Login

Project

#274 Cost-Effective Designs of Smart City Technologies for Vehicular Communications


Principal Investigator
Jon Peha
Status
Active
Start Date
July 1, 2019
End Date
June 30, 2020
Research Type
Advanced
Grant Type
Research
Grant Program
FAST Act - Mobility National (2016 - 2022)
Grant Cycle
2019 Mobility21 UTC
Visibility
Public

Abstract

In this research, we will identify appropriate application-layer quality of service metrics for each of the more prominent safety applications.  We will develop simulation software, and use that software to see how design decisions and external factors affect our measures of application-layer quality of service.  We will develop an engineering-economic model, and use that model to see how these same design decisions affect cost.    
Description
1. Introduction

Smart city technologies may enable applications to improve mobility of users and goods, to save fuel and reduce emissions, and to enhance road safety. Some of these applications are based on communications among vehicles (V2V), between vehicles and infrastructure (V2I), and even with pedestrians and cyclists. This is collectively called vehicle-to-everything (V2X) communications.  While considerable research is underway to determine how best to design V2X systems, and what is technically possible with emerging V2X standards, more research is needed to provide local, state and federal policymakers with guidance on when the use of this technology is actually cost-effective, i.e. when its benefits exceed its costs.  The proposed research will help to meet this need.

Government leaders have been motivated to adopt these technologies primarily for their safety benefits, and some studies have shown that safety benefits can be great, including fewer vehicle accidents, fewer deaths, and less property damage. However, the costs are also considerable.  For example, the U.S. Department of Transportation had previously suggested that state and local transportation agencies all over the nation should spend over a billion dollars on roadside infrastructure for V2X communications, and the federal government had no plan to provide funding for this purpose. Each individual local and state agency is left to determine whether the costs are justified in its particular region, and if so, what technical approach to take.  In addition, the U.S. Federal Communications Commission (FCC) has allocated 75 MHz of spectrum exclusively for a particular kind of V2X technology known as dedicated short-range communications (DSRC).  This resource is worth billions of dollars.  Makers of other kinds of wireless technology are now urging the FCC to make some of this spectrum available for purposes unrelated V2X, at the same time that developers of a new kind of V2X technology known as C-V2CX are saying that they need new spectrum of their own. Thus, the FCC too must decide how much of a limited resource to invest in V2X, and in which technology or technologies.  All of these actors need to understand whether and when benefits can exceed costs.

Important cost-benefit analysis has been done, but it is already outdated and inadequate in the face of a new and more complicated reality.  In particular, the Department of Transportation concluded [7] that the safety benefits from important V2X applications greatly exceeded the cost of deploying V2X roadside units (RSUs) throughout the nation, at least under a government mandate that all new vehicles come equipped with DSRC-based on-board units (OBUs).  Although these results are encouraging, they are based on many assumptions that are now in question.  One limiting assumption is that all vehicles would be equipped with OBUs, even though a federal mandate is no longer planned.  Moreover, DSRC technology was assumed for this mandate, but other technical approaches are now possible. This includes 3GPP’s emerging C-V2X standard that has been backed by some car companies (but not by the FCC), and even macrocell-based approaches which may work for some safety applications although not all.  Another limiting assumption is that all RSUs must be capable of supporting all applications, even though a less-capable but lower-cost approach might be more cost-effective in some locations, such as an RSU with little or no backhaul capability. (The PI’s work the with the City of Pittsburgh has shown that backhaul can be a surprisingly large portion of the cost of a smart city deployment, at least if the most demanding applications are to be supported.) A fourth limiting assumption was that RSUs would be built in all regions or none, when perhaps the infrastructure is cost-effective in some locations and not in others.  A fifth was that OBUs would be installed in all vehicle types or none, when perhaps the expense of an OBU is justified in vehicles that are more prone to accidents where lives are lost but not in all vehicles.  A sixth is that 75 MHz of spectrum would be available on an exclusive basis, and more importantly, at no cost; the Department of Transportation incorrectly viewed that cost as irreversibly sunk.  

 This project is part of broader research with a long-term goal of informing city leaders, federal, state and local transportation authorities, the Federal Communications Commission, automakers, and ISPs on what technical design choices and policy options are the most cost-effective for V2X road safety applications. At the broadest level, we seek to understand from the perspective of every such decision-maker what the range of technical design options are, how costs vary with design decisions, and how safety-related benefits vary with design decisions, so that informed decisions can be made about which approach is cost-effective.  These relationships may differ considerably from one decision-maker to another, e.g. urban is likely to be different from rural.  

In the coming year, we will advance these long-term objectives in several ways.  To better understand safety-related benefits, we will develop simulation and analysis tools that allow us to quantify quality of service (QOS) with a variety of technical approaches.  To make the results more directly relevant for quantifying safety-related benefits, we will define QOS metrics for a series of important safety-related applications, rather than relying on generic network-layer QOS metrics that do not necessarily affect safety benefits.  At the same time, we will quantify the burdens that these applications cause on different parts of the network, as this is critical in understanding how the applications supported relate to costs.  We will also continue to build engineering-economic models that allow us to relate technical design and policy choices to actual deployment and operating costs.  These efforts in combination advance our ability to relate network design choices to cost and benefit measures that are useful to policymakers.  

Our efforts to understand benefits and costs both build on our research of that last few years [8, 9, 15], where we developed extensive simulation capabilities to estimate performance of V2X systems and engineering-economic models to estimate costs.  However, all of our prior work was based on many of the same limiting assumptions described above that plagued the Department of Transportation studies, including the assumptions that the only available technology is DSRC, that all new vehicles would be equipped with V2X OBUs, and that all V2X roadside infrastructure is homogeneous and must be capable of supporting every application under consideration.  Our prior work will be discussed further in Section 4.

Section 2 will describe some of the design decisions that would benefit from greater cost-benefit analysis.  Section 3 describes the three major tasks we foresee to help meet that need.

2. Technical and Policy Design Choices to be Explored
One critical decision to be explored in our research is the choice of technology.  Cities must decide what infrastructure to deploy, including which technology or technologies to support.  The FCC must decide what V2X technology to allow in the Intelligent Transportation System (ITS) spectrum band.  Car companies and consumers must decide what technology of technologies to support on each car.  Moreover, it now appears that multiple technologies will exist, and a given safety application may run over one or more of these technologies.

Recently there has been a hot debate about which V2X technology is the most suitable for smart city applications. One is Dedicated Short Range Communications (DRSC), which enables both V2V and V2I communications over distances of a few hundreds of meters [1], [2]. DSRC has been tested extensively for several years by researchers and transportation agencies, mainly for safety applications [3]–[5]. DSRC is based on the IEEE family of standards and has similarities with Wi-Fi. More recently, a newer technology called cellular V2X (C-V2X) has emerged as an alternative. C-V2X enables V2V communications even where vehicles are not in coverage of cellular towers. Some claim that C-V2X incorporates technical innovations that result in superior performance compared to DSRC [6]. In addition, some smartphone-based connected vehicle applications have been deployed over cellular technologies such as 4G LTE, and in the future, 5G.  LTE chips could easily be placed in vehicles. In cellular communications, mobile devices such as phones connect to the network via downlink and uplink connections (DL/UL) with cellular towers. Not all safety applications run over these macrocells, because with this approach it is difficult for a device to communicate with other nearby devices without first discovering the addresses of these neighboring devices.  However, this limitation can be overcome by using a broadcast over macrocell capability, such as MMBS.  Such an approach consumes more resources every time it is used, and has disadvantages with respect to QOS, but the approach requires far less infrastructure to be deployed, so it is not obvious without analysis whether it is more or less cost-effective.

In addition to technology, smart city planners must make other decisions when deploying infrastructure, including whether and where to use RSUs as opposed to macrocellular connections, the number of RSUs to deploy, the types of locations for RSUs, and the type of backhaul provided for each.  Backhaul greatly affects cost, and it affects the range of potentially beneficial applications that be supported as well.  We will examine these tradeoffs.  

Similar tradeoffs occur in the capacity of the wireless connection to the vehicle, which depends in turn on the amount of spectrum allocated by the FCC, and how it is managed.  Through the same simulation tools, we can examine these tradeoffs as well.


3. Approach

This project will address both costs and benefits of V2X networks.  Costs are studied using an engineering-economic model. Benefits depend on QOS, so one task of this proposal is to develop simulation software that will allow us to quantify QOS.  However, real benefits are from safety outcomes, such as a reduction in accidents and lives lost, and the relationship between QOS as observed in the network and safety-related outcomes depends on the specific application in use.  As a result, in a separate task, we delve more deeply into the more promising safety applications to understand their demands on the network.  Each of these three tasks is explained below.


3.1. Determining the Network Demands of Leading Safety Applications

The conventional approach to determining requirements for V2X networks has little to do with the characteristics of the applications to be supported.  This simplifies the job of network designers, but it obscures some critical trade-offs that should be a part of a good cost-benefit analysis.  For example, requirements for network-layer metrics such as packet loss rate and maximum range have been established such that it appears that all safety-related applications will work reasonably well even in the worst case.  Any network with QOS that fails to meet all such requirements is inadequate, and any network with better QOS has been overdesigned.  Thus, if there is a requirement that V2V connections must work at 300 meters, a system is considered inadequate if its range is 280 meters, even though most applications don’t need a range greater than 100 meters.  As a result, some of the comparisons between technical approaches may focus on the irrelevant.

In reality, QOS can only be described probabilistically, and no network can meet any set of specific QOS requirements consistently when network load is too high, so specifications based on QOS requirements with no limit on utilization are meaningless.  How should we design systems that are cost-effective in a world where utilization is unpredictable, and occasional periods of congestion are possible?  Perhaps a system should be designed to support the highest possible utilization, as might occur when many cars are packed together in a traffic jam, but this can be expensive.  As an alternative to accepting these high costs, perhaps a system should support only the most beneficial applications, thereby reducing traffic per vehicle, and of course decreasing safety benefits as well.  To understand how much this helps, one needs to know the traffic generated for each application, and the benefit lost by dropping some applications.  Perhaps the system should instead support all applications, but accept QOS that is less than ideal during periods of high utilization. This too would reduce benefits, but not eliminate them for any application even though nominal QOS requirements are not met.  For example, if a driver gets the warning of impending danger 200 ms later than would be preferred, the probability that the warning will prevent an accident decreases, but some accidents will still be prevented.

To bring these complex tradeoffs into a cost-benefit analysis, we must understand each individual safety application, its impact on the network, and the impact of network QOS on the benefit (e.g. avoided accidents) that it can provide.  Achieving that goal is beyond the scope proposed here, but we will take an important step in that direction in the proposed research.  We begin by analyzing dozens of safety applications that have been described in detail by the U.S. Department of Transportation. We will later supplement these with applications emerging from the standards bodies and the research literature.  For each application, we will attempt to answer the following questions.   What measures should be used to describe QOS at the application layer in a way that can later be mapped into safety benefits (such as accidents avoided)?  How much traffic does this application generate in the ITS spectrum band that is allocated for wireless V2X communications, and how does this vary with conditions?  How much traffic does this application generate on backhaul connections, and how does this vary with conditions?   At what locations would this application be beneficial?  (e.g. at intersections, on tight curves, on roads that are prone to ice, ...)  Answers to these questions will be important for subsequent research tasks related to cost-benefit analysis.

This will be done as follows. A typical safety application might be designed to avoid crashes for a certain pre-crash scenario. We identify a metric based on that application and scenario.  For example, consider the left-turn assist application, which is intended to warn drivers of dangers when they make a left turn.  The characteristics of this application are known, e.g. packets are sent with a known frequency, and each packet contains time stamp, position, heading, speed, acceleration, etc.). We will use the scenario and the application characteristics to define what QoS variables are the most important for the particular safety application.  A good application-level metric for our subsequent analysis and simulation is the age of location information about the car that is most likely to cause a collision at the instant when the driver turns on the left turn signal.  This metric can eventually be translated into accidents prevented.  Traditional network-layer measures such as packet loss rate and latency are largely irrelevant, as is the state of a car that is still 300 meters away.  At minimum, we can determine the direction of the effect that a change in each application-layer QoS measure has on the safety outcomes of the associated V2X application. For example, if a QoS measure is the uncertainty in speed of the approaching vehicle, then a reduction in uncertainty (due to e.g. change in a design choice) should result in fewer crashes.  

For a limited set of well-studied safety applications, we can quantify more relevant application-layer metrics. For applications such as intersection movement assistance (IMA) and left-turn assistance (LTA), there is work [5] from which we can derive how our proposed application-layer QoS measure translates into application quality measures such as the uncertainty in the estimation of when that car will reach the intersection where an accident could occur.  For a small number of applications, this may even lead to estimates of accident probability and accidents avoided, as these previous studies have examined the relationship between accident probability and the speed at which cars are going, and having inferior application-layer QOS metrics is comparable to having the cars that are approaching the intersection move more quickly. In future, our approach can be extended to other applications as more data about their safety outcomes are made available. 

Another important part of this application by application examination is a characterization of the packets that will be exchanged on V2V links, V2I links, and RSU-network links.  This includes packet rate, packet size, retransmission behavior when packets are lost, location of transmitter at the time of transmission, and other factors that will guide our simulation analysis.  These will be important as we consider the resources that should be invested in both wireless communications and backhaul, and the associated costs and benefits.


3.2. Simulation Software to Estimate Quality of Service

We will develop simulation software to determine QoS for varying design choices and utilization, using both traditional network-layer measures of QOS and the application-layer measures developed in the task described above.  These simulation results are at the heart of a good cost-benefit analysis.  Greater application-layer QOS leads to greater benefits, such as fewer accidents and lives lost, while many of the design choices that can improve QOS in a given scenario will also increase cost.  However, it is not always a tradeoff; some strategies can produce given application-layer results at lower cost than other strategies, and simulation will help us identify these strategies.  

We have extensive experience simulating V2X networks that are based on DSRC technology [5]–[7] using software we developed on the ns-3 network simulator [4].  We also have extensive experience simulating macrocellular networks with technologies such as LTE on various platforms, including the Vienna simulator [20].  However, new approaches are needed to simulate C-V2X.  We have acquired a new open-source simulator proposed in [3], which does support C-V2X, and should be sufficient to meet our needs.  Nevertheless, we are also in ongoing discussions to gain access to the network simulator that Ford Motor Company had developed for its own use, which has advantages.  We are optimistic that Ford will agree to partner with us in this effort in 2019.  If so, we will use their software instead or in addition to the open-source alternative.

To better reflect realistic utilization conditions, vehicle mobility will be derived from mobility patterns of real vehicles from the city of Porto in Portugal, using a proprietary dataset we have already acquired for this purpose, and the placement methodology described in [8], [9]. In the future we may enrich this method of vehicle mobility by aggregating other datasets of vehicle positions, such as a dataset derived from connected vehicles in the city of Cologne in Germany [10].

We will derive application-layer QoS and network-layer QoS in a wide variety of scenarios. Each scenario will be defined by design choices and utilization. Design choices include quantity of infrastructure, capacity of infrastructure backhaul, amount of ITS spectrum, and the technical approach or approaches considered when deploying infrastructure (e.g. C-V2x, DSRC, LTE/MMBS). Utilization factors that define a scenario include population density, the quantity of vehicles equipped with V2X capability, the range of safety applications that are active and the data rates of these applications.

One result of this task is to show application-layer QoS as a function of different design choices and utilization. For example, we may show how the application-layer QoS measure for the left-turn-assist application varies with backhaul capacity of the RSU, with the other design choices and utilization fixed. Such a result would help cities determine the conditions under which high-capacity backhaul is worth the cost, which can be considerable.  As another example, we may compare application-layer QOS for DSRC and C-V2X with a variety of applications.  While other researchers have shown that there are pros and cons from the perspective of network-layer metrics [6], [11]–[13] , it is not clear whether or how these differences might matter for the applications that are expected to be most important from a safety perspective.
.

3.3. Engineering-Economic Model

We will develop an engineering-economic model that can be used to estimate the costs of providing V2X communications with different design choices. Good cost estimates are obviously essential to identifying which technical and policy strategies are more cost-effective.

We have already developed such a model for certain DSRC-based V2X networks, but this model must be greatly expanded to accommodate the broader range of design choices considered in this proposal.  An engineering-economic model of this kind consists of equations that relate design choices and scenario characteristics to costs, some of which can be fairly complex, as well as realistic estimations of numerical input variables and associated confidence intervals with evidence-based justifications.  In some cases, we will employ simulation results as well, e.g. when operating costs depend on traffic volumes or achieved quality of service.  

Using this model, costs will be estimated at the per-RSU, per-OBU and per-macrocell levels, at the citywide or regional level, and at the national level, under a wide variety of assumptions.  We will consider both deployment and operating costs, including costs of the RSU or macrocell, its backhaul (if any), monthly cellular transport fees, spectrum allocations, and more. 


4. Our Previous Work That Lays Groundwork

We have done work on cost-effectiveness, QoS, infrastructure deployment and spectrum issues related to V2X and other types of communications, as well as research on communications for public safety. We previously addressed issues that are different from those described in this proposal. However, we can rely on our previous work as base for the new research and apply our skills in the approach proposed here.

In [8], we showed that deploying infrastructure for V2X communications based on DSRC technology is more cost-effective than 4G macrocells for non-safety applications that exchange Internet traffic. This is true as long as DSRC OBUs are mandated in all new cars, such as proposed by the U.S. DOT [7], and for utilization scenarios that are representative or dense urban areas. Moreover, we focused on a specific design choice, the density of DSRC RSU infrastructure, and estimated optimal choices that maximize cost-effectiveness for several utilization scenarios of vehicle densities and data rates. Our approach was to simulate V2X communications with DSRC technology, based on assumptions derived from the Porto dataset. With the simulation we derived data throughput as the main network-layer QoS measure. QoS was then related to costs of using DSRC communications rather than 4G cellular.

In [9] we focused on other infrastructure-related design choices, which is whether to share infrastructure deployed for safety applications with providers of non-safety applications such as Internet access. Again, we used network simulation to estimate QoS and find the most cost-effective design choices with respect to density of RSUs either deployed exclusively or shared, for several scenarios of utilization.

Besides, the PI has done extensive research on spectrum issues for wireless communications in general, and V2X communications and particular. For example, in [14] he proposed several spectrum-related design choices. These choices are made by policymakers with respect to access, priority and cooperation rules that apply to users of shared spectrum, in order to enhance utilization of spectrum in efficient ways. For V2X communications in particular, in [15] we examined how much spectrum should be allocated for V2X communications, and whether part of that spectrum should be shared with non-vehicular devices such as Wi-Fi equipment.

The PI has also done work about using cellular infrastructure for safety communications. As shown in [16]–[19], a highly cost-effective way to provide communications for emergency responders involves sharing infrastructure between government and commercial cellular providers. These findings were derived using a detailed engineering-economic model of costs and design choices. This approach was adopted as “FirstNet,” a nationwide network for emergency responders which Congress funded in 2012 with $7 billion [18]. 


References

[1]	J. B. Kenney, “Dedicated Short-Range Communications (DSRC) Standards in the United States,” Proc. IEEE, vol. 99, no. 7, Jul. 2011.
[2]	U.S. Federal Communications Commission, “Wireless Services: Dedicated Short Range Communications (DSRC) Service,” 2004. [Online]. Available: http://wireless.fcc.gov/services/index.htm?job=about&id=dedicated_src. 
[3]	R. Uzcategui and G. Acosta-Marum, “Wave: A tutorial,” IEEE Commun. Mag., vol. 47, no. 5, pp. 126–133, May 2009.
[4]	U.S. Department of Transportation, “Planning for the Future of Transportation: Connected Vehicles and ITS,” 2015.
[5]	J. Harding et al., “Vehicle-to-Vehicle Communications: Readiness of V2V Technology for Application.” U.S. DOT NHTSA, 2014.
[6]	Qualcomm, “Accelerating C-V2X commercialization,” 2017.
[7]	U.S. Department of Transportation, Federal Motor Vehicle Safety Standards; V2V Communications - Notice of Proposed Rulemaking (NPRM). 2016.
[8]	A. K. Ligo, J. M. Peha, P. Ferreira, and J. Barros, “Throughput and Economics of DSRC-Based Internet of Vehicles,” IEEE Access, vol. 6, pp. 7276–7290, 2017.
[9]	A. K. Ligo and J. M. Peha, “Cost-Effectiveness of Sharing Roadside Infrastructure for Internet of Vehicles,” IEEE Trans. Intell. Transp. Syst., vol. 19, no. 7, pp. 2362–2372, 2018.
[10]	S. Uppoor, D. Naboulsi, and M. Fiore, “Vehicular mobility trace of the city of Cologne, Germany,” http://kolntrace.project.citi-lab.fr, 2016. .
[11]	A. Filippi, K. Moerman, V. Martinez, A. Turley, O. Haran, and R. Toledano, “IEEE802.11p ahead of LTE-V2V for safety applications,” 2017.
[12]	M. Fallgren et al., “Fifth-Generation Technologies for the Connected Car: Capable Systems for Vehicle-to-Anything Communications,” IEEE Veh. Technol. Mag., vol. 13, no. 3, pp. 28–38, 2018.
[13]	M. Boban, A. Kousaridas, K. Manolakis, J. Eichinger, and W. Xu, “Connected Roads of the Future: Use Cases, Requirements, and Design Considerations for Vehicle-to-Everything Communications,” IEEE Veh. Technol. Mag., vol. 13, no. September, pp. 110–123, 2018.
[14]	J. M. Peha, “Sharing spectrum through spectrum policy reform and cognitive radio,” Proceedings of the IEEE, 2009. 
[15]	A. K. Ligo and J. M. Peha, “Spectrum for Intelligent Transportation Systems: Allocation and Sharing,” in IEEE International Symposium on Dynamic Spectrum Access Networks, DySPAN, 2018.
[16]	R. Hallahan and J. M. Peha, “Compensating Commercial Carriers for Public Safety Use: Pricing Options and the Financial Benefits and Risks,” in 39th Telecommunications Policy Research Conference, 2011.
[17]	R. Hallahan and J. M. Peha, “Quantifying the costs of a nationwide public safety wireless network,” Telecommun. Policy, 2010.
[18]	J. M. Peha, “A Public-Private Approach to Public Safety Communications,” Issues Sci. Technol. Natl. Acad. Press., 2013.
[19]	R. Hallahan and J. M. Peha, “The business case of a network that serves both public safety and commercial subscribers,” Telecommunicatoins Policy., 2011.
[20]	M. Alotaibi, J. M. Peha, and M. A. Sirbu, "Impact of Spectrum Aggregation Technology and Spectrum Allocation on Cellular Network Performance," IEEE Conference on Dynamic Spectrum Access Networks (DySPAN), Sept. 2015.
    
Timeline
In the first six months, we will set up the software to simulate the more important scenarios, and to incorporate data derived from a DSRC deployment in Porto.
In parallel, we will conduct the application-by-application analysis.
In the next six months, we will obtain quantitative results from the simulation.
In parallel, we will work on the engineering-economic model.
    
Deployment Plan
Not applicable    
Expected Accomplishments and Metrics
We expect to produce concrete analysis that will inform important policy decisions by providing quantitative and qualitative results.  With one more year of funding, we hope to provide analysis on the impact of design choices about infrastructure deployment that will be useful to smart city and state policymakers, and design choices about spectrum allocation that will be useful to policymakers at the Federal Communications Commission and its counterparts.

In particular, we expect to achieve several accomplishments with this grant. We will identify application-layer quality of service (QOS) measures that relate to safety outcomes for each of the more prominent connected vehicle safety applications. We will develop simulation software, and use that software to quantify QoS as expressed in these application-layer measures under several scenarios representing different design choices and utilization.  We will quantify costs of infrastructure, spectrum, and OBUs under several scenarios representing different design choices.
    

Individuals Involved

Email Name Affiliation Role Position
peha@cmu.edu Peha, Jon Carnegie Mellon University PI Faculty - Tenured

Budget

Amount of UTC Funds Awarded
$30000.00
Total Project Budget (from all funding sources)
$

Documents

Type Name Uploaded
Data Management Plan Data_Management_Plan_final.docx Jan. 4, 2019, 6:24 p.m.
Publication Spectrum for V2X: Allocation and Sharing Sept. 27, 2019, 9:31 a.m.
Presentation Cellular Technology and Public Safety Networks Sept. 27, 2019, 9:36 a.m.
Presentation Security and Privacy Law for Connected Vehicles Sept. 27, 2019, 9:36 a.m.
Progress Report 274_Progress_Report_2019-09-30 Sept. 28, 2019, 5:37 a.m.

Match Sources

No match sources!

Partners

Name Type
Ford Motor Company 1
City of Pittsburgh None
Veniam None