INSTITUTE OF TRANSPORTATION STUDIES
UNIVERSITY OF CALIFORNIA, BERKELEY
High Accuracy Location Based Services Cost
Benefit Study: Final Report
Adam Goodliss, Christian Manasseh, Venkatesan
Ekambaram, Raja Sengupta, Adib Kanafani,
Kannan Ramchandran
Research Report
UCB- ITS- RR- 2011- 1
May 2011
1
2
Table of Contents
Figures........................................................................................................................ ......................................... 4
Tables ............................................................................................................................... ................................... 5
Acknowledgment ............................................................................................................................... ................ 6
I. Executive Summary ............................................................................................................................... .... 7
II. Motivation..................................................................................................................... ............................. 9
III. Literature Summary ............................................................................................................................ 11
A. Market Analysis ............................................................................................................................... ... 11
B. Published Benefit Analysis Reports of Various GNSS........................................................................ 11
C. Existing C‐ HALO Type Deployments................................................................................................... 13
IV. Benefit Assessment ............................................................................................................................. 18
A. Assumptions ............................................................................................................................... ........ 19
B. Sources of Data........................................................................................................................... ........ 20
C. Safety Applications .............................................................................................................................. 20
1. Curve Speed Warning ..................................................................................................................... 20
2. Forward Collision Warning............................................................................................................. 21
3. Merge/ Lane Change Warning........................................................................................................ 22
4. Intersection Collision Warning....................................................................................................... 23
5. Left Turn Assistant .......................................................................................................................... 24
6. Stop Sign Movement Assistant ...................................................................................................... 25
7. Highway/ Rail Collision Warning..................................................................................................... 26
D. Mobility Applications .......................................................................................................................... 27
1. Intelligent Traffic Flow Controls ( ITFC) .......................................................................................... 27
2. Free Flow Tolling........................................................................................................................ .... 27
E. Efficacy Literature Caveat ................................................................................................................... 28
F. Summary of Benefits ........................................................................................................................... 29
V. Separating HALO Costs into Good and Bad ( Dark) GPS areas .............................................................. 30
A. Data for Modeling Satellite Count...................................................................................................... 32
B. Hidden Markov Model to Predict Satellite Counts ........................................................................... 35
C. Use of the Model ............................................................................................................................... . 38
3
VI. Cost Assessment in Good GPS Areas ................................................................................................. 40
A. N‐ RTK Base Station Cost Estimation .................................................................................................. 41
VII. Further Research ............................................................................................................................... . 42
VIII. References ............................................................................................................................... ........... 43
Appendix A – Market Presence ( X‐ Major; O‐ Minor) ...................................................................................... 47
Appendix B – Process of Selecting ITS Application that Stand to Benefit from C‐ HALO.............................. 48
A. Initial Application Selection Matrix ( 1,2) ........................................................................................... 48
B. Applications that stand to benefit from C‐ HALO .............................................................................. 51
1. Safety ............................................................................................................................... ............... 51
2. Mobility....................................................................................................................... .................... 52
3. Emissions ............................................................................................................................... ......... 52
4. Mobility & Emissions ...................................................................................................................... 52
C. References ............................................................................................................................... ........... 52
Appendix C – Querying Methodology Matrix ................................................................................................. 53
Appendix D – Efficacy Rate Matrix................................................................................................................... 54
Appendix E – Cost Estimation Emails............................................................................................................... 55
4
Figures
Figure 1 HA‐ NDGPS Coverage Area ................................................................................................................. 14
Figure 2 N‐ RTK Deployments Reviewed ( 36 – 46) .......................................................................................... 15
Figure 3 Leica N‐ RTK Service Area map ( lower portion show covered spots in a three states: California,
part of Nevada and part of Arizona)................................................................................................................ 17
Figure 4 Technology adoption curve ............................................................................................................... 19
Figure 5 Benefits by Category .......................................................................................................................... 29
Figure 6 Shaded area represents study area .................................................................................................. 31
Figure 7 3D rendering showing building coverage in validation area ........................................................... 32
Figure 8 GPS data collection depicting satellite counts: > 6 are shown in green, between 4 and 6 are
shown in yellow, < 4 are shown in red ............................................................................................................. 33
Figure 9 Empirical and predicted HDOP values with 95% confidence intervals........................................... 34
Figure 10 Mask‐ Angle Representation using street width and building heights.......................................... 35
Figure 11 HMM for predicting satellite count ................................................................................................ 37
Figure 12 GPS data estimating satellite counts: > 6 or more are shown in green, between 4 and 6 are
shown in yellow, < 4 are shown in red ............................................................................................................. 38
Figure 13 Overlay of accident data on satellite count projection data......................................................... 39
5
Tables
Table 1 State N‐ RTK Deployments ( 36 – 46) ................................................................................................... 15
Table 2 ITS Applications that Benefit from C‐ HALO Deployment.................................................................. 18
Table 3 Injury Worth Percentages ................................................................................................................... 20
Table 4 Sources of GPS Errors .......................................................................................................................... 30
Table 5 Model Prediction Accuracy ................................................................................................................. 36
Table 6 HMM Satellite Count Accuracies ........................................................................................................ 37
Table 7 N‐ RTK Cost Estimation ( No Other ( Contingency) Costs) ( 47,48,49.60)............................................ 41
6
Acknowledgment
The authors gratefully acknowledge a gift from ATLIS Wireless, LLC that has, in part, made this work
possible. The contents of this report are determined solely by its authors.
7
I. Executive Summary
This report presents a benefit and cost study of a Cooperative High‐ Accuracy LOcation ( C‐ HALO) service.
The Department of Defense operating the GPS constellation guarantees a location service accurate to 7
meters 97% of the time. Using this as a baseline, we define the High Accuracy Location service to be sub‐meter
( alternatively, decimeter) level accuracy. This report is about the costs and benefits of realizing a
location service with sub‐ meter accuracy in the United States.
There are several GPS augmentation technologies aimed at enhanced accuracy. Some such as Inertial
Navigation Units are on the rovers and entail no new infrastructure. Others such as DGPS, N‐ RTK or
HAN‐ DGPS, build new infrastructure to assist new rover units. In our definition, these augmentations
only constitute a high accuracy location service, if they entail a promise like the one provided by the
DoD for the GPS constellation, e. g., Infrastructure X provides a location service accurate to 1 meter or
better Y% of the time. A High Accuracy Location service should be ubiquitous and reliable, like its LALO1
equivalent, the GPS constellation.
The business model of the Inertial Navigation Unit industry envisages no such guarantee. The ubiquity of
any higher accuracy outcome of GPS/ INS fusion is left entirely to the purchaser of the unit, if anybody at
all. On the other hand, our literature and market survey suggests, the N‐ RTK market does entail
guarantees or promises like the one from the DoD for GPS. The N‐ RTK service providers seek to assure
customers that the ubiquity or reliability of their high accuracy location service will be adequate. HAN‐DGPS
is a public sector infrastructure build that could also provide assurances. Hence this report
discusses the details and salient features of these infrastructures. We do not study the INS industry.
We have surveyed the literature on location services, and discovered a body of work on the benefits of a
high accuracy location service. The literature establishes the benefits of high accuracy location to
industries such as agriculture, construction, mining, and aerospace. Chapter III summarizes this
literature. We focus on road travel and the findings suggest this focus is worthwhile. Chapter IV of this
report quantifies the benefits of HALO services to road travel, and estimate benefits to be between $ 160
billion and $ 320 billion. This translates into 1.1 to 2.2 percent of the current US GDP. The estimated
benefits are gross benefits over a time horizon of 22 years discounted to the present day value. They do
not account for the costs of implementation. The range depends on whether one uses the low‐ level or
mid‐ level efficacy assumptions. The benefits arise from smoother traffic flow on the roads leading to
reduced travel delays, and fewer accidents leading to reduced injuries and saved lives. We expect HALO
to also yield climate‐ change and public health benefits through its enabling of smoother traffic flow but
have been unable to quantify these benefits for this report.
The assessment of the costs associated with these benefits has two challenges. One must assess where
new infrastructure is required and how much each installation might cost. To understand where HALO
1 Low Accuracy LOcation
8
services need new infrastructure it helps to think of places with “ good” GPS coverage as distinct from
“ bad” GPS coverage. This is the subject of Chapter V. The chapter presents a way to process data from a
county assessor’s office and divide the county into good and bad GPS areas. The method is illustrated by
application to San Francisco. When 6 or more satellites are visible and they are well spread out, GPS
accuracy is good and enhancements such as N‐ RTK or HAN‐ DGPS make it even better, resulting in
decimeter or even centimeter level accuracy. When the number of GPS satellites is lower, GPS errors
can be over a meter and these augmentations are less effective. Some other kind of technology, perhaps
pseudolites, is required. We call the area not well served by the GPS constellation, the “ bad GPS” area or
“ dark area.” We expect tall buildings in urban areas to reduce satellite visibility on streets. Likewise, tall
trees can occlude satellites. Thus urban canyons or parkways in wooded area would be part of the dark
area or bad GPS coverage areas.
The existence of the N‐ RTK market makes the distinction between good and bad GPS areas useful. Since
N‐ RTK realizes HALO in good GPS areas, HALO services for good GPS areas are more advanced than
HALO services for bad GPS areas. There are no significant location services sold in the US consumer
market for bad GPS areas. Therefore the work in chapter V computing the percentage of the road area
that is “ good” and “ bad” GPS coverage provides insight into how much of HALO needs fundamentally
new location infrastructure that is not GPS augmentation, and how much could potentially be achieved
by leveraging GPS augmentation methods such as HAN‐ DGPS or N‐ RTK market.
Finally, for good GPS areas, chapter V presents infrastructure cost numbers based on N‐ RTK. There are
enough N‐ RTK deployments to collect and compare these numbers. We assess the current cost to be
$ 560,000 to $ 1.6 million per base station over a time horizon of 22 years covering a 60x60 sq. km area. A
simple linear scaling, puts this at 1.6 to 4.4 billion nationwide, of the order of 1% of the benefits of HALO
estimated to accrue from road travel alone. We have been unable to do the same for bad GPS areas.
The good and bad GPS coverage classification method in chapter V is GIS based. It produces a map with
good and bad areas marked on it. Other GIS data can be over‐ layed to visualize the value of enabling
HALO services in a particular place. We overlay accident data. One can spot the bad GPS streets
coincident with higher accident rates. The idea is to spot the streets where one gets a bigger bang for
the buck. The other good news in our findings, is that out of the total 121 sq km. area of San Francisco
( 121 sq. km), only 0.3 to 4% of the San Francisco streets fall in the “ dark” area. This range is the 95%
confidence interval. Since the dark areas are few and we can spot them, one can target new investment
to areas yielding safety or mobility benefits. If an area is estimated by us to be dark, we know new high
accuracy location infrastructure beyond GPS augmentation methods is required with high confidence.
Our model does not say the converse with high confidence, i. e, GPS augmentation technologies like N‐RTK
will work in the “ light” areas. This is because even in the “ good” GPS areas with sufficient satellite
coverage the errors can be quite large due to multipath effects.
9
II. Motivation
The study is motivated by the possibility of the benefits of High Accuracy ( sub‐ meter) Location being
large, such a service being technologically feasible, and deployable at reasonable cost.
The high benefit argument rests on research showing vehicle‐ based collision avoidance systems can
reduce accidents ( 8). The systems require location services accurate enough for a vehicle to discriminate
its lane, which in turn requires sub‐ meter accuracy ( 30). Likewise high‐ speed tolling or intelligent traffic
light control also requires lane discrimination and sub‐ meter accuracy. Since the economic cost of lives
lost in accidents is high as is the economic cost of delays due to traffic congestion, the benefits of HALO
as a critical enabler of services mitigating accidents or congestion should also be high.
The technological feasibility argument rests on the proliferation of GNSS2 systems and services. GPS has
become a stable component of global economic activity. Other GNSS such as GLONASS and Galileo, are
becoming established. Technologies such as DGPS, GPS‐ WAAS, GPS+ INS, GPS‐ RTK, and Network RTK( 2,3)
have been developed and partially deployed in recent years. Limited‐ coverage pseudolite‐ based
systems are also available ( 61‐ 63). Ultra Wide Band systems coupled with RTK ( 66,67) are being
proposed to achieve high levels of accuracy in multipath rich environments. Wide‐ area multi‐ lateration
systems ( 4) are being substantially deployed around airports for aircraft and ground vehicle tracking.
Inertial Navigation Systems are rapidly dropping in cost even as the rise in accuracy. All of these suggest
there may be many options in the market or near‐ to‐ market that might be leveraged to realize a
ubiquitous and reliable HALO service.
Many of these new location options will rely on the ubiquitous availability of mobile communication
services. HAN‐ DGPS requires ubiquitous broadcast communication from base station to mobile, and N‐RTK
ubiquitous unicast communication between infrastructure and mobiles. Here too the rise of 3G, 4G
( 65), Dedicated Short Range Communications ( DSRC) ( 64), software defined radio ( 5), smart antenna
systems ( 6), and other techniques, along with use of vehicles as excellent mobile communication
platforms ( 7), may provide for advancements in positioning techniques and integrated systems.
The market analysis and literature review section indicates a lack of coverage of the benefits of C‐ HALO
technology to the Transportation domain in general and ITS in particular. From the literature presented,
efforts exist that attempt to quantitatively estimate the costs and benefits of high accuracy location
data. These efforts are mostly market wide, or specific to an individual industry that is not
transportation. In addition, plenty of literature exists to understand the costs and benefits of
implementing certain ITS applications like curve speed warnings ( 8), or intelligent signal control systems
( 9). This later effort usually addresses the particular application under study and does not analyze the
monetary benefits to the market; their results are presented in number of accidents reduced or total
vehicle miles saved by the deployment of the application.
2 Global Navigation Satellite System
10
This lack of coverage of the transportation industry motivates us to tackle the job of estimating the
economic benefits that could be reaped from a C‐ HALO nationwide deployment to the ITS sector of the
transportation domain. It is the objective of this analysis to develop an exhaustive list of ITS applications
that require high accuracy location data to realize the costs of implementation and the expected
economic benefits that accompany each application.
This study aims at providing a tool enabling government and private funding agencies to assess the
benefits of investing in a new breed of positioning technologies and wide‐ scale deployments to meet
the goals first noted above. It will shed light on the range of costs of the most promising technologies,
their integration, and phases of deployments leading to a nationwide C‐ HALO infrastructure.
11
III. Literature Summary
We have reviewed existing GNSS related market analyses and cost benefit studies done by others on
various sectors of the economy and in various parts of the globe. This section summarizes our findings.
A. Market Analysis
Rob Lorimer of Position One Consulting performed a three year projection on the GNSS global market in
his report titled: GNSS Market Research and Analysis September 2008 ( 10). Based on this report and
analysis, we created a table of global positioning companies, along with which industry( ies) each
company is involved in. The complete table is included as Appendix A.
The table identifies the three most ubiquitous providers of GNSS‐ based services as Leica Geosystems,
Trimble, and TopCon/ Sokkia. Omnistar is also relevant in many industries, but they are mainly focused
on precision augmentation services, while the other three are more vertically integrated, and typically
incorporate numerous levels of the value chain. Interviews conducted by Lorimer with the CEO’s of the
companies listed in Appendix A ( 10) provided insight into the industries that are major consumers of
location services. The biggest consumers are the Aerospace, Agriculture, Autonomous Vehicles,
Construction, Defense, Maritime, Mining, and Surveying industries. Clearly void from this list is the
transportation sector, which we choose to analyze as part of this CBA. Benefit estimates have been
completed in some of these industries and are discussed in more detail in the subsequent section.
B. Published Benefit Analysis Reports of Various GNSS
The Allen Group ( 11) estimated the economic benefits of C‐ HALO type technology in three specific
Australian industries: Agriculture, Mining, and Construction. The Allen Group determined the benefits to
be between $ 100 and $ 200 billion, approximately 10 to 20 percent of Australian GDP. These three
markets make up approximately 10 to 13 percent of GDP. Assuming that the U. S. transportation market
makes up 5 percent of GDP, a simple linear scaling of the Allen Group’s numbers suggests the HALO
benefits derived from the transportation sector alone should be 4 to 9 percent of GDP, which would be
approximately $ 560 to $ 1.2 billion in benefits. We find $ 160 to 300 billion. These benefit numbers
appear conservative in relation to the Allen Group study.
We have incorporated a key piece of the Allen Group report in our method. The adoption rate for the C‐HALO
technology is represented by this studies’ industry‐ wide national rollout adoption scenario.
A socio‐ economic benefit study was commissioned by US Department of Commerce ( DoC) ( 12) to
determine where there is value added by the CORS and GRAV‐ D systems. The study is focused on the
benefits derived from the increased vertical accuracy of GPS. We do not consider this dimension at all.
The study suggests the surveying and mapping industry will be the most significantly impacted, but goes
on to list other possible industries like construction, agriculture, environmental science, and
transportation. Again this reiterates the fact that researchers are continuing to view transportation as a
realm for potential benefits from C‐ HALO technology. The US DoC study assesses benefits utilizing the
12
productivity methodology, which is typical and similar to the methodology used in our study and many
others contained in the literature review. One slight difference to our methodology is that their time
horizon is 15 years while ours is 20 years.
Alcantarilla, et al. analyze the benefits of a multi‐ constellation system, versus a stand‐ alone GNSS
system, and ultimately a SBAS approach ( 14). A piece that may be of importance to us when discussing
the costs is the distribution of the number of satellites in view. They conduct a simulation of an urban
environment and contend that with GPS & Galileo 65% of the area is covered by more than 3 satellites,
while 20% is covered by 3, and 15% by less than 3. They then go on to qualitatively discuss the principal
pieces of a future GPS system along with the envisioned benefits of multi‐ constellation GNSS SBAS
augmentations. Similar analysis is carried out by Zabic et al. ( 68) but with actual data in Copenhagen.
They estimate the average satellite availabity in Copenhagen through extensive data collection and use
simulation tools to predict the improvement in satellite availability with the addition of Galileo.
Swann, et al. discuss the qualitative benefits of location‐ based services, the architectural issues involved
in multi‐ constellation systems, and the market aspects that need to be addressed for deploying multi‐constellation
systems ( 15). They focus on the benefits of reliability of a combined GPS/ Galileo signal
where availability is at 99.7% in their Stuttgart analysis. In addition, they estimate the GNSS service
provision market to be 135 billion Euros by 2015 with a significant portion of that residing in the
transportation industry. This is significantly higher than what Lorimer’s report quotes for the U. S.
market by 2012, which is around $ 9 billion.
Vollath, et al. aimed to look at how NRTK and the third frequency to be offered by Galileo will interact
( 16). Vollath et al. present the value of the Galileo third frequency in allowing higher horizontal accuracy
and increasing distances between base stations among other things. NRTK, however, still proves to be
more accurate in the vertical direction. Ultimately, they do not assess the monetary benefits, but only
the technical reliability. They conclude that NRTK will not be replaced by the Galileo new third
frequency, but that the two could be used as complimentary technologies.
Arthur, et al. delve deeper into the impacts of Galileo by going beyond cost benefit analyses and
conducting specific input‐ output models ( 17) which actually predict economic output rather than just
analyzing costs and benefits. They even go as far to suggest that some ‘ market externality’ impacts, like
induced effects, could be twice as large as the direct impacts. They also suggest how to enhance a CBA
by including innovation effects ( through supply‐ push or demand‐ pull forces), or market and social
externalities. These types of analyses could be worthwhile as future work. They are not included in this
report.
Brennan, et al. wrote National PNT Architecture: Interim Results to facilitate the decision making
process on a national PNT architecture for the United States by 2025( 18). It does not focus on costs or
benefits in quantitative terms. It does however evaluate many different technological options to achieve
their stated goals. Ultimately, they want to put together a transition plan from an “ as is” architecture to
a “ should be” architecture. Unfortunately, this is not directly related to our CBA.
13
C. Existing C­HALO
Type Deployments
In order to understand the existing C‐ HALO deployments and technologies, we reviewed the initiatives
undertaken by the government agencies. The material here is based on reports ( 56) and ( 59) provided
by the Federal Highway Authority. The earliest deployments were the Differential GPS ( DGPS) base
stations by the US Coast Guard for maritime services. These base stations broadcast the actual and
measured pseudo‐ range differences of the received code measurements from the different satellites.
These error measurements are used by GPS receivers to calibrate their own measurements resulting in
accuracies as high as 1m under good line‐ of‐ sight conditions. The corrections are broadcast typically in
the longwave frequency range between 285kHz and 325kHz. The U. S. Army Corps of Engineers
( USACOE) later realized the benefits of accurate localization and efforts were made to increase the
coverage of the DGPS base stations ( 56). This resulted in the N‐ DGPS or nationwide DGPS program
under which a total of around 137 base stations were to be installed nationwide to provide accurate
localization services. The defense establishment also found need for decimeter and centimeter level
accuracies. This could be obtained by sending corrections to the carrier phase received by the DGPS
stations, as the carrier frequency of GPS is 1000 times higher than the frequency of the modulated code
sequence. Hence one could obtain very high accuracies by measuring the carrier phase. This technology
came to be known as RTK or Real Time Kinematic positioning and the proposed system implementation
by government agencies has come to be known as HA‐ NDGPS – High Accuracy NDGPS ( 56).
One of the challenges of HA‐ NDGPS is that the allocated bandwidth does not suffice for broadcasting the
carrier measurements for all the satellites ( 59). This requires compression of the phase measurements.
This work is still in progress. Prototypes of this system were deployed and evaluated ( 56). During
deployment it was found that, if a receiver obtained corrections from more than one base station, a
combination of the measurements provided higher accuracies. More sophisticated combination could
provide still higher accuracies, and this is the proprietary technology used in N‐ RTK or Network RTK, a
service, provided by companies such as Leica, Trimble etc. The N‐ RTK service has two methods of
operation ( 71). The Virtual Reference Station ( VRS) method as adopted by agencies like Trimble is a
unicast system where the GPS receiver contacts a central server, which in turn computes the corrections
from the set of receiver stations in the vicinity of the receiver and gives an estimate of the receiver’s
location. The Master Auxiliary Concept ( MAC) method allows for a broadcast system wherein a single
master reference station amongst a cluster of reference stations in a cell, broadcasts the corrections.
The rover in turn interpolates these corrections to estimate the corrections at its location. The MAC
method also allows for a two‐ way mode where the reference station calculates the corrections for the
rover as in the case of VRS. In our opinion, the question of whether one would want to adopt a unicast
system or a broadcast system depends on the application. For a large‐ scale application like Intelligent
Transportation Systems, it might be desirable to have a broadcast system and have all the intelligent
processing done at the GPS receiver as compared to a central server. If every vehicle is required to know
its location accurately, it is more efficient to broadcast the error measurements to all the vehicles in
contrast to every vehicle contacting a centralized server to compute its location estimate since the error
measurements would be common to all the vehicles in a particular region of interest.
14
HA‐ NDGPS is the technology that is being standardized by the federal DOT as the technology of choice
for achieving high accuracy positioning for ITS applications. The federal DOT has commissioned a couple
of pilot programs to improve on this technology to achieve cm level accuracies nationwide. The pilot
sites are in Maryland and Pennsylvania and the research is being headed by the Turner Fairbanks
Highway Research Center. The current and planned coverage areas are in the map below:
Figure 1 HA‐ NDGPS Coverage Area
Additionally, we have researched state run, cooperative and private run positioning and augmentation
services. Most of these services are N‐ RTK corrections. Figure 2 shows the states with N‐ RTK
deployments we have found as of December 2010.
15
Figure 2 N‐ RTK Deployments Reviewed ( 36 – 46)
The red states denote N‐ RTK deployments partnered with Trimble, while the blue states denote N‐ RTK
deployments partnered with Leica. The green states partners were either unidentifiable or only
explored, but never actually deployed, an N‐ RTK the network. Within this group of states the State run
programs and Private/ Public Cooperatives are as follows:
Table 1 State N‐ RTK Deployments ( 36 – 46)
State DOTs Public/ Private Cooperatives
Utah Texas
Ohio Washington
Iowa Midwest ( Indiana)
Oregon Alabama
California
Michigan
Minnesota
Wisconsin
Throughout these deployments there are many similarities in infrastructure. The first implementations
were in the early 1990s and have continued through the 2000s. From an infrastructure standpoint the
industry standard seems to place N‐ RTK base stations 60km to 70km apart. Most of the deployments
have from 50 to 80 base stations. Some of the cooperative deployments continue to grow due to
increasing membership, and in addition, some of the nascent state DOT’s deployments also have
16
expansion plans in place. All of the deployments offer centimeter level accuracy within their network.
( 36 – 46)
The networks differ in their access rules. Currently all state DOT networks charge no fee for usage,
except for Utah, which just changed policies and began charging $ 400 annually. The cooperative
networks typically charge between several hundred and several thousand dollars annually. On top of
this, users must purchase a receiver and applicable cellular plan for the data flow. Cellular plans typically
range in the order of $ 100 while receivers range from several hundred to several thousand depending
on capability. ( 36 – 46)
These costs seem bearable by markets such as Agriculture, Surveying, and Construction services, due to
their high use of these state‐ run and cooperative networks. Only one state, Minnesota, had
implemented and deployed N‐ RTK for transportation purposes. They use the network for snowplows
and inner city bus routes. ( 42)
Three states were questioned for cost information: Iowa, Ohio and Washington. These systems range
between $ 50K and $ 115K in expenditures per base station to perpetuity. These costs are discussed in
further detail in a separate chapter.
To gain further understanding of the availability of C‐ HALO services, we review private services offered
by Omnistar and Leica. ( 47,48 & 49)
Leica has SmartNet, which is N‐ RTK coverage, in many states across the United States. Based on
SmartNet’s service agreement ( 54, 55), Leica offers 1‐ 2 cm horizontal accuracy and 2‐ 3 cm vertical
accuracy under conditions of good satellite coverage, good geometry, and low multipath environments.
However we have not been able to locate, from Leica, the percentage of time those conditions are
satisfied within their areas of coverage. The map below documents many of the states in which Leica
has some private coverage available. Typically this coverage is provided through private investment, and
partnerships with other Leica network deployments. The service agreement ( 54) explicitly mentions that
Leica geosystems disclaims warranty to the accuracy of the data created by or passing through the
SMARTNET Reference Station Network. Omnistar currently claims 99% availability of C‐ HALO services in
the United States. This is offered using DGPS technology and entails an annual subscription service as
well as investment in a GPS receiver. The subscription services range from $ 800 for the least accurate
( sub‐ meter) to $ 2500 for the most accurate ( centimeter) per receiver. The receivers generally cost
around $ 5000 and are available from Trimble, Novatel, Raven, Topcon and others ( 50).
17
Figure 3 Leica N‐ RTK Service Area map ( lower portion show covered spots in a three states: California, part of Nevada and
part of Arizona).
18
IV. Benefit Assessment
Our approach is to determine a suite of ITS applications that require a high accuracy location service,
find the benefits of these applications, and associate them to C‐ HALO. The ITS applications analyzed are
those listed by the FHWA ( 19). A comprehensive list of these applications appears in appendix B. This list
is analyzed for its location accuracy requirements. This filters down the application list to 8 groups of
applications. If the applications require 1m or less accuracy, the applications and their benefits are
analyzed, and associated to C‐ HALO.
Each application is explored independently to determine the efficacy rate, and the monetary benefit
from reducing accidents ( and in turn injuries and fatalities), VMT, travel times, emissions, and the like
depending on the application. This type of methodology is similar to those used in other CBA’s
completed by the USDOT and other international governmental agencies. The method we use takes into
account the cash flow estimates of the benefits over a 22 year period, and discounts those into
“ today’s” worth via a discount rate that is proposed for this type of analysis by the congressional
budgetary office. The analysis is similar to that adopted by the Allen Group ( 11).
The final list of applications can be seen in the Table 2 below:
Table 2 ITS Applications that Benefit from C‐ HALO Deployment
ITS Applications Type Included in
Benefit
Analysis
Curve Speed Warning Safety Y
Forward Collision/ Braking Warning Safety Y
Emergency Electronic Brake Lights
Cooperative Forward Collision Warning
Merge/ Lane Change Applications Safety Y
Highway Merge Assistant
Lane Change Warning
Blind Spot Warning
Blind Merge Warning
Left Turn Assistant Safety Y
Stop Sign Movement Assistant Safety Y
Highway/ Rail Collision Warning Safety Y
Intersection Collision Warning Safety Y
Corridor Management Mobility Y
Intelligent Traffic Flow Control
Free‐ Flow Tolling
19
A. Assumptions
Some overall assumptions have to be made to estimate the benefits. Overall assumptions cover
predictions we make about the national economy into the next 20 years, and general assumptions on
how the new technology would be adopted by the ITS sector. We later on make application‐ based
assumptions to estimate the particular efficacy of each application.
Technology Adoption Rate – The shape of this curve determines how quickly the fleet will adopt new
technology, in this case C‐ HALO. The s‐ curve used in this analysis is leveraged from a report, by the Allen
Group ( 11), which analyzes the benefits of high accuracy location data in non‐ ITS industries. The general
shape of the curve is in Figure 4.
Figure 4 Technology adoption curve
This curve is applied over a project horizon of 22 years, 2008 – 2030. In calculating benefits, this
adoption rate was typically used to determine the correct portion of benefits accumulated in a given
year.
Discount Rate – This rate is used to discount future cash values to current day terms by taking into
account inflation and a risk free rate of return, the higher the rate the more significant the discount to
future cash values. For this analysis, a discount rate of 5 percent is used, and is taken from the Office of
Budget and Management ( 20). They also suggest using a range from 3 to 7 percent.
Value of Time – The value of time is used in quantifying reductions in delay into monetary benefits.
Again, the Volpe study quotes two figures, one for local travel, $ 11.20, and the other for intercity travel
$ 15.60. These figures are from guidance from the office of the Secretary of Transportation ( 21). In our
analysis, we take both figures and average them since in our data. The resulting figure is $ 13.40.
20
Delay Growth – The delay growth is calculated using figures from the Traffic Congestion and Reliability
Report prepared by Cambridge Systematics for the FHWA in 2005 ( 22). Using twenty‐ year historical data
( hours of delay per traveler) and trend analysis a growth rate of 6.5 percent is calculated.
B. Sources of Data
Accident Data – For the Safety applications, all accident data is culled from the GES database ( 35), which
includes all types of accidents, not just accidents including fatalities. This database is then queried to
ensure the appropriate accidents are being accounted for with regards to each individual application.
Please see Appendix C for the querying methodology for each application class. We have also examined
the FARS database ( 23) which includes fatal accidents.
Accident Growth Rate – The accident growth rate is used to project accident counts for years 2009 –
2030. The Volpe VII report projects accident rates based on VMT estimates and increased safety
measures. These yearly accident rates are used to calculate the compound annual growth rate over the
project horizon ( 21) This rate is calculated to be ‐ 0.2 percent.
Fatality Worth – This value is used in determining the benefit of reducing the count of fatal accidents.
The Office of Management and Budgets put forth a memorandum in 2008 that suggests to the DOT that
$ 5.8 million be used for the value of a life. It also suggests using a range of $ 3.2 million to $ 8.4 million
( 24).
Injury Worth – These values are based on percentages of the fatality worth. Again there is a standard,
and that is the Maximum Abbreviated Injury Scale. Typically there are 5 injury levels not counting a
fatality ( 24). In the FARS database only three levels of injuries are reported not counting fatalities.
Therefore averages were taken first and second level and the third and fourth levels to determine the
three percentages used in this analysis. The percentages used are in Table 3.
Table 3 Injury Worth Percentages
Injury Worth (% of Fatality Worth)
Incapacitating 47.50%
Non‐ Incapacitating 5.80%
Possible/ Light Injury 0.90%
C. Safety Applications
As part of the safety analysis, seven applications are analyzed: Curve Speed Warning, Forward Collision
Warning, Merge/ Lane Change Warning, Left Turn Assistants, Stop Sign Movement Assistant,
Highway/ Rail Collision Warning, and Intersection Collision Warning. All of these applications are focused
on reducing accidents, and in turn fatalities and other injuries.
1. Curve Speed Warning
Curve speed warnings would aid drivers in negotiating curves at appropriate speeds. This is aimed at
reducing single and multi‐ vehicle accidents in curves due to unsafe speeds.
21
To quantify the benefits of such a system we aimed to determine the number of accidents that could be
reduced, then by using the assumptions laid out in previous sections, calculate a monetary benefit for
reducing accidents.
To begin this process, the GES database was queried for specific accident data related to the application
in question. For instance, all accidents that took place in curves, and were related to speed were
included in this analysis.
In 2008, there were 1048 fatalities, and ~ 29000 other injuries where this type of application may be
applicable. To determine the benefit of this system an efficacy rate must be determined to see how
much of a reduction from these figures can be expected.
Through another literature review, several reports were found discussing how effective curve speed
warnings could be. The three reports and results are summarized briefly below:
• Field Evaluation of the Myrtle Creek Advanced Curve Warning System ( Oregon DOT 2006) –
Empirical analysis of I‐ 5 implementation near Myrtle Beach, over 75 percent of people reduced
speeds entering the curves with dynamic message signage. The FHWA report ( 21) uses this value as
a measure of efficacy of the curve speed warning applications when assessing the benefits of
wireless communication to ITS.
• Rural ITS Toolbox ( FHWA 2001) – Empirical study for trucks in Colorado. Speeds were reduced by 25
percent.
• An Evaluation of Dynamic Curve Warning Systems in the Sacramento River Canyon: Final Report ( CA
DOT 2000) ‐ Empirical analysis of five locations on I‐ 5 in California, over 70 percent of people
reduced speeds entering the curves with dynamic message signage.
Using these sources as references, we chose to use 40% accident reduction as a mid‐ level efficacy rate.
A low level would be 20% while a high efficacy level would be 70%. For a matrix of the efficacy rates
please see Appendix D.
With all of this data, a discounted yearly monetary benefit can be calculated. The equation would as
follows:
Where B is monetary benefits, n is the year, j is the application, and i is the injury level ( fatal, serious,
etc.).
Using this formula and the low efficacy rate, preliminary benefits of ~$ 54 Billion were estimated.
2. Forward Collision Warning
Forward collision warnings alert a driver when a forward vehicle brakes hard ( deceleration is above a
predetermined threshold). This is very similar to Cooperative Forward Collision Warning which is used to
preemptively avoid rear‐ end collisions with vehicles in front of the subject vehicle.
22
To quantify the benefits of such a system we aimed to determine the number of accidents that could be
reduced, then by using the assumptions laid out in previous sections, calculate a monetary benefit for
reducing accidents.
To begin this process, the GES database was queried for specific accident data related to the application
in question. For instance, all accidents that took place in curves, and were related to speed were
included in this analysis.
In 2008, there were 241 fatalities, and ~ 109000 other injuries where this type of application may be
applicable. To determine the benefit of this system an efficacy rate must be determined to see how
much of a reduction from these figures can be expected.
Through another literature review, several reports were found discussing how effective forward collision
warnings could be. The three reports and results are summarized briefly below:
• Evaluation of an Automotive Rear‐ End Collision Avoidance System ( Volpe 2006) – A study that
analyzed data from a field operation test and the results suggest that 10% of all rear‐ end collisions
could be reduced.
• Integrated Vehicle Based Safety Systems: A Major ITS Initiative ( FHWA 2005) – A study on IV systems
that suggests these types of applications could reduce rear end, run off road, or lane change
collisions by 48%.
• The Evaluation of Impact on Traffic Safety of Anti‐ Collision Assist Applications ( Sala, Gianguido &
Lorenzo Mussone, 1999) – A simulation study that suggests between 10 and 60% accident reduction
could be attainable depending on the adoption rate of the technology. This is very interesting and
one of the only studies that addresses changes in effectiveness due to technology adoption.
Using these sources as references, we chose to use 25% accident reduction as a mid‐ level efficacy rate.
A low level would be 10% while a high efficacy level would be 50%.
With all of this data, a discounted yearly monetary benefit can be calculated. The equation would as
follows:
Where B is monetary benefits, n is the year, j is the application, and i is the injury level ( fatal, serious,
etc.).
Using this formula and the low efficacy rate, preliminary benefits of ~$ 28 Billion were estimated.
3. Merge/ Lane Change Warning
These warnings would alert vehicles on highway on‐ ramps if another vehicle is occupying its merging
space ( or in its blind spot). This is similar to Blind Merge Warning where warnings are used for vehicles
attempting to merge with limited sight distance, and another vehicle is predicted to occupy the merging
space. In addition, this system could warn the subject driver if a lane change is likely to cause a collision,
triggered by turn signal activation.
23
To quantify the benefits of such a system we aimed to determine the number of accidents that could be
reduced, then by using the assumptions laid out in previous sections, calculate a monetary benefit for
reducing accidents.
To begin this process, the GES database was queried for specific accident data related to the application
in question. For instance, all accidents that took place in curves, and were related to speed were
included in this analysis.
In 2008, there were 13 fatalities, and ~ 3500 other injuries where this type of application may be
applicable. To determine the benefit of this system an efficacy rate must be determined to see how
much of a reduction from these figures can be expected.
Through another literature review, several reports were found discussing how effective merge or lane
change warnings could be. The four reports and results are summarized briefly below:
• Integrated Vehicle Based Safety Systems: A Major ITS Initiative ( FHWA 2005) – A study on IV systems
that suggests these types of applications could reduce rear end, run off road, or lane change
collisions by 48%.
• Freightliner to Offer Collision Warning on New Truck Line ( Inside ITS 1995) – Empirical study of
Transport Besner Trucking Co, which reduced its at‐ fault accidents by 34%.
• Dutch Field Operational Test Experience with “ The Assisted Driver” ( Alkim, Boostma, and
Hoogendoorn 2007) – Empirical study of 20 vehicles in the Netherleands equipped with warning
systems which were driven for five months. It found that unintentional lane changes were reduced
by 35% on arterials, while it was reduced by 30% on highways.
• Run‐ Off Road Collision Avoidance Using IVHS Countermeasures: Final Report ( NHTSA, 1999) – A
simulation study that looked at lane departure warnings. Suggests passenger vehicle lane
departures would decrease by 10%, while heavy trucks would decrease by 30%.
Using these sources as references, we chose to use 35% accident reduction as a mid‐ level efficacy rate.
A low level would be 15% while a high efficacy level would be 60%.
With all of this data, a discounted yearly monetary benefit can be calculated. The equation would as
follows:
Where B is monetary benefits, n is the year, j is the application, and i is the injury level ( fatal, serious,
etc.).
Using this formula and the low efficacy rate, preliminary benefits of ~$ 2.1 Billion were estimated.
4. Intersection Collision Warning
Intersection Collision Warning applications provide warnings to drivers that a collision is likely at the
upcoming intersection either due to their own speed or inattention, or that of another driver.
24
To quantify the benefits of such a system we aimed to determine the number of accidents that could be
reduced, then by using the assumptions laid out in previous sections, calculate a monetary benefit for
reducing accidents.
To begin this process, the GES database was queried for specific accident data related to the application
in question. For instance, all accidents that took place in curves, and were related to speed were
included in this analysis.
In 2008, there were 88 fatalities, and ~ 37000 other injuries where this type of application may be
applicable. To determine the benefit of this system an efficacy rate must be determined to see how
much of a reduction from these figures can be expected.
Through another literature review, several reports were found discussing how effective intersection
collision warnings could be. The two reports and results are summarized briefly below:
• Field & Driving Simulator Validations of System for Warning Potential Victims of Red‐ Light Violators
( Inman, Vaughan TRB 2006) – A Field and Simulation study that tested participants in a driving
simulator and on a closed track. In the simulator, 90% stopped or avoided the collision, while on the
track, 64% stopped or avoided the collision.
• Intersection Collision Avoidance Study ( FHWA Office of Safety 2003) – An in depth analysis of
literature and operational concepts of specific ICAS systems, and they state that 100% reduction in
accidents is not unrealistic, however a more conservative estimate would be a 50% reduction in
accidents.
Using these sources as references, we chose to use 50% accident reduction as a mid‐ level efficacy rate.
A low level would be 25% while a high efficacy level would be 75%.
With all of this data, a discounted yearly monetary benefit can be calculated. The equation would as
follows:
Where B is monetary benefits, n is the year, j is the application, and i is the injury level ( fatal, serious,
etc.).
Using this formula and the low efficacy rate, preliminary benefits of ~$ 33 Billion were estimated.
5. Left Turn Assistant
Left Turn Assistants provide drivers information about oncoming traffic when trying to take a left‐ hand
turn at an unprotected intersection.
To quantify the benefits of such a system we aimed to determine the number of accidents that could be
reduced, then by using the assumptions laid out in previous sections, calculate a monetary benefit for
reducing accidents.
25
To begin this process, the GES database was queried for specific accident data related to the application
in question. For instance, all accidents that took place in curves, and were related to speed were
included in this analysis.
In 2008, there were 26 fatalities, and ~ 24000 other injuries where this type of application may be
applicable. To determine the benefit of this system an efficacy rate must be determined to see how
much of a reduction from these figures can be expected.
Since the application is very similar to that of intersection collision warnings, the literature used to
determine an efficacy rate for that application were leveraged for this application as well.
Using these sources as references, we chose to use 50% accident reduction as a mid‐ level efficacy rate.
A low level would be 25% while a high efficacy level would be 75%.
With all of this data, a discounted yearly monetary benefit can be calculated. The equation would as
follows:
Where B is monetary benefits, n is the year, j is the application, and i is the injury level ( fatal, serious,
etc.).
Using this formula and the low efficacy rate, preliminary benefits of ~$ 21 Billion were estimated.
6. Stop Sign Movement Assistant
Stop Sign Movement Assistants alert vehicles about to cross an intersection, after stopping, of cross
traffic.
To quantify the benefits of such a system we aimed to determine the number of accidents that could be
reduced, then by using the assumptions laid out in previous sections, calculate a monetary benefit for
reducing accidents.
To begin this process, the GES database was queried for specific accident data related to the application
in question. For instance, all accidents that took place in curves, and were related to speed were
included in this analysis.
In 2008, there were 110 fatalities, and ~ 10000 other injuries where this type of application may be
applicable. To determine the benefit of this system an efficacy rate must be determined to see how
much of a reduction from these figures can be expected.
Since the application is very similar to that of intersection collision warnings, the literature used to
determine an efficacy rate for that application were leveraged for this application as well.
Using these sources as references, we chose to use 50% accident reduction as a mid‐ level efficacy rate.
A low level would be 25% while a high efficacy level would be 75%.
26
With all of this data, a discounted yearly monetary benefit can be calculated. The equation would as
follows:
Where B is monetary benefits, n is the year, j is the application, and i is the injury level ( fatal, serious,
etc.).
Using this formula and the low efficacy rate, preliminary benefits of ~$ 10 Billion were estimated.
7. Highway/ Rail Collision Warning
Highway/ Rail Collision warnings provide alerts to reduce the likelihood of a collision between vehicles
and trains on intersecting paths.
To quantify the benefits of such a system we aimed to determine the number of accidents that could be
reduced, then by using the assumptions laid out in previous sections, calculate a monetary benefit for
reducing accidents.
To begin this process, the GES database was queried for specific accident data related to the application
in question. For instance, all accidents that took place in curves, and were related to speed were
included in this analysis.
In 2008, there were 0 fatalities, and ~ 0 other injuries where this type of application may be applicable.
To determine the benefit of this system an efficacy rate must be determined to see how much of a
reduction from these figures can be expected.
Through another literature review, a report was found discussing how effective Highway/ Rail Crossing
Warnings could be. The report and results are summarized briefly below:
• Second Train Coming Warning Sign Demonstration Projects ( TCRP Research Results Digest, 2002) – A
demonstration study of two sites, one in Baltimore and the other in LA, where warnings were placed
for approaching trains. 26% of drivers reduced the most risky behavior.
Using these sources as references, we chose to use 25% accident reduction as a mid‐ level efficacy rate.
A low level would be 10% while a high efficacy level would be 50%.
With all of this data, a discounted yearly monetary benefit can be calculated. The equation would as
follows:
Where B is monetary benefits, n is the year, j is the application, and i is the injury level ( fatal, serious,
etc.).
Using this formula and the low efficacy rate, preliminary benefits of ~$ 0 Billion were estimated.
27
D. Mobility Applications
As part of the mobility analysis, two applications are analyzed: Intelligent traffic flow control and free
flow tolling. Both of these applications are focused on reducing delay. Both those applications require
lane‐ level positioning accuracy to operate and therefore would benefit from a C‐ HALO nationwide
deployment.
1. Intelligent Traffic Flow Controls ( ITFC)
ITFC uses real‐ time data to adjust signal phases to an optimal level. These applications could also include
Green Light Optimal Speed Advisory, which would provide the subject vehicle with the optimal speed
given signal phase timing at upcoming intersections.
To quantify the benefits of such a system two additional pieces of information are needed to complete
the calculation. The first is to determine how much delay is currently realized at signalized intersections.
This was done through a literature review, and Temporary Losses of Highway Capacity and Impacts on
Performance ( Phase 2), written by Oak Ridge National Laboratory for the Department of Energy,
discusses sub‐ optimal signal timing specifically. Through surveying and significant quantitative modeling
they determine that there is, as of 1999, ~ 295 million hours of delay at signalized intersections.
Lastly, the efficacy of these new systems needs to be estimated. Through another literature review,
several reports were found discussing how much more optimal signal timing assisted in reducing delay.
The three reports and results are summarized briefly below:
• Preliminary Evaluation Study of Adaptive Traffic Control System ( LA DOT 2001) – Empirical study in
LA with 375 intersections, reduced delay by ~ 21%
• Realizing Benefits of Adaptive Signal Control at an Isolated Intersection ( Park and Change 2002) – A
simulation study on a hypothetical intersection of two one‐ way streets. Reductions in delay were
between 18‐ 20%
• ITS Benefits: The Case for Traffic Signal Control Systems ( Skabardonis 2001) – Empirical study of
multiple California implemented systems, reductions of delay close to 25%.
Using these sources as references, we chose to use 15% delay reduction as a conservative efficacy rate.
With all of this data, a discounted yearly monetary benefit can be calculated. The equation would as
follows:
Where B is monetary benefits, n is the year, and TVoM is the monetary value of time.
Using this formula, preliminary benefits of ~$ 10 Billion were estimated.
2. Free Flow Tolling
Toll collection without toll plazas reducing stop and go traffic surrounding current toll plazas, also
beneficial, but not included in this analysis is the fact that in tolling situations, costs are actually saved by
not having to build facilities. In this exercise we only look at reduced delay.
28
To calculate the delay reduced by free tolling systems, some metrics needed to be deciphered. Average
delay at a toll facility, the total revenue of all tolling facilities, and the average toll for toll roads in the
U. S is three metrics needed to calculate total delay due to toll facilities. Again, this was done through a
literature review, and Temporary Losses of Highway Capacity and Impacts on Performance ( Phase 2),
written by Oak Ridge National Laboratory for the Department of Energy, discusses average toll delay.
Through thorough quantitative analysis, they determine the average tolling delay to be 11.9 sec per
vehicle.
With this figure, only the number of vehicles would be necessary to determine overall delay. To
determine the number of vehicles using toll facilities, total tolling revenues and average toll were
sought. In the Highway Statistics 2007 published by the FHWA, the total revenues of toll facilities was
$ 7.7 billion, while in the Toll Facilities in the U. S. August 2009, the average toll is calculated to be $ 3.89
( 25). Using these two figures, an annual vehicle count of ~ 2 billion was determined. This was grown on a
year‐ to‐ year basis at a rate of 1.65% ( 26).
Lastly, the efficacy of these new systems needs to be estimated. Through another literature review,
several reports were found discussing how much free tolling systems reducing delay. The two reports
and results are summarized briefly below:
• Evaluation of Impacts from Deployment of an Open Road Tolling Concept for a Mainline Toll Plaza
( Klodzinski 2007) – Twenty‐ month empirical study done around UCF which reduced delays by
approximately 50 percent.
• Operational and Traffic Benefits of E‐ Zpass to the New Jersey Turnpike ( NJ Turnpike Authority 2001)
– EZ‐ pass empirical study that showed 85 percent reductions in delay.
Using these sources as references, we chose to use 70% delay reduction as a conservative efficacy rate.
With all of this data, a discounted yearly monetary benefit can be calculated. The equation would as
follows:
Where B is monetary benefits, n is the year, and TVoM is the monetary value of time.
Using this formula, preliminary benefits of ~$ 0.6 Billion were estimated.
E. Efficacy Literature Caveat
The ITS application benefit numbers are from the RITA ITS Benefits database online. Since ITS funding is
part of RITA’s budget, we have found and checked benefit numbers from some of these applications in
documents from m the GAO ( 27), RAND ( 28), and CBO ( 29). These do not challenge the assumptions
made and published by RITA with respect to the analyzed applications. The RITA database is the most
comprehensive.
29
F. Summary of Benefits
After completing all these individual analyses, the sum of these benefits ranges from $ 160 billion to
$ 320 billion. This range depends on whether one uses the low‐ level safety application efficacy rates or
the mid‐ level efficacy rates. This translates into 1.1 to 2.2 percent of GDP. The safety benefits in the
analysis dominate, making up over 90 percent of the total benefits calculated.
Figure 5 Benefits by Category
30
V. Separating HALO Costs into Good and Bad ( Dark) GPS areas
In order to accurately estimate the cost of delivering C‐ HALO services in areas of need, we estimate the
accuracy of existing location services on the ground and determine the area without good GPS coverage
today We refer to this area as the “ dark area” or the area, that may not realize HALO using N‐ RTK or HA‐NDGPS,
and to be targeted by a newer C‐ HALO infrastructure.
This chapter describes the methods and tools we have developed to estimate dark area. Emerging
technologies such as the penetration of INS systems in vehicles could mitigate this gap. For example, we
know from our prior work ( 30) that GPS augmented with INS can dead reckon to lane level precision for
about 20 seconds if there are no sudden lane changes or turns at intersections ( 31,32).
Several reports exist on the causes of errors when measuring position on the ground using the GPS
system ( 33). These reports address the theoretical values of the various types of errors. Table 4 below
shows the possible values for the different errors attributed to locating objects on the ground using GPS.
Table 4 Sources of GPS Errors
Source Effect ( m) Signal ISSσσETMoiaRRprgun tohCPnelopte(/ lasiloYmAplip ts) AAa heep trr rehhcrrirle siiiodvv rceciaai srecklltr f eoPCfeoefr( r/ rfctYrAseito ) oc s rt nss ±±±±±±±±± 001223566..... 03557
As part of this study, we set out to estimate the size of the “ gap” using empirical and data modeling
techniques to arrive at a more accurate assessment of GPS accuracy on the ground. Our method for
doing this relies on understanding the satellite coverage and the visibility of satellites at a Point‐ of‐
Interest ( POI) on the ground. When the POI is in an open space environment, the GPS receiver is capable
of communicating with several satellites ( 6 or more) and is able to locate the POI with good accuracy ( 1‐
3m). When comparing this POI with another POI in an urban setting with several high‐ rise buildings, the
number of satellites viewed drops significantly resulting in lower location accuracy.
Our effort rests on modeling the relation between position accuracy and number of satellites‐ in‐ view by
incorporating the PDOP values, the height of buildings near the POI, and the open‐ space area ‐ as
31
represented by street widths ‐ into the model. The method is tested on data from the city of San
Francisco. We collected validation data in the San Francisco Downtown blocks highlighted in Figure 6.
Figure 6 Shaded area represents study area
This area encompasses 2 sq. km of buildings of various heights providing us with a variety of satellite
counts. Figure 7 is a Google Earth 3D rendering showing the structures in this area as of 2009.
32
Figure 7 3D rendering showing building coverage in validation area
A. Data for Modeling Satellite Count
In order for us to systematically replicate the modeling of the gap across various cities, we rely on data
that is easily accessible in the public domain. The International Association of Assessing Officers ( IAAO)
in collaboration with the Urban and Regional Information Systems Association ( URISA) have been among
the leading efforts in enabling GIS use by cities all over the world. As a result almost all major cities in
the US have implemented GIS3 and are affiliated with either of those two organizations. In San
Francisco, the assessor office manages the SFParcel GIS system which holds information on close to
198,000 parcels . Of those we were able to obtain clean data on 160,000 parcels covering approximately
86% of the built area of San Francisco. In the remaining 14%, the height data could not be verified.
These are dropped from the model ( visualized as grey points in the plots below). The 86% that is used
covers only the parts of San Francisco that are registered with the assessor’s office. This does not
include open spaces, public gardens, etc. Those areas ( aka Park Acres) are estimated by the San
Francisco County’s office to be 0.19% of the total 121sq. km area of the City of San Francisco. So for the
purposes of this model, we will be assume them negligible, and the clean data we have on San Francisco
from the assessor’s office will be assumed to cover all the 121sq. km.
The model is constructed using the ESRI GIS software ArcMap. Data for the model includes:
• Building heights as reported by the SFParcel GIS system controlled by the County of San Francisco
• Street width as measured using the ArcMap GIS software
3 According to a 2003 survey by Public Technology Inc ( a national not‐ for‐ profit that works with local governments)
97% of cities with a population greater than 100,000 have a GIS system in place. ( Public Technology Inc., “ 2003
Survey on the Use of GIS Technology in Local Governments,” December 2003.)
33
Thus building heights and street width at a Point Of Interest ( POI) are “ known” variables in the model
and could be obtained from the GIS system of most city assessor’s office. The “ unknown” variable is the
satellite count. To calibrate the model we measure satellite count on the ground in the proposed area.
This is done by driving around with a GPS equipped Smartphone. We developed an application on the
Windows Mobile 6.5 operating system and deployed it on two HTC phones, namely, the HTC Diamond
and HTC Touch Pro 2. The application logs the following values:
• GPS Longitude and Latitude
• Number of Satellites Visible
• Number of Satellites Connected
• Vertical Dilution of Position ( VDOP)
• Horizontal Dilution of Position ( HDOP)
The preliminary data collected is visualized in Figure 8.
Figure 8 GPS data collection depicting satellite counts: > 6 are shown in green, between 4 and 6 are shown in yellow, < 4 are
shown in red
The number of satellites at a POI can be taken as an indicator of the GPS accuracy. However the
Horizontal Dilution of Precision ( HDOP) is a better indicator of the localization accuracy of the GPS. For
example, given a fixed number of satellites, the accuracy is better at a POI where the satellites are seen
34
well spread out as compared to a place where the satellites are more clustered together. The HDOP
captures this. Given the total number of operational satellites ( N = 30 ( 57)) and the predicted number of
satellites ( s) at a POI, the HDOP can be theoretically calculated as follows
Figure 9 shows the theoretical and the empirical HDOP values obtained from the data set along with the
95% confidence intervals for the empirical HDOP. The empirical HDOP values were obtained from the
data set collected in the city of San Francisco. The theoretical HDOP is obtained from the equation
above.
Figure 9 Empirical and predicted HDOP values with 95% confidence intervals.
The question of what values of HDOP is good for high‐ accuracy localization would depend on the
receiver, ionospheric conditions etc. Typically, under normal conditions, HDOP values below 4 are
considered to be good ( 52). However, for high accuracy applications, we would require the HDOP values
35
to be lesser than 2 or 1. Using this as a rule of thumb, based on Figure 9, we roughly categorized the
satellite counts as < 4, 4 to 6 and > 6 and the model predictions were carried out for these three
categories. The use of 6 as a threshold for good GPS coverage is empirically supported by a 100 miles of
driving data ( 30). In the next section we will describe the models used to predict the satellite counts and
evaluate the performance based on the collected data.
B. Hidden Markov Model to Predict Satellite Counts
This section describes the method used to predict the number of satellites at a POI given the GIS data
i. e. building heights and street widths. The quantity being predicted is illustrated by Figure 10.
The estimate of the satellite count at the point of interest is obtained as follows. We think of the
satellites as being placed on the surface of a hemisphere with a radius R centered at the POI. We assume
the POI is occluded from the satellites only by buildings at the sides of the street and there is visibility in
the forward and backward directions as in Figure 10. The mask angle alpha ( shown in Figure 10) is
calculated based on the heights of the buildings and street width as follows.
Figure 10 Mask‐ Angle Representation using street width and building heights
The satellites visible at this point, are essentially the ones lying on a strip of the hemisphere with angular
width alpha. The fraction of these satellites is given by
36
where N is the total number of satellites in orbit. The satellite count data collected in downtown San
Francisco is compared against the predicted count computed as described above. The data set
consisted of 1657 data points. These were a subset of the 13822 data points collected overall. The
subset was chosen by excluding data points that were not part of downtown San Francisco as in Figure 8
and data points that did not have corresponding meaningful building heights in the SFParcel GIS system.
Out of these 1657 data points, 34 data points had satellite counts < 4, 568 points had satellite counts 4
to 6 and the rest had satellite counts > 6.
Table 5 shows the prediction accuracy of the model. Each column in this table is the prediction accuracy
for the corresponding category of satellites. For example column 1 shows the percentage of data points
having < 4 satellites being predicted as < 4, 4 to 6 and > 6 number of satellites.
Table 5 Model Prediction Accuracy
Satellites True < 4 True 4 to 6 True > 6
Predicted < 4 0.74 0.16 0.02 Predicted 4 to 6 0.24 0.25 0.07
Predicted > 6 0.02 0.59 0.91 The overall prediction accuracy is around 69%. The prediction accuracy is calculated by adding the
fraction of data points in each of the categories multiplied with the corresponding diagonal entry. We
next add a Hidden Markov Model ( HMM) ( 51) to improve the prediction accuracies. The HMM captures
the statistical dependence of the satellite count at a POI on the building heights at neighboring points as
well.
The idea of the HMM modeling is as follows. Depending on the time of day, climatic conditions, or
scatter in the environment, the number of satellites visible at a point could vary significantly. These
random parameters lead to a stochastic dependence between the building heights and the number of
satellites at a given point. Furthermore, the number of satellites at a particular point would be
dependent on the number of satellites in nearby points. These dependencies could be captured by a
HMM as shown in the figure below.
37
Figure 11 HMM for predicting satellite count
The nodes corresponding to the heights are values that are known. The nodes corresponding to the
satellite count are the hidden nodes that need to be estimated. The hidden nodes are connected to their
neighbors to model the dependency between satellite counts in adjacent regions. The transition
probabilities between the satellite count variables are modeled using a sticky Markov chain ( 51). This is
validated using the empirical data collected. The distribution of the building height given the satellite
count is modeled as a Gaussian random variable with mean and variance empirically determined from
the collected data. The mean height and variance for the Gaussian model and the transition
probabilities between the states were obtained empirically from the collected data. 95% confidence
intervals were calculated for the estimated parameters of the model. The satellite counts were
predicted by taking the mean of the estimated parameters and the corresponding prediction accuracies
for this model are as shown in Table 6.
Table 6 HMM Satellite Count Accuracies
Satellites True < 4 True 4 to 6 True > 6
Predicted < 4 0.70 0.01 0
Predicted 4 to 6 0.08 0.41 0.11 Predicted > 6 0.22 0.58 0.89
The results of this model are 87% accurate. Experimental results [ 53] have shown that with less than 7
satellites, GPS estimates have errors of the order of 1 meter. Therefore we focus on a 7 satellite
threshold. Our model predicts that 0.3 to 4 % of the streets of San Francisco has satellite coverage of
less than 7 satellites with 95% confidence. The prediction is made for all the streets of San Francisco
where we have the building height data from the assessor’s office. Figure 12 shows the results of the
model. The figure is drawn by aggregating and averaging the values of the model into 10m x 10m grids
that lie on roads. Each grid is given a color based on the average satellite count in that grid: red if < 4,
yellow if between 4 and 6 and green if > 6. The 14% of San Francisco for which we do not have height
data is not modeled. It appears as grey areas in the figure. We would also like to note that the effects of
multipath are not taken into account in this modeling. Thus the green areas do not
38
Figure 12 GPS data estimating satellite counts: > 6 or more are shown in green, between 4 and 6 are shown in yellow, < 4 are
shown in red
necessarily reflect regions of high accuracy. Even though there is a satellite visibility of > 6, multipath can
cause significant errors. However, we can say with high confidence that the dark areas are regions of
bad location accuracies. Table 6 can be related to some numbers in the literature. The methods in the
literature use precise LiDAR data to yield good prediction accuracies [ 58]. The authors of [ 58] evaluated
their method by two test cases. With a 5m RADAR digital surface map, the mean error in the predicted
number of satellites using their model is 2.7 and 4.85 satellites with the corresponding error
percentages being 46% and 82% in the two test cases. This improves with a more precise digital surface
map. However it would be very expensive to obtain precise LiDAR data for an entire city and all the cities
in the US. Under the same error metric, the mean error in the predicted number of satellites using our
model is 1.85 satellites and the error percentage is 29.14%, which seems good. We use only building
height data that can be obtained from the city planning department or a similar agency. However the
comparison of the models is to be taken with caution given that the test data under consideration is
vastly different for the two approaches. Our accuracies could be improved by having more data points.
Extensive data collection during different times of the day and in different regions can help build and
evaluate better models.
C. Use of the Model
The approach we adopted to obtain the satellite count on the surface in San Francisco can be repeated
to any city in the US. The method is based on a GIS model constructed from the building heights and
street width of the whole city – information easily accessible for all urban centers in the US via the local
39
assessor’s office. This would quantify the area where decimeter level accuracy cannot be achieved today
with the current GPS and DGPS technology in many cities nationwide.
Our method produces a figure such as figure 12. Such a figure can guide the phased deployment of C‐HALO
infrastructure and provide insight into the full extent of new infrastructure required. The red areas
are candidates for a first phase C‐ HALO deployment. Fortunately, they are also few in number,
suggesting one might reap considerable improvements in location accuracies for moderate initial
investment.
The benefits of deploying in red or yellow areas can be better understood by using GIS tools to overlay
other statistics such as the distribution of accidents on figure 12. We have done this for San Francisco.
Figure 13 shows a quadrant of San Francisco which include 1000m x 1000m grids of 2008 accident data
as reported by NHTSA where red are areas of high accident counts, yellow those of medium counts and
green of low or no accidents. This type of plot could be repeated by overlaying emissions data, or
congestion data or other types of data relevant to the benefit measures guiding C‐ HALO pilot
deployments or initial infrastructure investments.
Figure 13 Overlay of accident data on satellite count projection data
40
VI. Cost Assessment in Good GPS Areas
This chapter quantifies the new infrastructure investment required to realize a C‐ HALO service in areas
with good GPS coverage based on N‐ RTK technology. This cost does not include the wireless
communication technology between vehicles, but just the cost of deploying the infrastructure to
provide the service. For the purposes of this analysis we explored the cost of implementing N‐ RTK
infrastructure. This of course, is an upper bound on the cost estimate of the infrastructure since in
reality some areas of the U. S. are already covered by N‐ RTK service, while others areas may not need it
( i. e. some areas may already have C‐ HALO capability without N‐ RTK).
The present N‐ RTK system consists of a set of references stations and servers installed and maintained
by companies/ governmental agencies offering the service. Customers use the service by paying a
subscription fee. The NRTK servers provide the rovers with the RTCM corrections as and when requested
by the rover. A typical N‐ RTK system as implemented by companies like Trimble and adopted by the
present DoT’s, consists of the following components4:
1. N‐ RTK base stations with geodetic and communication capabilities
2. Server( s) that handle incoming NRTK requests and RTK corrections, process the data and
transmits the correction data to the rovers.
3. Communication links between reference stations and server( s) and the rovers and the server( s).
The capital costs involved in setting up such a system would include
a) Hardware ‐ NRTK reference stations (*) and the servers.
b) Software on the servers and reference stations. This should also have the ability to handle
secure communication.
c) Design ( hardware, site selection etc), testing and installation of the reference stations (*).
d) Predicted hardware and software upgrades (*).
Variable costs include
a) Hardware and software maintenance costs for the server and reference stations (*)
b) Rent/ value of facility for the reference stations (*) and servers
c) Link costs for the communication from reference stations to server (*) and from server to
rovers
d) Power supply to reference stations (*) and servers.
e) Customer support
The cost estimates in Table 7 are for the installation and maintenance of a single base station and
include the costs marked (*) in the NRTK system components.
4 The OSI National NRTK solution, http:// www. fig. net/ pub/ athens/ papers/ ts11/ TS11_ 5_ Bray_ Greenway. pdf
41
A. N­RTK
Base Station Cost Estimation
To begin estimating the infrastructure cost of deploying N‐ RTK infrastructure, discussions, via email and
phone, were held with employees of three current N‐ RTK deployments, 2 state DOT’s ( Iowa [ Steve
Milligan] and Ohio [ Dave Beiter) and one Cooperative ( Washington [ Gavin Schrock]). During these emails
and conversations the costs associated with infrastructure cost requirements, as well as maintenance
and operating costs were focused on ( 47,48,49,60). We also obtained concrete documents on invoices
and cost reports for the hardware, servers, services etc ( 69,70) from these DOT’s. These costs are
summed and determined over the 22‐ year horizon using a 5% discount rate. The calculations are as
follows.
Table 7 N‐ RTK Cost Estimation ( No Other ( Contingency) Costs) ( 47,48,49.60)
This is the representative cost given average levels of all the above costs. There are low and high
estimates for each cost category, including the useful life of the hardware, which ranges from 7 to 15
years. This useful life changes the 22‐ year horizon cost of the hardware.
The range of infrastructure costs is from $ 220K to $ 615K per base station for the life of the system.
Using a range of annual contingency expenses from $ 25K to $ 70K the range of infrastructure costs
increases to $ 570K to $ 1.6M per base station for the life of the system. If one were to provide N‐ RTK
coverage over the entire US land mass for the horizon of this project, approximately 2,730 base stations
would be needed. Using this figure, nationwide N‐ RTK coverage would cost between $ 1.6B to $ 4.4B.
This may be compared to benefits ranging between $ 160 and $ 300 billion from the Intelligent
Transportation Systems Sector alone.
42
VII. Further Research
While accomplishing many things during this process, there are still many areas of research that could
improve the analysis and enhance scope to incorporate more levels of detail. Areas of further
explorations are briefly discussed below:
• Further model refinement:
Continue to update the model with more empirical data and expand use to other cities and use
nationwide to determine nationwide ‘ dark area.’ Different models than the HMM could be
investigated for better accuracy. Obtaining real data that spans across weather conditions and time
would enable the model to provide time‐ based projection of satellite coverage, yielding a statement
such as satellite count less than 6 for x % of the time. The use of variations on the satellite count,
such as number of satellites used to calculate position could offer a better understanding of multi‐path
errors in the region.
• Cost of infrastructure adjustment:
Utilizing the new ‘ dark area’ estimation a more accurate infrastructure cost estimate can be made
based on average coverage area of a base station.
• Communication Technology Research:
Further research needs to be done on how the actual augmentation services will be communicated
to the vehicles and between vehicles. This analysis has not been included in this report, but is
integral in realizing the benefits of the new ITS applications.
• Benefit Refinement:
Ultimately, the benefits calculations could be expanded to include environmental benefits.
• Technology Assessment:
To achieve a more thorough understanding of where N‐ RTK stands in terms of cost effectiveness a
more complete technology assessment needs to be completed. As part of this, the cost of
infrastructure for each technological alternative needs to be completed, as well as analyzing the
capabilities of each technology. Once this analysis is complete, the technologies can be compared
and a prudent decision going forward could be made.
• Other Economic Stimulus:
Analysis could be completed on what type of economic development may be induced due to these
applications, specifically the mobility applications since the main component of the benefits is saved
time. Typically if users are saving time, they are using that time to create benefits in another
industry or realm. These effects need to be explored more fully to get a better estimate of the full
benefits of implementing C‐ HALO services.
43
VIII. References
1. Department of Defense. Global Positioning System 2008: A Report to Congress.
http:// pnt. gov/ public/ docs/ 2008/ biennial2008. pdf. 2008
2. Brain, Marshall and Harris, Tom. How GPS Works. How Stuff Works.
http:// electronics. howstuffworks. com/ gadgets/ travel/ gps4. htm. Accessed August 23, 2010.
3. Wanninger, Lambert. Introduction to N‐ RTK. International Association of Geodesy. June 16
2008. http:// www. wasoft. de/ e/ iagwg451/ intro/ introduction. html. Accessed August 23,
2010.
4. PRNewswire. Sensis Multilateration Systems Meeting the Demands of Continually Changing
Airports. Airport Business. July 13, 2010http:// www. airportbusiness. com/ web/ online/ Top‐
News‐ Headlines/ Sensis‐ Multilateration‐ Systems‐ Meeting‐ the‐ Demands‐ of‐ Continually‐
Changing‐ Airports/ 1$ 37952. Accessed August 23, 2010
5. Tuttlebee, Walter. Advances in Software Defined Radio. Annals of Telecommunications. Vol
57, No 5‐ 6, p 314‐ 337. http:// www. springerlink. com/ content/ 5784533018364762/.
Accessed August 23, 2010.
6. Sun, Chen; Cheng, Jun; and Ohira, Takashi. Handbook for Advancements in Smart Antenna
Technologies for Wireless Networks. IGI Global. July 2008.
http:// my. safaribooksonline. com/ 9781599049885. Accessed August 23, 2010
7. Hiraiwa, et al. Mobile Communication Platforms for ITS Services. Hitachi Review Vol 52, No
1. 2003
8. California Department of Transportation. An Evaluation of Dynamic Curve Warning Systems
in the Sacramento River Canyon: Final Report. 2000
9. Skabardonis, Alex. ITS Benefits: The Case for Traffic Signal Control Systems. 2001.
10. Gakstatter, Eric and Lorimer, Robert. GNSS Market Research and Analysis. Position One
Consulting Pty Ltd. September 2008.
11. Economic Benefits of High Resolution Positioning Services. The Allen Consulting Group Pty
Ltd. November 2008.
12. Leveson, Irving. Socio‐ Economic Benefits Study: Scoping the Value of CORS and GRAV‐ D.
Leveson Consulting. January 2009.
13. Levson, Irving. Benefits of the New GPS Civil Signal. Leveson Consulting. July/ August 2006.
14. Alcantarilla, I., Porras, D., Tajdine, A., Zarraoa, N., Lévy, J. C., " The Benefits of Multi‐constellation
GNSS Augmentations," Proceedings of the 19th International Technical
Meeting of the Satellite Division of The Institute of Navigation ( ION GNSS 2006), Fort Worth,
TX, September 2006, pp. 930‐ 938.
15. Swann, J., Chatre, E., Ludwig, D., " Galileo: Benefits for Location‐ Based Services,"
Proceedings of the 16th International Technical Meeting of the Satellite Division of The
Institute of Navigation ( ION GPS/ GNSS 2003), Portland, OR, September 2003, pp. 1603‐
1612.
16. Vollath, Ulrich, Patra, Richard, Chen, Xiaoming, Landau, Herbert, Allison, Timo,
" GALILEO/ Modernized GPS: A New Challenge to Network RTK," Proceedings of the 17th
International Technical Meeting of the Satellite Division of The Institute of Navigation ( ION
GNSS 2004), Long Beach, CA, September 2004, pp. 2855‐ 2863
44
17. Arthur, Daniel, Jenkins, Bryan, von Tunzelmann, G. Nick, Styles, Jon, " The Macroeconomic
Impacts of Galileo," Proceedings of the 18th International Technical Meeting of the Satellite
Division of The Institute of Navigation ( ION GNSS 2005), Long Beach, CA, September 2005,
pp. 381‐ 389.
18. Brennan, Shawn M., Van Dyke, Karen, " National PNT Architecture ‐ Interim Results,"
Proceedings of the 63rd Annual Meeting of The Institute of Navigation, Cambridge, MA,
April 2007, pp. 614‐ 623.
19. The CAMP Vehicle Safety Communications Consortium. Identify Intelligent Vehicle Safety
Applications Enabled by DSRC. March 2005.
20. Nussle, Jim. Memorandum for the Heads of Departments and Services. Office of
Management and Budget. December 12, 2008.
21. Economic and Industry Analysis Division, John A. Volpe National Transportation Systems
Center, United States Department of Transportation. Vehicle‐ Infrastructure Integration
Initiative: Benefit‐ Cost Analysis Version 2.3. May 8, 2008.
22. Cambridge Systematics. Traffic Congestions and Reliability: Trends and Advanced Strategies
for Congestion Mitigation. September 2005.
23. The Fatality Analysis Reporting System. http:// www‐fars.
nhtsa. dot. gov/ QueryTool/ QuerySection/ SelectYear. aspx. Accessed August 23, 2010.
24. Revised Departmental Guidance: Treatment of the Value of Preventing Fatalities and
Injuries in Economic Analyses. Office of Management and Budget. 2008.
25. Toll Facilities in the United States. Office of Highway Policy Information. August 2009.
http:// www. fhwa. dot. gov/ ohim/ tollpage. htm. Accessed August 23, 2010
26. Traffic Volume Trends. Federal Highway Administration.
http:// www. fhwa. dot. gov/ ohim/ tvtw/ tvtpage. cfm. March 2010 & March 2009. Accessed
August 23, 2010.
27. Foresight Issues Challenge DOT’s Efforts to Assess and Respond to New Technology‐ Based
Trends. US GAO. October 2008. http:// www. gao. gov/ new. items/ d0956. pdf. Accessed
August 23, 2010.
28. Liability and Regulation of Autonomous Vehicle Technologies. Anderson, James; Kalra,
Nidhi; Wachs, Martin. RAND. 2009.
http:// www. rand. org/ pubs/ external_ publications/ EP20090427/. Accessed August 23, 2010.
29. High‐ Tech Highways: ITS and Policy. Congressional Budget Office. October 1995.
http:// www. cbo. gov/ doc. cfm? index= 16& type= 0. Accessed August 23, 2010
30. S. Rezaei and R. Sengupta " Kalman Filter‐ Based on Integration of DGPS and Vehicle Sensors
for Localization," IEEE Transactions on Control Systems Technology, November 2007, Vol.
15, No. 6, pp. 1080‐ 1088.
31. Chen, X., U. Vollath, H. Landau, K. Sauer. Will Galileo/ Modernized GPS Obsolete Network
RTK?. European GNSS 2004 Conference, May 16‐ 19, 2004, Rotterdam, The Netherlands .
32. LeLievre, Jaqueline. GPSnet NRTK Corrections and Dynamic Accuracy Trials. GNSS Technical
Support Newsletter. Issue 29, June 2008. pp. 6
33. Moahn, Brij and Jain, Pranay. A Seminar Report on GPS. The University of Rajasthan, Jaipur.
http:// www. scribd. com/ doc/ 6698521/ GPS‐ Final‐ Report. Accessed August 23, 2010
45
34. De Smith, Goodchild, and Longley. Geospatial Analysis, a Comprehensive Guide. 3rd Edition.
http:// www. spatialanalysisonline. com/ output/ html/ InversedistanceweightingIDW. html.
Accessed August 23, 2010.
35. General Estimates System Database FTP Site. ftp:// ftp. nhtsa. dot. gov/ GES/. Accessed
December 17, 2010.
36. CRTN: California Real Time Network. http:// sopac. ucsd. edu/ projects/ realtime/. Accessed
December 17, 2010.
37. Iowa Real Time Network. Iowa Department of Transportation.
http:// www. iowadot. gov/ rtn/. Accessed December 17, 2010.
38. ODOT’s VRS RTK Network. Ohio Department of Transportation.
http:// www. iowadot. gov/ rtn/. Accessed December 17, 2010.
39. Washington State Reference Network. A Regional Cooperative of Real Time GPS Networks.
http:// www. iowadot. gov/ rtn/. Accessed December 17, 2010.
40. The Utah Reference Network Global Positioning System. Utah Department of
Transportation. Accessed December 17, 2010.
41. Oregon Real Time Network. Oregon Department of Transportation.
http:// www. theorgn. org/. Accessed December 17, 2010.
42. Mn/ DOT CORS GPS Network. Minnesota Department of Transporation.
http:// www. dot. state. mn. us/ surveying/ CORS/ CORS. html. Accessed December 17, 2010.
43. MDOT CORS. Michigan Department of Transporation. http:// www. mdotcors. org/. Accessed
December 17, 2010.
44. Wisconsin Continually Operating Reference Station. Wisconsin Department of
Transportation. https:// wiscors. dot. wi. gov/. Accessed December 17, 2010.
45. The RTK Cooperative Network. Western Data Systems. http:// www. txrtk. com/. Accessed
December 17, 2010.
46. Midwest RTK Networks. Seiler Instrument Company. http:// www. mwrtk. net/. Accessed
December 17, 2010.
47. Email from Steve Milligan, to Adam Goodliss. Re: RTK/ CORS Network Cost Information ‐ UC
Berkeley Research. Received October 24, 2010.
48. Email from Dave Beiter, to Adam Goodliss. Re: RTK/ CORS Network Cost Information ‐ UC
Berkeley Research. Received October 21, 2010.
49. Phone Interview with Gavin Schrock. Re: RTK/ CORS Network Cost Information ‐ UC Berkeley
Research. Conducted October 25, 2010.
50. Phone Interview with John Pointon, Director of Sales, Omnistar. Re: High Accuracy Service
Capabilities. Conducted December 9, 2010.
51. http:// en. wikipedia. org/ wiki/ Hidden_ Markov_ model
52. http:// www. developerfusion. com/ article/ 4652/ writing‐ your‐ own‐ gps‐ applications‐ part‐
2/ 2/
53. Sengupta R, Rezaei S, Shladover S. E, Cody D, Dickey S, Krishnan H, Cooperative collision
warning systems: Concept definition and experimental implementation, Journal of
Intelligent Transportation Systems.
54. http:// smartnet. leica‐ geosystems. us/ 9SmartNetSubAgree. html
55. http:// smartnet. leica‐ geosystems. us/ 2eFAQ. html# whatisthertkaccuracy
56. http:// www. fhwa. dot. gov/ publications/ research/ operations/ 02110/ index. cfm
46
57. http:// en. wikipedia. org/ wiki/ Global_ Positioning_ System
58. Surface Modelling for GPS Satellite Visibility: George Taylor, Jing Li, David Kidner and Mark
Ware http:// www. springerlink. com/ content/ p85w747422684377/
59. http:// www. fhwa. dot. gov/ publications/ research/ operations/ its/ 05034/ 05034. pdf
60. Iowa Department of Transportation Purchasing Contract. Iowa Statewide RTK‐ GPS Network
Project. January 2008.
61. J Barnes, C Rizos, J Wang, D Small, G Voigt and N Gambale, Locata: A new positioning
technology for high precision indoor and outdoor positioning, ION GPS/ GNSS Sept 2003.
62. J Barnes, C Rizos, J Wang, D Small, G Voigt and N Gambale, Locata: The positioning
technology of the future, July 2003.
63. http:// www. locatacorp. com/ index2. html
64. http:// www. standards. its. dot. gov/ Documents/ advisories/ dsrc_ advisory. htm
65. http:// www. 3gpp2. org/
66. David S. Chiu, Glenn MacGougan, Kyle O’Keefe: UWB Assisted GPS RTK in Hostile
Environments, ION NTM 2008.
67. Paul Richardson, Weidong Xiang, Dan Shan : An outdoor UWB tracking system to improve
safety of semi‐ autonomous vehicle operations, International Journal of Ultra Wideband
Communications and Systems, 2010
68. M. Zabic and O. A. Nielsen, An analysis of stand‐ alone GPs quality and simulated GNSS
quality for Road Pricing.
69. FOIA responses from Ohio State Department of Transportation – invoices for the server and
other hardware purchases and reports.
70. FOIA responses from Seattle Public Utilities for the Washington State Reference Network –
cost estimation and annual budget reports.
71. Volker Janssen, New Ways to Network RTK: How do VRS and MAC measure up?
http:// eprints. utas. edu. au/ 9327/
47
Appendix A – Market Presence ( X­Major;
O­Minor)
Aero. Agri. AV Constr. Def. Mari. Mine Survey
Automated Positioning Systems X
Atair Aerospace X
Axio‐ Net GmbH O O O O O
Biscarosse BV X
C& C Technologies X
Credent Technologies X
Crossbow Technology O O O O O O O
DataGrid Inc. O O O O O X
Fugro/ Omnistar X O O X X X X
GeoKosmos X
GPS Ag X
Grumman X X
Hemisphere GPS X O O O
Honeywell X
Javad GNSS X X
John Deere/ Navcom Technology X O
Leica Geosystems/
Novatel/ Hexagon
O X X O X
Locata Corporation O X X
Magellan X
MMIST Inc. X
NavSys Corportation X O X O O
Novariant O X O O X
New Zealand Aerial Mapping X
Septentrio BV X X
Stara Technologies, Inc. X
Subsea 7/ Veripos X X X
Suzhou FOIF X X
TopCon/ Sokkia O X X O X X
Trimble O X O X O X X X
48
Appendix B – Process of Selecting ITS Application that Stand to Benefit
from C­HALO
A. Initial Application Selection Matrix ( 1,2)
ITS
Applications Safety
Location
Services
Necessary
Accuracy
Requirements
Transmission
Min
Frequency
Allowable
Latency
Required
Range
Units
( Y/ N) ( Y/ N) ( m) ( Event/ Periodic) ( Hz) ( msec) ( m)
Traffic Signal
Violation
Warning*
Y Y 1 P 10 100 250
Stop Sign
Violation
Warning
Y Y 1 P 10 100 250
Curve Speed
Warning* Y Y 1 P 1 1000 200
Emergency
Electronic Brake
Lights*
Y Y 1 E 10 100 300
Adv. Warning
Info/ Weather &
Road Conditions
Approaching
Emergency
Vehicle Warning Y Y 5 E 1 1000 1000
Emergency
Vehicle Signal
Preemption Y Y 5 E N/ A 1000 1000
SOS Services
Y Y 25 E 1 1000 400
Post- Crash
Warning Y Y 1 E 1 500 300
In- Vehicle
Signage Y Y 5 P 1 1000 200
Work Zone
Warning Y N N/ A P 1 1000 300
In- Vehicle Amber
Alert Y N N/ A E 1 1000 250
Safety Recall
Notice Y N N/ A E N/ A 5000 400
JIT Repair
Notification N N N/ A E N/ A N/ A 400
Low Parking
Structure
Warning Y Y 5 P 1 1000 100
49
ITS
Applications Safety
Location
Services
Necessary
Accuracy
Requirements Transmission
Min
Frequency
Allowable
Latency
Required
Range
Units
( Y/ N) ( Y/ N) ( m) ( Event/ Periodic) ( Hz) ( msec) ( m)
Wrong Way
Driver Warning Y Y 1 P 10 100 500
Low Bridge
Warning Y Y 5 P 1 1000 300
V2V Road
Feature
Notification
Y Y 5 E 2 500 400
Cooperative
Glare Reducation N Y 1 P 1 1000 400
Instant
Messanging N N N/ A E N/ A 1000 50
Vehicle- Based
Road Condition
Warning Y Y 25 E 2 500 400
Visibility
Enhancer Y Y 1 P 2 100 300
Road Condition
Warning Y N N/ A E 1 1000 200
Highway Merge
Applications
Highway Merge
Assistant Y N N/ A P 10 100 250
Intelligent On-
Ramp Metering N N N/ A E 1 1000 100
Blind Merge
Warning Y N N/ A P 10 100 200
Left Turn
Assistant* Y Y 1 P 10 100 300
Stop Sign
Movement
Assistance* Y Y 1 P 10 100 300
Pedestrian
Crossing
Information
Y Y 1 P 10 100 200
Collision Warning
Applications
Pre- Crash
Sensing* Y Y 1 E 50 20 50
Cooperative
Forward Collision
Warning* Y Y 1 P 10 100 150
50
ITS
Applications Safety
Location
Services
Necessary
Accuracy
Requirements Transmission
Min
Frequency
Allowable
Latency
Required
Range
Units
( Y/ N) ( Y/ N) ( m) ( Event/ Periodic) ( Hz) ( msec) ( m)
Cooperative
Collision Warning Y Y 1 P 10 100 150
Highway/ Rail
Collision Warning Y Y 1 E/ P 1 1000 300
Intersection
Collision Warning Y Y 1 P 10 100 300
Adapative
Headlight Aiming Y Y 1 P 1 1000 200
Adaptive
Drivetrain
Management N Y 5 P 1 1000 200
Lane Change
Warning* Y Y 1 P 10 100 150
Blind Spot
Warning Y Y 1 P 10 100 150
Corridor
Management
Cooperative
Vehicle- Highway
Automation
System
Y Y 5 P 50 20 100
Cooperative
Adaptive Cruise
Control Y Y 5 P 10 100 250
Intelligent Traffic
Flow Control N Y 5 E 1 1000 250
Free- Flow Tolling
N Y 1 E N/ A 50 50
Private
Applications
Enhanced Route
Guidance &
Navigation N Y 1 E N/ A 1000 200
Point of Interest
Notification N Y 5 P 1 1000 400
Map Downloads
& Updates N N N/ A E/ P 1 1000 400
GPS Correction
N N N/ A P 1 1000 400
51
B. Applications that stand to benefit from C­HALO
1. Safety
a) Curve Speed Warning
Aid drivers in negotiating curves at appropriate speeds.
b) Emergency Electronic Brake Light
Warns a driver when forward vehicle brakes hard ( deceleration is above a predetermined threshold).
This is very similar to Cooperative Forward Collision Warning which is used to preemptively avoid rear‐end
collisions with vehicles in front of the subject vehicle.
c) Highway Merge Assistant
Warns vehicles on highway on‐ ramps if another vehicle is occupying its merging space ( or in its blind
spot). This is similar to Blind Merge Warning where warnings are used for vehicles attempting to merge
with limited sight distance, and another vehicles is predicted to occupy the merging space.
d) Blind Spot Warning
Warns subject driver if another vehicle is occupying his/ her blind spot during an intended lane change
maneuver.
e) Lane Change Warning
Warns subject driver if a lane change is likely to cause a collision. Triggered by turn signal activation.
f) Intersection Collision Warning
Provides warnings to drivers that a collision is likely at the upcoming intersection
g) Cooperative Collision Warning
Warns vehicles when a collision is likely with surrounding vehicles.
h) Left Turn Assistant
Provides drivers information about oncoming traffic when trying to take a left‐ hand turn at an
unprotected intersection.
i) Stop Sign Movement Assistance
Warns vehicles about to cross an intersection, after stopping, of cross traffic.
j) Highway/ Rail Collision Warning
Provides warnings to reduce the likelihood of a collision between vehicles and trains on intersecting
paths.
k) Pedestrian Crossing Information
Alerts vehicles if there is danger of a collision with a pedestrian in a crosswalk.
52
2. Mobility
a) Free Flow Tolling
Toll collection without toll plazas reducing stop and go traffic surrounding current toll plazas.
3. Emissions
a) Adaptive Drivetrain Management
Allows vehicles to anticipate shift change patterns, and assist engine management systems to stabilize
the transmission. Effects should be seen in increased gas mileage, reduced emissions, and improved
shifting performance.
4. Mobility & Emissions
a) Intelligent On­ramp
Metering
Uses real‐ time data to adjust ramp metering signal phases
b) Intelligent Traffic Flow Control
Use real‐ time data to adjust signal phases to an optimal level. Could also include Green Light Optimal
Speed Advisory, which would provide the subject vehicle with the optimal speed given signal phase
timing at upcoming intersections.
c) Private Applications ( Etc.)
Enhanced Route Guidance & Navigation
Drive‐ thru Payments
Parking Lot Payment/ Spot Locator
C. References
1. Lockheed Martin Federal Systems. ITS Performance and Benefits Study. June 1996.
2. The CAMP Vehicle Safety Communications Consortium. Identify Intelligent Vehicle Safety
Applications Enabled by DSRC. March 2005.
53
Appendix C – Querying Methodology Matrix
54
Appendix D – Efficacy Rate Matrix
Low Mid High
Curve Speed Warning 20.00% 40.00% 70.00%
Emergency Electronic Brake Lights 10.00% 25.00% 50.00%
Highway/ Rail Collision Warning 10.00% 25.00% 50.00%
Intersection Collision Warning 25.00% 50.00% 75.00%
Left Turn Assistant 25.00% 50.00% 75.00%
Merge/ Lane Change Applications 15.00% 35.00% 60.00%
Stop Sign Movement Assistance 25.00% 50.00% 75.00%
Free Flow Tolling 70.00%
Intelligent Traffic Control 15.00%
55
Appendix E – Cost Estimation Emails
Email from David Beiter to Adam Goodliss:
Adam, my answers to your questions below are in Red
David J. Beiter, PE / SIT, Transportation Engineer 2
ODOT Office of Aerial Engineering and Surveying
1602 West Broad St, Columbus, Ohio 43223
Voice: ( 614)- 275- 1372 FAX: ( 614)- 275- 1673
e- mail: dave. beiter@ dot. state. oh. us
Adam Goodliss < agoodliss@ gmail. com>
10/ 20/ 2010 11: 59 AM
To < Dave. Beiter@ dot. state. oh. us>
cc
Subject Re: RTK/ CORS Network Cost Information UC Berkeley Research
Dave,
Thanks so much for getting back to me. I just want to clarify what you wrote to make sure I understand you
correctly, and then ask a couple questions to follow up.
From your summary, I understand the base towers cost ~$ 2K fully loaded ( installation, crew, fixed cost, etc.) and
the mount cost fully loaded, if put on a building, cost $ 500. This does not include the receiver or antenna.
The $ 2k and $ 500.00 cost detailed was for materials only, labor costs are not figured in
Follow up questions:
Are there any estimated maintenance costs of such a system? Are the on a per tower basis? Maintenance is a
fairly low cost, the only " normal" maint items are the battery back- ups that are replaced every 2 yrs approx
$ 150.00 / station.
Am I correct in assuming that the receiver is what is used by the end user? The end user is therefore responsible
for the cost of this equipment and any data plan necessary. For the rovers, yes. Each base ( CORS) has a GPS
receiver and antenna, and yes the end users are required to pay for their own data plan.
What is the antenna’s purpose? Are those the responsibility of the end user or the DOT? The antenna is the device
that collects the GPS signal, each CORS has one. These are the responsibility of the DOT
Do you have any range of costs for receivers and antennas? I am assuming the range may be significant
depending on what brand and function it is meant to serve. A high end GNSS receiver and antenna ( Geodetic
Grade) ( Like what we have on our stations) will run around $ 18k
How many users are on your system currently? How long has it been operational? Is there an industry
( agriculture, construction, etc.) that dominates the network? We currently have 964 users on our system. It has
been operational since 2004. Our users are, in order of prevalence : Surveyors, Agricultural, GIS users,
Construction machine control. Surveyors are, right now our biggest users but the Ag market has come on strong
in that last few years and I expect it to be our biggest user in the near future.
Adam, if you have any further questions, or need clarification, just let me know or give me a call
Thanks
DJB
56
Again, thanks for the timely response. Look forward to hearing from you soon.
Cheers,
Adam
On 10/ 18/ 10 5: 29 AM, " Dave. Beiter@ dot. state. oh. us" < Dave. Beiter@ dot. state. oh. us> wrote:
We built our Trimble VRS network completely in house, the Concrete monuments with foundations extending down
at least 10' at 3' diameter cost us approximately $ 2000.00 each in materials, with a three man crew for 3 partial days.
Our building mount stations cost about $ 500.00 in materials and only took us 1 day to complete with the 3 man
crew. Please note that these costs do not include the receiver or antenna.
If you need anything else, let me know
Thanks
DJB
David J. Beiter, PE / SIT, VRS/ CORS system Manager
ODOT Office of Aerial Engineering and Surveying
1602 West Broad St, Columbus, Ohio 43223
Voice: ( 614)- 275- 1372 FAX: ( 614)- 275- 1673
e- mail: dave. beiter@ dot. state. oh. us
Adam Goodliss < agoodliss@ gmail. com> 10/ 14/ 2010 07: 29 PM
To Adam Goodliss < agoodliss@ gmail. com>
Subject: RTK/ CORS Network Cost Information - UC Berkeley Research
Hello, My name is Adam Goodliss and I am currently a graduate student, in transportation engineering, at the
University of California, Berkeley. I have been working on a research project revolving around analyzing the costs
and benefits of high accuracy location data with regards to intelligent transportation systems. You can find more
information at this website, http:// ucbchalo. wordpress. com/ < http:// ucbchalo. wordpress. com/> . I have contacted
you specifically due to your involvement with CORS/ RTK networks. As part of our research we are looking into N-RTK
solutions among others. I was wondering if you had additional cost details ( most not available on your
website) from an infrastructure/ maintenance point of view, ie. the cost of a tower, installation cost, etc. Ultimately, I
want to estimate the cost of implementing N- RTK networks of differing sizes and possibly on a national scale.
I also realize that you could have possibly worked with Trimble or Leica Geosystems so if they would be more
suited to deal with this type of question please advise accordingly. If you do not have this information, but know
someone within your organization that does I would greatly appreciate being put in touch with them.
Thanks in advance, and if you have additional questions please do not hesitate to be in touch via email or at
781.888.8033. - Adam
Email from Steve Milligan to Adam Goodliss:
Adam,
The Iowa network cost approximately $ 1.70 million, and it consists of 80 reference stations. Each reference station
cost approximately $ 21,250 which included installation.
57
In the first 1.75 years of operation it cost approximately $ 60,000 for equipment repairs and replacements due to
lightning/ power surges. Software and firmware upgrades run approximately $ 80,000/ year.
Let me know if you need any additional information.
Regards,
Steve Milligan
Statewide RTN Coordinator
515- 239- 1981
515- 290- 2831 cell
From: Adam Goodliss < agoodliss@ gmail. com>
To: Milligan, Steven [ DOT]
Sent: Fri Oct 15 12: 21: 08 2010
Subject: Re: RTK/ CORS Network Cost Information - UC Berkeley Research
Steve,
Thanks for the timely response. I appreciate any information you are able to share with me.
- Adam
On 10/ 15/ 10 6: 47 AM, " Milligan, Steven [ DOT]" < Steven. Milligan@ dot. iowa. gov> wrote:
Adam,
I'm not in the office today, but I will get this information together and send it to you.
Steve Milligan
Statewide RTN Coordinator
515- 239- 1981
515- 290- 2831 cell
From: Adam Goodliss < agoodliss@ gmail. com>
To: Adam Goodliss < agoodliss@ gmail. com>
Sent: Thu Oct 14 18: 19: 09 2010
Subject: RTK/ CORS Network Cost Information - UC Berkeley Research
Hello,
My name is Adam Goodliss and I am currently a graduate student, in transportation engineering, at the University of
California, Berkeley. I have been working on a research project revolving around analyzing the costs and benefits of
58
high accuracy location data with regards to intelligent transportation systems. You can find more information at this
website, http:// ucbchalo. wordpress. com/.
I have contacted you specifically due to your involvement with CORS/ RTK networks. As part of our research we
are looking into N- RTK solutions among others. I was wondering if you had additional cost details ( most not
available on your website) from an infrastructure/ maintenance point of view, ie. the cost of a tower, installation cost,
etc. Ultimately, I want to estimate the cost of implementing N- RTK networks of differing sizes and possibly on a
national scale.
I also realize that you could have possibly worked with Trimble or Leica Geosystems so if they would be more
suited to deal with this type of question please advise accordingly. If you do not have this information, but know
someone within your organization that does I would greatly appreciate being put in touch with them.
Thanks in advance, and if you have additional questions please do not hesitate to be in touch via email or at
781.888.8033.
- Adam