Evaluation of traffic data obtained via GPS-enabled mobile phones: The Mobile Century field experiment

TL;DR: Results suggest that a 2-3% penetration of cell phones in the driver population is enough to provide accurate measurements of the velocity of the traffic flow, demonstrating the feasibility of the proposed system for real-time traffic monitoring.

Abstract: The growing need of the driving public for accurate traffic information has spurred the deployment of large scale dedicated monitoring infrastructure systems, which mainly consist in the use of inductive loop detectors and video cameras On-board electronic devices have been proposed as an alternative traffic sensing infrastructure, as they usually provide a cost-effective way to collect traffic data, leveraging existing communication infrastructure such as the cellular phone network A traffic monitoring system based on GPS-enabled smartphones exploits the extensive coverage provided by the cellular network, the high accuracy in position and velocity measurements provided by GPS devices, and the existing infrastructure of the communication network This article presents a field experiment nicknamed Mobile Century, which was conceived as a proof of concept of such a system Mobile Century included 100 vehicles carrying a GPS-enabled Nokia N95 phone driving loops on a 10-mile stretch of I-880 near Union City, California, for 8 hours Data were collected using virtual trip lines, which are geographical markers stored in the handset that probabilistically trigger position and speed updates when the handset crosses them The proposed prototype system provided sufficient data for traffic monitoring purposes while managing the privacy of participants The data obtained in the experiment were processed in real-time and successfully broadcast on the internet, demonstrating the feasibility of the proposed system for real-time traffic monitoring Results suggest that a 2-3% penetration of cell phones in the driver population is enough to provide accurate measurements of the velocity of the traffic flow

Summary

1 Introduction

• Before the era of the mobile internet, characterized in particular by the emergence of location based services heavily relying on GPS, the traffic monitoring infrastructure has 1Corresponding author: Assistant Professor.
• Electronic devices traveling onboard cars are appealing for this purpose, as they usually provide a cost-effective and reliable way to collect traffic data.
• The fundamental limitations of this system is the cost to install the infrastructure , its limited coverage, and the fact that only travel time between two locations can be obtained.
• In [21] and [22], the authors conclude that if GPS-equipped cell phones are widely used, they will become more attractive and realistic alternative for traffic monitoring.
• Section 4 presents the main results obtained from the data.

2.1 Sampling and Data Collection

• As explained earlier, a variety of sampling techniques can be used to collect data from GPS enabled mobile devices.
• In the case of the Nokia N95, the embedded GPS chip-set is capable of producing a time-stamped geo-position (latitude, longitude, altitude) every three seconds.
• To manage privacy concerns, in addition to pseudo-anonomization of the trajectory data, the data can be further degraded until a sufficient level of privacy is attained.
• Common degradation approaches include (i) spatial obfuscation (i.e. blocking data collection from particular regions, such as home), (ii) increasing uncertainty in the data through noise addition, and (iii) location discretization approaches, which round the measurement to the nearest discrete grid point.
• Mobile devices monitor their speed and location using GPS and use the locally stored VTLs to determine when a VTL crossing occurs.

2.2 System Architecture

• A prototype system architecture was implemented to test VTL based sampling strategies (shown in Figure 1).
• On each participating mobile device (or client), an application is executed which is responsible for the following functions: downloading and caching trip lines from the VTL server, detecting trip line traversal, and filtering measurements before transmissions to the service provider.
• These VTL updates are transmitted to the ID proxy server over a secure channel.
• Thus the authors prevent any single entity from observing both the identification data required by the network operator, and the sensing data.
• The traffic report server then sends data to information consumers through a mapping interface on a web site.

3 Experimental design

• The experiment was conceived as a proof of concept of the system described in the previous section.
• Drivers were instructed to drive as they would normally, on one of the three routes.
• It presents interesting traffic properties, which include alternating periods of free-flow and congestion throughout the day (which thus satisfies the requirements of Goal 2).
• Table 1 presents the main features of the loops used during the experiment, also shown in 7At the present, all dual loops on this experiment site are treated as single loops by PeMS for the purpose of computing speeds.
• First, each Nokia N95 cell phone was storing its position and velocity log every 3 seconds, which allows for the computation of every equipped vehicle trajectory.

4 Experimental results

• This section analyzes the main results derived from the experiment.
• Unless otherwise noted, the rest of this section focuses on the highway segment covered by the afternoon loops in the northbound (NB) direction.
• The data obtained in the experiment using the system architecture described in Section 2 were processed in real-time.
• As can be seen from the two subfigures in Figure 3, the extent of congestion estimated by their algorithm9 and based on the GPS data only match closely the 511.org display, which uses a combination of data sources for velocity and travel time calculation including loop detectors, FasTrak-equipped vehicles, and speed radars.
• Comparisons with the 511.org speed map at other times during the experiment showed similar results, which confirm that the GPS cell phone based technique and the system described in Section 2 can produce reasonable speed estimates for the section of interest, at least for the day of the experiment.

Trajectory data

• Each phone stored its position (latitude and longitude) and a velocity log every 3 seconds.
• The propagation of the shockwave generated by the accident is clearly identified from this plot as well.
• The size of the influence area depends on the proximity of neighbor detector stations.
• The qualitative agreement between subfigures a) and b) is evident – in terms of bottlenecks location, and their spatial and temporal extent.
• When sampled in time (every 3 seconds in this case), mobile sensors can provide with spatial information – such as the backward propagation of congestion – that would only be available with a high density of loop detector stations.

VTL data

• In addition to the trajectory data stored by each phone, VTL data were collected during the experiment using the system architecture described in Section 2.
• By placing VTLs on existing loop detector locations (17 in total), velocity measurements collected by a loop detector every 5 minutes can be compared to the ones provided by a VTL at the same location.
• Note that loop detector measurements are usually considered as ground truth.
• Drivers hired for the experiment are not necessarily a proper statistical sample of the population.

5 Conclusions

• The Mobile Century field experiment presented in this article was conceived as a proof of concept for a traffic monitoring system based on GPS-enabled mobile phones.
• The prototype system exploits the extensive coverage provided by mobile phones and the high accuracy in position and velocity measurements provided by GPS units.
• The comparison suggests the presence of some bias in the velocity estimation for some loop detectors, showing sometimes substantial differences with the VTL measurements.
• Cover Northern California [28] in its initial phase.
• Therefore, the potential errors, inaccuracies, and/or biases observed in the data will be addressed to compute travel time estimates or other features extracted from it as clearly as shown for the raw data, with the proper flow models of highway traffic and corresponding inverse modeling techniques.

Figures

Figure 2: Stretch of highway I-880 CA, used in the Mobile Century experiment.

Figure 5: Loop detector locations for the NB direction. Numbers indicate postmiles. Traffic flows from left to right, and loop detectors have been numbered sequentially from 1 (upstream) to 17 (downstream).

Figure 10: Penetration rate map (in %).

Figure 11: a) and c): Average penetration rate over time at existing detector station locations during the morning and the afternoon. The range is one standard deviation below and over the mean. Traffic flows from left to right. b) and d): Histogram of the penetration rate including all the 17 locations during the morning and the afternoon.

Figure 4: Vehicle trajectories in NB direction extracted from the data stored by 50% of the cell phones. The propagation of the shockwave from the accident can clearly be identified from this plot. The red lines in the subfigure were drawn by hand by fitting a line through the points where trajectories change slope.

Figure 7: Travel time (in minutes) between Decoto Rd. and Winton Ave. Dots correspond to individual vehicle travel times (4268 in total), collected manually using video cameras at the ends of the section of interest. Black dash-dotted lines correspond to the standard deviation of the average travel time obtained from the video cameras in 5-minute windows. The time in the x-axis is the entry time to the section of interest.

Figure 3: Live traffic feed at 10:52am on February 8, 2008, after an accident on the NB direction of I880 occurred, provided by the proposed system and 511.org (inside). Numbers in circles correspond to speed in mph.

Figure 8: Loop detector vs. VTL velocity measurements (all locations). Dotted lines are the ±5 mph thresholds.

Table 1: Features of the loops used in the experiment.

Figure 1: Mobile Century system architecture overview. The system consists of vehicles equipped with GPS-enabled smartphones (Nokia N95), a cellular network provider, a data collection infrastructure and a traffic estimation engine, and an information display system. A live tracking infrastructure (shown in dashed green) was also required for the safety of the UC Berkeley student drivers during this experiment, but it is not part of the core system (shown in solid black).

Figure 9: Loop detector versus VTL velocity data collected at detectors a) 1, b) 4, c) 17, and d) 7. Subfigure e) shows the penetration rate at these four locations during the day.

Figure 6: Velocity field in (mph) using a) 17 loop detector stations, b) vehicle trajectories and Edie’s speed definition, c) 17 VTLs at the loop detector locations, and d) 30 VTLs equally spaced.

Full text

UC Berkeley
Recent Work
Title
Evaluation of Traffic Data Obtained via GPS-Enabled Mobile Phones: the Mobile Century
Field Experiment
Permalink
https://escholarship.org/uc/item/0sd42014
Authors
Herrera, Juan C.
Work, Daniel B.
Herring, Ryan
et al.
Publication Date
2009-08-01
eScholarship.org Powered by the California Digital Library
University of California

Evaluation of Traffic Data Obtained via GPS-Enabled Mobile
Phones: the Mobile Century Field Experiment
Juan C. Herrera, Daniel B. Work, Ryan Herring, Xuegang (Jeff)
Ban, and Alexandre M. Bayen
WORKING PAPER
UCB-ITS-VWP-2009-8
August 2009

Ev
aluation of Traffic Data Obtained via GPS-enabled Mobile
Phones: the Mobile Century field experiment
Juan C. Herrera
1
, Daniel B. Work
2
, Ryan Herring
3
,
Xuegang (Jeff) Ban
4
, and Alexandre M. Bayen
2
Abstract
The growing need of the driving public for accurate traffic information has spurred
the deployment of large scale dedicated monitoring infrastructure systems, which
mainly consist in the use of inductive loop detectors and video cameras. On-board
electronic devices have been proposed as an alternative traffic sensing infrastructure,
as they usually provide a cost-effective way to collect traffic data, leveraging existing
communication infrastructure such as the cellular phone network. A traffic monitoring
system based on GPS-enabled smartphones exploits the extensive coverage provided
by the cellular network, the high accuracy in position and velocity measurements pro-
vided by GPS devices, and the existing infrastructure of the communication network.
This article presents a field experiment nicknamed Mobile Century, which was con-
ceived as a proof of concept of such a system. Mobile Century included 100 vehicles
carrying a GPS-enabled Nokia N95 phone driving loops on a 10-mile stretch of I-880
near Union City, California, for 8 hours. Data were collected using virtual trip lines,
which are geographical markers stored in the handset that probabilistically trigger
position and speed updates when the handset crosses them. The proposed prototyp e
system provided sufficient data for traffic monitoring purp oses while managing the pri-
vacy of participants. The data obtained in the experiment were processed in real-time
and successfully broadcast on the internet, demonstrating the feasibility of the pro-
posed system for real-time traffic monitoring. Results suggest that a 2-3% penetration
of cell phones in the driver population is enough to provide accurate measurements of
the velocity of the traffic flow.
1 Introduction
Before the era of the mobile internet, characterized in particular by the emergence of
location based services heavily relying on GPS, the traffic monitoring infrastructure has
1
Corresp
onding author: Assistant Professor. Departamento de Ingenier´ıa de Transporte y Log´ıstica,
Pontificia Universidad Cat´olica de Chile. 2. Systems Engineering, Department of Civil and Environmen-
tal Engineering, UC Berkeley. 3. Department of Industrial Engineering and Operations Research, UC
Berkeley. 4. Department of Civil and Environmental Engineering, Rensselaer Polytechnic Institute (RPI)
1

mainly
consisted of dedicated equipment, such as loop detectors, cameras, and radars.
Installation and maintenance costs prevent the deployment of these technologies for the
entire arterial network and even for highways in numerous places around the world. More-
over, inductive loop detectors are prone to errors and malfunctioning (daily in California,
30% out of 25000 detectors do not work properly [1]).
For this reason, the transportation engineering community has looked for new ways
to collect traffic data to monitor traffic. Electronic devices traveling onboard cars are
appealing for this purpose, as they usually provide a cost-effective and reliable way to
collect traffic data.
Radio-frequency identification (RFID) transponders, such as Fastrak in California or
EZ-Pass on the East Coast
2
, can be used to obtain individual travel times based on vehicle
re-identification [2], [3]. Readers located on the side of the road keep record of the time
the transp onder (i.e. the vehicle) crosses that location. Measurements from the same
vehicle are matched between consecutive readers to obtain travel time. The fundamental
limitations of this system is the cost to install the infrastructure (readers), its limited
coverage, and the fact that only travel time between two locations can be obtained.
License Plate Recognition (LPR) systems are composed of cameras deployed along the
roadway which identify license plates of vehicles using image processing techniques. When
a vehicle is successfully identified crossing two sensors, a measurement of the vehicle’s
travel time is obtained. Example deployments include TrafficMaster’s passive target flow
management (PTFM) on trunk roads in the United Kingdom [4], and Oregon DOT’s
Frontier Travel Time project [5]. Like RFID systems, LPR system coverage is limited by
the cost to deploy the cameras.
Global Positioning System (GPS) devices found in the market can obtain position and
instantaneous velocity readings with a high accuracy, which can be used to obtain traffic
information. In [6] the authors addressed some of the key issues of a traffic monitoring
system based on probe vehicle rep orts (position, speeds, or travel times), and concluded
that they constitute a feasible source of traffic data. In [7] the authors also investigated
the use of GPS devices as a source of data for traffic monitoring. Two tests were performed
to evaluate the accuracy of the GPS as a source of velocity and acceleration data. The
accuracy level found was good, even though the selective availability
3
feature was still
on. The main drawback of this technology is that its low penetration in the population is
not sufficient to provide an exhaustive coverage of the transportation network. Dedicated
probe vehicles equipped with a GPS device represent added cost that cannot be applied
at a global scale. An example of such program at a small scale is HICOMP
4
in California,
which uses GPS devices in dedicated probe vehicles to monitor traffic for some freeways
and major highways in California. However, as pointed out in [8], the penetration of
HICOMP is low and the collected travel times are not as reliable as other systems such
2
F
astrak and EZ-Pass are electronic transponders used to pay road tolls electronically.
3
Selective availability is the intentional inclusion of positioning error in civilian GPS receivers. It was
introduced by the Department of Defense of the U.S. to prevent these devices from being used in a military
attack on the U.S. This feature was turned off on May 1, 2000.
4
HIghway COngestion Monitoring Program.
http://www.dot.ca.gov/hq/traffops/sysmgtpl/HICOMP/index.htm
2

as
PeMS. Other approaches have investigated the possibility of using dedicated fleets of
vehicles equipped with GPS or automatic vehicle location (AVL) technology to monitor
traffic [9, 10, 11], for example FedEx, UPS trucks, taxis, buses or dedicated vehicles. While
industry models have been successful at gathering substantial amounts of historical data
using this strategy, for example Inrix, the use of dedicated fleets always poses issues of
coverage, penetration, bias due to operational constraints and specific travel patterns.
Nevertheless, it appears as a viable source of data, particularly in large cities.
In the era of mobile internet services, and with the shrinking costs and increased
accuracy of GPS, probe based traffic monitoring has become one of the next arenas to
conquer by industries working in the field of mobile sensing. Increasing penetration of
mobile phones in the population makes them attractive as traffic sensors, since an extensive
spatial and temporal coverage could potentially soon be achieved. GPS-enabled cellular
phone based traffic monitoring systems are particularly suitable for developing countries,
where there is a lack of resources for traffic monitoring infrastructure systems, and where
the penetration rate of mobile phones in the population is rapidly increasing. By the end
of 2007, the penetration rate of mobile phones in the population was over 50% in the
world, ranging from 30-40% in developing countries (with an annual growth rate greater
than 30%) to 90-100% in developed countries [12].
Multiple technological solutions exist to the localization problem using cell phones.
Historically, the seminal approach chosen for monitoring vehicle motion using cell phones
(prior to the rapid penetration of GPS in cellular devices) uses cell tower signal infor-
mation to identify handset’s location. This technique usually relies on triangulation,
trilateration, tower hand-offs, or a combination of these. Several studies have investi-
gated the use of mobile phones for traffic monitoring using this approach (see for example
[13], [14], [15], [16] and [17]). The fundamental challenge in using cell tower information
for estimating position and motion of vehicles is the inherent inaccuracy of the method,
which poses significant difficulties to the computation of speed. Several solutions have
been implemented to circumvent this difficulty, in particular by the company Airsage,
which historically developed its traffic monitoring infrastructure based on cell tower in-
formation [18, 19, 20]. Based on the time difference between two positions, average link
travel time and speed can be estimated. In [21], the authors conduct a field experiment
to compare the performance of cell phones and GPS devices for traffic monitoring. The
study concludes that GPS technology is more accurate than cell tower signals for tracking
purposes. In addition, the low positioning accuracy of non-GPS based methods prevents
its massive use for monitoring purposes, especially in places with complex road geometries.
Also, while travel times for large spatio-temporal scales can be obtained from such meth-
ods, other traffic variables of interest, such as instantaneous velocity are more challenging
to obtain accurately.
A second approach is based on GPS-enabled smartphones, leveraging the fact that
increasing numbers of smartphones or PDAs come with GPS as a standard feature. This
technique can provide more accurate location information, and thus more accurate traffic
data such as speeds and/or travel times. Additional quantities can potentially be obtained
from these devices, such as instantaneous velo city, acceleration, and direction of travel.
In [16], the authors use cell phone for traffic monitoring purposes, and mention the need of
3

Frequently asked questions

Q1. What have the authors contributed in "Evaluation of traffic data obtained via gps-enabled mobile phones: the mobile century field experiment" ?

A traffic monitoring system based on GPS-enabled smartphones exploits the extensive coverage provided by the cellular network, the high accuracy in position and velocity measurements provided by GPS devices, and the existing infrastructure of the communication network. This article presents a field experiment nicknamed Mobile Century, which was conceived as a proof of concept of such a system. The proposed prototype system provided sufficient data for traffic monitoring purposes while managing the privacy of participants. Results suggest that a 2-3 % penetration of cell phones in the driver population is enough to provide accurate measurements of the velocity of the traffic flow. 

Q2. How can the authors determine which velocity measurements are more likely to be closer to ground truth?

Velocity fields constructed using 17 VTLs and 17 loop detector stations can be integrated to compute travel time11, which can be used to assess which velocity measurements are more likely to be closer to ground truth. 

Q3. How many KB of update data is generated per day?

The current VTL implementation generates approximately 1KB of update data for every two minutes per client while driving on a major road. 

Q4. What is the main argument for the use of cell phones in traffic monitoring?

Increasing penetration of mobile phones in the population makes them attractive as traffic sensors, since an extensive spatial and temporal coverage could potentially soon be achieved. 

Q5. How can the authors artificially recreate VTL data at different locations?

Note that since all vehicle trajectories can be reconstructed, it is possible to artificially recreate VTL data off-line at different locations. 

Q6. How many standard deviations are produced from the velocity fields produced from VTLs?

if the velocity fields produced from VTLs and loop detector data are integrated to estimate travel times, the travel times produced from VTLs are more likely to fall within one standard deviation of the mean travel time observed in the field. 

Q7. what are the features of interest for traffic monitoring systems such as mobile millennium?

Specific features of interest for traffic monitoring systems such as Mobile Millennium include travel time on a link or a route, robust range of arrival time, variance in travel time along a link or a route. 

Q8. What are the main reasons why electronic devices are appealing for traffic monitoring?

Electronic devices traveling onboard cars are appealing for this purpose, as they usually provide a cost-effective and reliable way to collect traffic data. 

Q9. Why do the different methods of VTL measurements exhibit more variability than loop detector measurements?

Because of the different 5-minute aggregation methods used, VTL measurements exhibit more variability than loop detector measurements. 

Q10. What is the main issue with the use of GPS?

Another issue is the knowledge of vehicle position and velocity provided by this technology, which needs to be used in a way which does not infringe privacy. 

Q11. What is the fundamental challenge in using cell tower information for estimating position and motion of vehicles?

The fundamental challenge in using cell tower information for estimating position and motion of vehicles is the inherent inaccuracy of the method, which poses significant difficulties to the computation of speed. 

Q12. Why were the loop detectors not used as benchmark?

For this reason, loop detector velocity data were not used as benchmark, and only a comparison with travel times was carried out to assess accuracy of the data.