017

7. Underlying Technology
Vehicle sensor systems
Speed
RPM
Throttle Level
Engine Temperature
Gear Position
There are various sensor systems in our vehicles that can help measure variables that can help us get an idea of a person’s driving style. They can be measured from the car’s OBD port.
On-Board Diagnostic (OBD) Port
5/17/2017

8. Introduction of New Technology Through i2D – “Intelligence to Drivers”
Collaboration with partners in
Portugal (TUL and ITDS)
Additional Sensors
GPS Coordinates
3D Acceleration
Barometric altimeter
Enhanced Processing,
Storage, and Communications
Simple Processor
4GB SD Card
Cellular Network
5/17/2017

9. i2D Technology Integration Process
Web
Access
Vehicles
Sensors
Cloud
Database
OBD
Port
i2D Device
Additional
Sensors
Cellular
Network
Memory
Processor
5/17/2017

10. Summary of the Collected Data
About 40 volunteer drivers
Gathered over 39 million seconds of data over 3 years ( ) for 46,828 trips.
Recorded over 360K miles of driving behavior
All data have been archived on an ITRE server (DaTA Lab- Driver and Transportation Analytics)
We are in process of analyzing and interpreting collected data using
ArcGIS
Independent Platforms
5/17/2017

11. HERE/ INRIX data
HCM service measure for LOS on arterials is the ratio of travel speed to free flow speed.
Measured speeds on the study segments were used to assess the LOS during the trips.
ITS probe data, provided by HERE and INRIX,
Data collected for three years ( ).
All data have been archived on an ITRE server (DaTA Lab- Driver and Transportation Analytics)
5/17/2017

12. methodology
5/17/2017

13. Methodology Site Selection Data Cleaning and Integration
First, trip data was filtered for only the selected road segments, using an internally developed Arc GIS Tool
INRIX and HERE Speed data was downloaded for the years for the selected TMC segments
Data Integration of these two data sets was done by matching the timestamps of trip data and probe data
Outlier Removal was done using K-means clustering
Exploratory Data Analysis was done using R and Tableau
Sample size – Distribution of number of trips and different drivers
Trends in acceleration across different LOS
Frequently traveled road segments
From a set of 20 sites, we selected one site with good distribution of LOS observations (A through F)
Site Selection
Frequently traveled road segments with a good distribution of LOS observations
Data Cleaning and Integration
Data extraction on road segments ArcGIS and Data Integration and formatting using R
Data Mining and Visualization
Exploratory Data Analysis and visualization using Tableau and R

14. Preliminary results and observations
5/17/2017

15. Selected Site Western Blvd WB near NC State University Western Blvd
TMC
Western Blvd
TMC:
Length of Segment: 0.7 miles
Includes 4 signalized intersections (3 HCM Segments)
Speed Limit: 45 mph
Number of Lanes: 2
48,952 rows of data
5/17/2017

16. Number of Trips and Drivers - Western Blvd WB
Year
Number of Drivers
Number of Trips
2014 15
121
2015 18
205
2016 12 50
376 total trips
5/17/2017

17. Distribution of Trips and LOS observed over Time of Day
5/17/2017

18. Sample Size of data over different LOS
5/17/2017

19. Number of Zero Acceleration Seconds over different LOS
5/17/2017

20. Average and Standard Deviation of Acceleration and Deceleration
5/17/2017

21. Maximum Acceleration and Deceleration by trips over different LOS
5/17/2017

22. Lateral Acceleration over different LOS
5/17/2017

23. 3D combined Acceleration over different LOS
Average
Standard Deviation
5/17/2017

24. Conclusions & Future Work
5/17/2017

25. Future Work
Characterize the impact of weather and incidents in analysis.
Consider other measures of aggressiveness
Further data mining
Outlier analysis
Predictive modeling of driver behavior
Validate findings from Driver Simulator
5/17/2017

26. 5/17/2017

27. Appendix
5/17/2017

28. DEFINING LEVEL OF SERVICE HCM Chapter 16
LOS
Based on Speed Limit of 45 mph A
>36 B
>30 C
>25 D
>20 E
>15 F
<=15
5/17/2017

29. Future Work K-Means Clustering Visualization
Site A
5/17/2017

30. Driving Simulator Capabilities
The hardware is very similar to a real car and consists of the driver seat, cockpit, steering wheel, acceleration and brake pedals, etc.
A road network can be developed in the software and drivers are able to select their own route between origin and destination.
The software has a capability of generating up to 5000 vehicles per lane on the road, mixes of cars, buses, trucks, and motorcycles
It collects data of the driver’s car every 1/100 seconds including but not limited to geographic positions, speed, lane, distance traveled, offset from road’s shoulder, acceleration, brake, yaw/pitch/roll.
Different scenarios of traffic situations, environment and information provision can be simulated.
5/17/2017

31. Driver Simulator - LOS Network
The developed arterial in the driving simulator
A 4.5-mile arterial in Baltimore, MD is developed in the driving simulator
High congestion level in this network is assured mainly by growing traffic volume and decreasing traffic speed
50 participants are expected to be recruited to drive
All participants are expected to drive all levels of service
A screenshot of the developed network
5/17/2017

32. Contrasting i2D and Smart Phone Sensors
i2D allows connection to vehicle sensors
Improved accuracy of sensor readings (e.g. 3D accelerometer) by installing device in vehicle
Enhanced and organized database in the cloud
Web access to download organized data
i2D personal users web access
Demonstration of driving behavior
Managing fleet members
i2D research platform
Several methods to access and download detailed data
i2D VIV Control Panel
5/17/2017

33. Number of Trips and Day of Week
5/17/2017