• Information Resources
  • Volpe Library
  • Volpe Journal
  • Volpe Highlights
  • Points of Pride
  • Bibliographies

Advanced Wireless Communication for the Transportation Sector

Previous Section | Contents | Next Section

---

3.0 Mobile Ad-Hoc Networks (MANETs)

3.1 Overview of Trends

MIT’s David Reed presented an overview of trends in the area mobile ad-hoc networks (MANETs) and their potential to provide alternatives to existing approaches. Mobile ad-hoc networks have emerged in response to mobility and networking challenges faced by traditional networks, including the "command and control" architecture model found in many Department of Defense systems.

Reed’s expertise is in an area that might be referred to as "slightly mobile ad-hoc social networks." Due to the inexpensive nature of radio technology and the extensive adoption of wireless devices by general users, it is reasonable to project that there will be a great deal of radio communication in the human centered network. The "out of the box" view of vehicular networks assumes the perspective that users will be carrying in-car radios. Thus, the "killer app" may not be about cars, but about the cars talking to the users; however, can the car talk to the user intelligently?

An important observation about the history of MANETs is that radios can send signals multiple hops further than they can point-to-point. There are advantages to eliminating points of failure and inefficiencies in order to achieve more capacity. If a system is deployed with a lot of short distance links, you get more capacity in the system. Currently, with cell towers there is only so much coverage available, such that cell networks are using handset-to-handset signaling to maximize available cell tower capacity. According to Reed, in terms of the transportation sector, when we talk about MANETs (based on radios that adapt and cooperate) we can provide a map of the environment around us. Industry leaders and researchers are already starting to build integrated systems in the indoor communications space; this can also happen in the vehicular space.

3.2 Role of Standards

Systems with a reasonable amount of standards can take advantage of the increased capability afforded by MANETs. In places like Intel labs and Motorola labs, there is a lot of work being done to make systems that basically exploit all of the radios in the space, and also use other systems, whether they are broadband, Wi-Fi, or traditional resources. A system that has a reasonable amount of intelligence can actually take advantage of these resources. What makes this work are radios that can understand each other.

The minimum capability of either standardizing on analog wave form or standardizing some form of capability that can be adapted to each other is what is called software-defined radio. The advantage of software-defined radio is, in fact, all of these modern radios are being pulled that way, whether they expose it or not. Although they may be defined in terms of analog wave form, they are generating waveforms digitally. So, you can take advantage of those, for example, to adapt to rapidly changing populations. For programmers this means that the digital realm gets programmable access to the (analog) radio spectrum. In that sense a wealth of new applications can evolve around the principle of software-defined radios.

The crucial element of packet-systems is that the links can be short-distance. This is extremely different from the traditional radio problem of establishing channels. Much of the work on fixed wireless mesh networks and new adaptable systems shows that they can work well for many kinds of short distance applications. For many of those applications, "you’d have to stand on your head to use traditional radio systems." The result is that the safety element can be extremely dynamic with short packets.

The field is starting to see users deploy add-on devices in cars that use Wi-Fi (i.e., Dash Express GPS³, although they are not very useful because they only function with a connection. An issue with Dash and other Wi-Fi devices (which are like memory chips with Wi-Fi antennae) is that they are missing critical standards. Reed’s assessment is that it is worth considering options beyond command-and-control networks, but for vehicle applications, it is important to understand channels and standards necessary for safety-specific applications.

An important note about packet systems is that the need for a consistent channel is very short. A microsecond or two has been read into packet systems to make them work; this actually establishes a channel condition that stays stable for a reasonable period of time. This is very easy and quite normal for a Wi-Fi radio that is flying past another Wi-Fi radio. If the right processor is available, there is enough time to exchange a fairly large amount of data with multiple packets going through it.

Packet-based radio systems that dynamically build connections on a type of opportunistic connectivity can be built today. This is one reason Reed is concerned about the confusion with the term "mesh." There has been a lot of work on fixed wireless, but vehicular communication "hop by hop" can actually work quite well for many kinds of naturally "ad-hoc" applications for short distances (such as measuring traffic, interacting with a parking meter, etc.).

3.3 Advanced Wireless Applications for the Transportation Sector

Reed’s overview provided a foundation for three important speakers to discuss advances in and uses of mesh networks and MANETs. The discussants (Robin Chase, L. Aaron Kaplan, and Tim Krout) provided new perspectives and details on different aspects of advanced wireless technologies and potential transportation applications. The discussion summary has been organized around several broad topic areas and includes an overview of presentations and question-and-answer sessions.

3.3.1 The Business Case for Open Platforms and Open Networks

Robin Chase (Meadow Networks and ZipCar founder) offered her perspective on strategies for leveraging wireless communications technologies for transportation safety financing, congestion mitigation, and innovative commercial applications. Chase noted that in terms of the transportation sector, applications will be adopted in a piecemeal fashion in specific geographies across the country and rely on after-market devices.

Furthermore, she believes there is an opportunity to change the outcome of DOT transportation investments by ensuring that technical investments use open platforms that can then be multi-purposed and accessible to others to build upon. Open platforms also support innovation: new ideas can be tried, tested, and improved upon at lower cost and with lower stakes (because no entity would have to build the entire set of applications or the entire infrastructure on their own). In a rapidly changing technology environment, Chase believes the desired DOT outcomes can be financed and achieved at a lower cost and at a faster pace of both innovation and adoption if open platforms and open networks are used.

In a positive trend, there are large scale wireless network experiments underway in Minnesota, Ft. Lauderdale, and other metropolitan areas, with the expectation that billions of dollars will be spent on these projects over the next several decades. The vast majority of the larger markets have deployed (or are considering deploying) closed proprietary solutions on closed networks. The result will be expensive single purpose devices and networks that have little likelihood of connectivity with others being built around the country, and enormous amounts of embedded excess capacity in the devices and infrastructure. These types of closed systems will not make use of the latest technologies because of the lengthy time cycle between development and deployment to the market. Additionally, one can anticipate no innovation on such systems. The system provider has little incentive to innovate or improve systems since they have built to initial specifications; and other companies cannot get into these systems in order to improve or update their capabilities.

The transportation sector now has an opportunity to change the outcome by making multi-purpose investments and leveraging open platforms to spark innovation. This strategy can result in rapidly improved environments and identification of the "killer app" (according to Chase, this is congestion and road pricing). Strategies for leveraging advanced wireless communication systems must include discussion of privacy issues, an issue at the forefront of the conversation about platforms, software, and hardware.

Chase identified several historical and current barriers to enabling advanced wireless communication for transportation safety, financing, and congestion applications:

• Centralized, top-down, massive scale projects – London’s congestion pricing initiative used existing technologies, drew a cordon around the central business district (CBD), and used cameras to capture license plates. While there was a 90 percent positive license capture rate, there was a 10 percent error rate, which had a huge negative impact on the cost-effectiveness of the system. The city had spent approximately £400 million on the original system, but would now like to expand the cordon around the city (at a cost of several hundred million more pounds). Had London used a vehicle-based open mesh platform, their initial investment would have more cost-effective transactions (no camera system generating high error rates), accommodated the expanded cordon with little incremental cost, and would have provided the foundation for a "wireless London" built on the back of the congestion pricing infrastructure. Instead, because an open platform did not exist, London chose traditional closed proprietary systems for its first congestion pricing build, and opted to maintain the status quo in its expansion plans.
• A hardware environment that prevented improvements – New York City modeled its congestion pricing initiative after London’s. As New York City considered congestion pricing, it became clear that it would also be politically easier to leverage the installed base of EZ pass transponders (a closed proprietary system). This promise immediately eliminated opportunities for innovative solutions that are newly possible since EZ pass was first created more than a decade ago. It also alienated the 25 percent of New Yorkers who have not adopted EZ pass because of privacy concerns (EZ freely gives location information of specific cars to those who ask) and the system’s requirement for cameras.
• Political barriers and jurisdictional issues. Chase used New York City’s recent nine-month long political discussion as an example. While there are many reasons for the city’s inability to ultimately gain support for congestion pricing, the rigid technology constraints, posed by the closed proprietary systems proposed, were major contributors: a system that required flat fixed rates along a cordon surrounding downtown Manhattan alienated the other boroughs and made the system appear unfair. The camera technology was much contested by those seeking to protect the geographic privacy of drivers circulating in the city. Adoption of an open mesh-based platform in the car would have also reduced jurisdictional friction raised by installation of fixed hardware throughout the city; it was also considered operationally risky since such a platform had not yet been built and used at the scale required for Manhattan. This is the opportunity that can be corrected and addressed by the DOT.

Chase also noted that open platforms can be "a magnet of opportunity for the business realm." For example, if New York City established a mobile communications system, not only could it enable safety applications, but also other types of convenience or commercial transactions.

British Telecom offers a good example of this type of openness when it told broadband buyers that if business users opened their broadband access points to other BT users within reach of their access points, they too could be able to roam and use the wireless networks of other BT business users. This offering improves BT’s network efficiencies (see www.BTFON.com). Ultimately, the planned expectation is that users will be switching networks to use the most cost-effective network for a given activity. Doing this requires an open system.

Closed platforms typically associated with huge fixed/mobile systems (e.g. bus rapid transit (BRT) in Los Angeles, HOT lanes in Florida, etc.) cannot test new ideas or be platforms for innovation. An open standard platform is needed for trials that can scale with less risk, and an open platform with flexible open architecture allows the private sector to enter into the space. (Another example of a missed opportunity for a large-scale open platform trial is in New Jersey. Rather than attempt to use an open platform, New Jersey Transit (NJT) is opting, conservatively, to explore possibilities based on "tried and true" devices and networks, which limit NJT’s opportunities for future innovation and interconnectivity.)

Chase closed by reviewing the handout she provided to the group, pointing out the benefits in terms of costs and time to implementation that could be gained by pursuing standards-based, open access points and ad-hoc networks that leverage both existing and new devices and infrastructure.

3.3.2 Technical Options and Considerations

Both L. Aaron Kaplan (Funk-Feuer Initiative, www.olsr.org) and Tim Krout (CenGen) offered their viewpoints on technical considerations for enabling advanced wireless communication in the transportation network. Kaplan, whose connection to this field started when he retrofitted a laptop for Wi-Fi in Vienna, now oversees a mesh network that uses the Optimized Link State Routing Protocol⁴ (OLSR) consisting of more than 400 nodes (still a relatively small network compared to the other existing mesh networks in Europe).

Larger networks exist in Athens (2,000 nodes, fully meshed), around Barcelona (~5500 nodes, Wi-Fi but non-meshed), and in the Czech Republic, which has a wireless network covering the whole country (www.czfree.cz – Czech language site). Kaplan asserted that it is a myth that Wi-Fi networks need to be small in size. There are density issues, but the Czech Republic experience shows that it is possible to have a national network. However, Kaplan noted that complete mesh networks will be available only "in the distant future" as large mesh segments may be a more promising approach. In Vienna, the network Kaplan created is a Wi-Fi mesh network, based on inexpensive off-the-shelf 802.11x Wi-Fi equipment, reprogrammed with open source code (openwrt) and feeding off of one single fiber uplink. The uplink provides bandwidth for approximately 400 nodes. The average operational expenditure (OPEX) costs for this network are a few hundred Euros/month (~1-2 Euro per user per month). The bandwidth in the Vienna mesh network goes up to 30 mbit/sec netto. Uplink is 100 Mbit/sec netto.

Kaplan clarified that he also directs the current research and development of the open source OLSR.org project.⁵ Current trends in OLSR research include integrating GPS into routing decisions (relative speed of vehicles will be used to adapt the rate of routing packets), reducing the number of packets in the air in order to conserve bandwidth, and reducing the complexity of fast and dynamic route changes.

The OLSR.org Project has recently demonstrated that the routing daemon scales well in an 800 node mesh network on an embedded web camera with only 20 MHz CPU power (roughly equivalent to an Intel 286 processor from 1982). Scalability of OLSR to thousands of nodes is currently a reality.

Kaplan later noted that in Europe many of the community networks use standard Wi-Fi equipment and the available 2.4 GHz frequencies. One major obstacle of current Wi-Fi technology is the scarcity of the spectrum. Based on his own experience building real world Wi-Fi mesh networks, Kaplan believes that self-interference of nodes and interference with neighboring nodes is the main stumbling block for successful deployment of any kind of Wi-Fi network (meshed or non-meshed). More and better spectrum will be needed.

Overall, Kaplan agreed with Ramming’s assessment of the state of the current technology. Network coding, better layer 1 support, MIMO, more spectrum and advances in routing metrics and algorithms all combined show that very large mobile mesh networks can now be feasible.

Tim Krout (CenGen) offered an overview of the Department of Defense (DoD) innovations, including a MANET based in Iraq that includes approximately 100,000 nodes. CenGen’s vision is to serve the tactical high mobile user with the "pointiest end of the spear," including "killer apps" to improve situational awareness and provide vehicle-based video feeds.

There are several radios available in the DoD environment, including:

• JTRS – Joint Tactical Radio System
• WINT project – War Fighter Information Network-Tactical
• EPLAR – Enhanced Position Location and Reporting in-vehicle radios
• HNS – High Beam Networking System
• SRWF – Soldier Radio Wave Form (uses and open platform)
• Blue Force Tracker

DoD and specifically Defense Advanced Research Projects Agency (DARPA) have "pushed the envelope" in terms of next-generation radio communications technologies. Advances in EPLAR will make it easier to configure and program, will be dynamic and meant to be operated and maintained by the average soldier. Within the next two to three years, Krout expects to see small JTRS radios available for use in robotic vehicles, while CenGen is currently developing an open source version of a JTRS radio.

At the same time, HNS technology is an in-vehicle system delivering 50mbps per sec and has been deployed in "a few hundred vehicles." The primary research developer for the network is Cisco Systems, which suggests that there is a larger commercial market available outside of DoD.

Blue Force Tracker, which uses an l-band, is a satellite communications based device that leverages the same technology as Iridium phones. There are currently thousands of vehicles being tracked with the device. Prior to the Iraq effort, EPLAR was the preferred system, however, approximately six weeks into the Iraq war production of Blue Force Trackers was increased in order to outfit more vehicles. While not used predominantly in Iraq, the device uses a commercial-off-the-shelf (COTS) cellular card with a Wi-Fi chip. There is no transparency to the user regarding which network is functioning (cellular or Wi-Fi). DoD has been encouraging MANETs in order to expand connections to vehicles. Krout noted the importance of understanding network coding, which is a new way of passing information between nodes.

J. Christopher Ramming (Consultant) presented a detailed look at some of the technological considerations related to mesh networks and advanced wireless communication technology. Ramming recently finished a term of service at DARPA, which is attempting to make mesh networks a reality. His work asked key questions such as "is the technology ready," "what are obstacles, risks, etc."?

DARPA is part of DoD and has a razor sharp focus to take results and accelerate them into practice; DARPA solves point problems, but it does not build huge networks.

"Did DoD finish the job of building the Internet?"

According to Ramming, "not really" – there is still a lot of work to do and support for mobility is currently missing. There is an exciting opportunity to create the mobile Internet and to use it for driving applications. In the transportation sector, there are problems to solve that force us to take the next step.

Ramming then described several known technical challenges and the evolution of the technology:

Mesh networks and cellular systems are quite similar, but meshes do not rely on fixed infrastructure. The intended benefits are identical: both try to achieve ubiquitous computing ("anytime, anywhere" solutions). This goal forces a shift to a wireless medium.

Importantly, the wireless medium is challenging and very different from traditional wired infrastructure. First, "the network abstraction may not be a perfect fit for the wireless systems that are in development." Nodes and links do not adequately represent interacting electromagnetic fields. Second, Ramming cited Tim Krout’s observation that MANETs routinely experience more than 10 percent packet loss. And third, Ramming suggested that participants recall that in designing IP technologies, latency or efficiencies are not a high priority – a single wire will provide 12,800 GHz of bandwidth relative to at most 3 GHz of useable mobile spectrum. The wireless medium forces a look at efficiencies in design stage, and this is difficult to handle from a systems architecture perspective.

The Internet is one line of development that allows interconnectivity; its common packet format has been transformational and has sparked a great deal of innovation. However, it is important to remember that the Internet has not fully addressed mobility. Ramming suggested a look at how the cellular industry has gone beyond interconnectivity to support mobility. DoD MANETs are another line of development in support of mobility. "Killer apps" include hard real-time control, but this is challenging in terms of data security and data assurance.

In terms of moving forward, Ramming asserted that DOT can select its own timeline, but it will face challenges similar to DoD. Some challenges may be "postponable." He suggested developing an evolutionary approach based on prioritizing "a taxonomy of applications that can include many dimensions: vehicle-to-vehicle, comfort and convenience, hard safety versus soft safety, scheduled versus event-driven versus on-demand, location aware versus location independent, unicast and multicast, elastic traffic streaming, public sector applications versus private sector applications, latency sensitive versus latency insensitive, and single hop versus multi-hop versus completely arbitrary, point-to-point, and point-to-any-point."

Support for Mobility – Ramming discussed several ways in which the original Internet design did not include support for mobility. The seven-layer OSI model⁶ differs from the four-layer Internet model in important ways, and in particular the four-layer model does not include rich support for wireless mobility. For example, there is no formal session layer that allows for temporary suspension of a session if there is some kind of mobility event. In addition, many protocols including Transmission Control Protocol (TCP) were designed for a regime with low packet error and latency (had wireless mobility been an earlier requirement for the Internet design certain important protocols might have evolved differently). And finally, Internet Protocol (IP) addresses represent both locations and identity, with the impact that topological mobility is challenging to implement.

However, mobility is increasingly understood and the Internet Engineering Task Force (IETF) is now exploring solutions to these challenges. Example solution elements include Mobile IP for global mobility via a home agent concept, which is similar to approaches used in cellular networks but that can involve high handoff latency (up to two seconds have been reported). To address fast handoff, various approaches to local mobility are being explored, including as Hierarchical Mobile IPv6 (HMIPv6), Fast-Handovers for Mobile IPv6 (FMIPv6), and work in the IETF Network-based Localized Mobility Management working group (NETLMPP) to minimize latency and packet loss in local handoff.

With respect to routing, traditional single-path wide-area Internet routing protocols are not very appropriate for wireless MANET. However, there has now been a huge amount of work on MANET routing protocols. Ramming pointed out that there is a good deal of mesh networking using OLSR that is appropriate for fairly static regimes. Efforts like Ad hoc On-Demand Distance Vector (AODV) routing assume much higher rates of mobility; at the extreme mobility time scales where the solution is not routing, intelligent flooding becomes the key enabler of mesh networks.

Finally, NACK-oriented reliable multicast (NORM), backpressure-based transport algorithms, Disruption Tolerant Networking, and the recommendations for TCP tuning in RFC3481 are all a step toward transport protocols that are wireless-friendly. With respect to wireless mobility, Ramming’s recommendation is that DOT chooses a set of initial protocols and put an architectural "stake in the ground." While this may be controversial, it seems important for progress in this area to have a specific "straw man" architecture for community consideration. Ramming believes that "no other Federal agency appears to have the specific challenge or the responsibility that DOT has."

L. Aaron Kaplan (Funk-Feuer Initiative) noted in a later clarification that link state algorithms were integrated into the Vienna OLSR network. Additional practical experimentation will integrate GPS-based devices so that intelligent flooding based on relative speeds of vehicles can occur. In a later clarification, Kaplan also explained that there are two aspects of the term "mobility": (1) mobility for small mobile devices such as cell phones or PDAs. These require AODV since AODV saves battery life, however, AODV does not permanently send packets, and, (2) mobility in terms of vehicular networks where fast handovers are a strict requirement. In these settings, AODV is not suitable since it needs to (reactively) determine the complete route before sending the first packet. Subsequent requests will also be quick, but if the topology changes rapidly (as in car meshes), it starts a new cycle of asking for the best route.

Kaplan believes that this is why CenGen is using OLSR for vehicular networks rather than AODV: they are more mobile (in the second meaning). For saving battery life, AODV is the optimal solution. And thus it is the algorithm of choice for some scenarios, like the $100 laptop that tries to save power. Additional information can be found at folk.uio.no/kenneho/studies/essay/.

Sascha Meinrath pointed out that link state work was undertaken 2001 and 2002, and that DoD is now rediscovering these routing protocols that are "half a decade old." He noted that by 2004 or 2005, developers working with OLSR+Fisheye (Fisheye is HSLS ported into Linux) began to merge it with the best of AODV and HSLS. This presents a classic example of where open source is far ahead of real world implementation.

Babak Daneshrad (University of California – Los Angeles) noted that it is important to remember layer 1– the physical layer, which is "subservient to everything else above it." In terms of Bluetooth, Wi-Fi, and cellular, the physical layer is different in each scenario for a reason because usage scenarios are different.

Daneshrad noted that participants need to address these issues of different types of usage scenarios (from in-car video to the passing of tiny messages back and forth between vehicles and infrastructure). For example, if a system operated on 2.4 GHz band, interference mitigation could happen at the physical layer but also at higher communications layers. He asked, "Can we look at interference mitigation and layer 1?" If Wi-Fi is ubiquitous, how does the field deal with delivering very sensitive and critical information in an "interference-rich environment?"

Daneshrad asked participants to think about notion of "elastic links" – stretching out a link until the network catches up and then performs the crossover. For example, a link can be stretched from 100 megabits per second in a room, to 100 kilobits per second on long distance links stretching several kilometers.

Richard Howard (Rutgers University) echoed Daneshrad’s comments and noted that users can move from texting on typewriter to YouTube because there was so much additional capacity available on the network. However, with radio (GSM), users cannot move from data or audio to video without using the entire spectrum from one channel. Howard considers this more than an "engineering problem" and a fundamental issue.

Ramming believes that mobility as a whole is an "addressable problem" with the technology available today, though difficult choices will have to be made. However, he offered that the "fully connected mobile ad-hoc network" is a myth and that wireless-friendly reliable transport protocols (for example Disruption Tolerant Networking) will need to replace or augment TCP in certain future mobile environments and applications.

Transport Protocols for Sparse and Dense MANET Deployments – Ramming went on to discuss two important regimes:

1. Sparse and partial MANET deployments (early stages)
2. Dense deployments (later evolutionary stages)

Ramming noted that in the early days of deployment and currently, there is always downtime in DoD’s MANETs. Even on I-80 in LA [sic] at night, the potential for disconnection in a fully equipped vehicular network could be as much as 35 percent. Disruption-tolerant networking can be an enabler of applications in these sparse regimes. Disruption tolerance allows a shift away from the idea that there needs to be a link from the source to the destination at all times in order to make communications work. Systems do not merely copy-and-forward at intermediate nodes; rather, the model is copy-carry-store-forward. This approach can vastly improve situational awareness application discussed earlier and is even used in developing countries to enable communications. DOT could consider using similar protocols for certain delay-insensitive applications.

The dense regime is also important to consider and is equally interesting. In an extremely heavily loaded network, it is important to look at network coding (the primitives for networking: first there were circuits in a telephone network, then packets in the Internet; network coding is the next step operating on bits and information). The network coding concept is not link oriented but broadcast oriented and allows information to be transmitted even with unreliable connections. This is a form of intelligent flooding that may allow for improved reliability in the future for certain safety applications but should be viewed as a "bleeding edge."

Information Security – An additional issue is information assurance and security. Wireless networks are changing the cyber-defense paradigm because there is no perimeter to defend. This transforms the problem of defense from perimeter to insider threat. DARPA has several programs underway to address this, but the problem is challenging. The key approach may be to focus on the security solutions that are available now, rather than attempting to address every possible threat. For example, protocol leashes, watchdogs, and reputation systems allow parts of the network to police each other and identify harmful behavior. These solutions force all parts of the network to "stay within the rules." Some ideas for dealing with network integrity and cyberattack are not completely ready to deploy, but this may be tolerable in the short run because some of the "soft safety" applications that are being considered by DOT may not require extreme assurances.

Krout (CenGen) agreed, noting that security issues are not only present in vehicular networking, but that the problem is being addressed in the commercial world. The available techniques approach the solution by making problems visible and transparent, such as policing cars. Once harmful actions are visible, the problem can be reduced.

Ramming offered that authentication, authorization, and accountability are other issues to consider. There have been advances in ideas like proof-carrying authorization, physical layer authentication and fingerprinting that allow system administrators to determine if information is coming from the node that is broadcasting (making it much more difficult to spoof an automobile or some other center of information). However, there is no complete solution, and many vulnerabilities lie with human error and misconfiguration. System administrators must take some risks, and complete security may not be achievable.

Michail Bletsas (One Laptop per Child) framed security issues in terms of encryption and privacy, not information integrity. He cited the effectiveness of end-to-end approaches, and noted that users do not rely on the layer 2 encryption for credit card transactions; rather users rely on Secure Sockets Layer (SSL)—the end-to-end encryption approach.

He continued with a word about MANETs: if they are married to network coding at the application level, one "killer app" is social situational awareness. This allows users to make intelligent decisions about traffic, but does not require the user to know his or her own position, only those of others around them. He posited that there is no efficient way to use cellular technology to propagate that information down the network, so MANETs have advantages. He suggested that participants refrain from attempting to use MANETs to replace solutions that already exist and work well. MANETs can extend or augment what is available.

Scalability – Ramming noted that the field has only have seen small deployments of MANETs, and that there is a pernicious assumption that MANETs will be used like a telephone network (where users can call anywhere, anytime and use all of the multiple hops). This results in dismal scaling results in research papers. On the other hand, many of the DOT applications only need to have communications within a few hops. So long as DOT does not try to create a MANET-based general-purpose cellular network, there is hope that current scaling laws will not be problematic.

Latency and Other Recent Developments – Ramming noted that DARPA’s Tactical Targeting Network Technology (TTNT) tries to achieve 2 millisecond packet transfer at 100 nautical miles (very low latency). He further noted that MIMO and directional networking may also provide benefits for DOT applications, and that emerging hop-by-hop protocols may resolve problematic interactions between MAC scheduling and congestion management techniques in protocols like TCP.

Ramming underscored the need for standardized protocols for mobility applications and consideration of dynamic network coding (like DTN) to improve robustness in the short term.

Following the technical presentations, the floor was opened for questions and discussion. Note that the dialogue is organized by topic areas, not by speakers’ names or chronological order of questions. Discussants’ affiliations are noted in parentheses.

Kevin Lu (Telecordia) observed that his company realized it would have been too costly to install optical fiber in every house, and shifted the focus to cable and DSL modems. Lu noted that The Telecommunications Industry Association is ready to work with DOT to develop standards for vehicular telematics. Some examples of relevant technologies and industry players include OnStar, ATX, Connexis, CrossCountry, and Qualcomm. Lu suggested that there may be opportunities to team up with automakers for commercial applications, such as the way Ford partnered with OnStar (the deployment took 10 years, but there were 4.5 million calls from active subscribers last year alone.

Robin Chase noted that given the long time to turn over the fleet in America (25 years), innovation will necessarily be in after-market devices being installed in the existing 240 million vehicles already on the road.

In terms of ITS solutions available today, Francois Simon (ARINC) asked, "Is MANET safe for safety messages?" and how can the number of nodes required be minimized? These questions remained unresolved (see Section 4.2). Simon posited that perhaps layers can be identified and separated to use different systems, so that you separate the DSRC layer from the MAC layer and put everything above DSRC on a MANET. The concept is that applications are placed in separate "buckets" that can leverage different types of systems.

Cisco Systems’ Howard Lock agreed and suggested that DOT should be looking for scalable physical systems and radios that can be isolated from the upper architecture. He further suggested separating the routing layer from the stack, and isolating the service layer.

Lock then pointed out that security remains a key concern, and that open systems like MANETs are potentially vulnerable to malicious code and mischievous behavior. As an anecdote, he noted that he would not want to rely on a mesh network to transmit safety information or personal critical data.

3.3.3 Business Models

Krout discussed various business models and opportunity areas for leveraging advanced wireless communication technologies:

• Hard and soft safety applications – What makes the transportation applications unique in terms of vehicular applications is the small amount of time relevant to situations, for example, only 50 milliseconds between a few vehicles can avoid a crash.
• Singe radio systems – These may be viable where the "80 percent solution" is acceptable.
• Wi-Fi approach for safety applications – Safety applications in a few thousand vehicles could be deployed fairly quickly using a Wi-Fi based approach using two, three, or four-hop communications. A MANET could be technologically feasible in an application that requires quick reaction and low latency, but it is probably not workable for safety applications.

There were several comments from participants related to business models and the business ecosystem.

While discussing the services approach to vehicle-based technologies, Dennis Govoni (Sun Microsystems) offered that the automobile is not an open system, and that it will be challenging (if not impossible) for open source communications to get into the automobile market. Govoni suggested leveraging private sector resources to help solve technology problems. He noted that the Internet can serve as a model for solving problems for two key reasons: first, the Internet has a very simple protocol (packet protocol), and second, the Internet "doesn’t remember anything" – it is a fluid system, and it does not matter if the network is Wi-Fi, copper wire, fiber, etc. Thus, vehicles must have an open interface to map to this.

Govoni pointed out that recently purchased vehicles include a "maintenance minder" that connects to a network. "Automakers are already thinking about solutions, and a potential next step for [the group] is to select several good use cases and see what infrastructure is really required as a platform for selected applications."

Howard Lock also pointed out that as DOT moves towards demonstration and deployment, small scale test are not relevant for large scale roll-out. His suggestions for moving forward include:

• Broaden the conversation to include other industries (such as hospitality, building trades);
• Involve other Federal departments who can participate through governance or funding; and
• Pursue innovation through a business model; do not approach the issue as a technology problem.

Lock also suggested an interesting approach to use: "the consumer carries their devices and the auto builder simply produces a platform [the user] plugs into." This leverages the "increased sophistication of consumer devices" and shifts the risk and responsibility for communications tools from automakers to consumers.

From a local government perspective, some municipalities have benefitted from extensive fiber optic and Wi-Fi infrastructure. Dana McDaniel (City of Dublin, Ohio) noted that the surge towards Wi-Fi being successfully deployed uses an anchor-tenant model based on departments like DOT, State DOTs, or cities to determine whether or not to use Wi-Fi to deliver efficient service. He offers central Ohio as a test site, as Dublin owns 25 percent of its network.

Howard Lock (Cisco Systems) offered some final advice on business models:

1. Avoid technology religion: There are many different perspectives on the issue, and players should be open to many solutions. For example, when asked about security, Lock said it needs to be pervasive in the network on a end-to-end basis. It’s less important to be "stuck" on a particular technology, but to focus instead on the business model.
2. Do not be too far ahead of the curve: Cisco recognized that a company or application may fail if too far ahead of the curve. If DOT is making investments in VII and the timing is right, he suggests a small-scale investment and a focus on "ruthless execution."
3. Connect the dots: Steve Jobs (Apple, Inc.) talks about connecting the dots, meaning, "killer apps" are almost never clear looking forward, only looking back. They will appear after a system is built, so it is important to rethink the need to identify the "killer app" before launching the system.

3.3.4 Practical Considerations for Field Applications

The group as a whole raised many practical considerations for field applications of advanced wireless technologies. Tim Krout (CenGen) noted that actionable steps are important. He identified five to six cities that received large amounts of research dollars to investigate these technologies.

Krout agreed that DOT could put a "stake in the ground" in terms of mobility protocols. From a deployment perspective, he suggested a model where 5,000 transceivers are installed in vehicles to create a test bed in a small-sized city. This could allow for advancement and innovation. A generic Linux processor could be used in a shoe-box-sized system to create a test bed with "amazing results" and substantial improvement in all areas.

Stan Pietrowicz (Telecordia) asked participants to consider Paul Brubaker’s opening remark about how to accelerate deployment of mobility and safety applications in this space. He noted that mobility and safety may not be served by the same platform, system, or same radio. Safety applications for vehicle-to-vehicle operations are different from mobility applications. Safety requires communications between strangers, low latencies, and features privacy issues. The first mandate for safety applications is that at least two cars have to have the system in place. At the normal rate of vehicle attrition, that equates to a 20- to 25-year fleet turnover cycle.

He continued that in-vehicle message signage indicating a crash two cars ahead is shortsighted, and privacy considerations create unique requirements and authentication challenges. It is not desirable to transmit inaccurate data, especially if vehicles (not only their drivers) are reacting to that information. Thus, intrusion detection systems will be important – not necessarily for encryption but for authentication (with the complex follow-on issue of determining how to remove intruders from the system). With regard to mobility, vehicle routing and traffic information, these could be subscriber-based "opt-in" information services. This would complement mandatory safety applications.

Christopher Wilson (TeleAtlas) offered an idea regarding authentication: ranging signals built into the communications signals either on OFDM, 802.11, or on top of DSRC systems. David Reed (MIT) noted that OFDM ranging is not a current challenge, but questions remain about how to extract data from the MAC layer off of a Wi-Fi card. He asserted that if you can measure the time, you can measure the phase. Professor Reed agreed that this approach might be possible, and noted that ranging, not at speed, is something some vendors are looking at.

Practical Procurement and Maintenance Issues – Ferdinand Milanes (CALTRANS) offered a sobering perspective on procurement and maintenance, stating that it is difficult for public entities (such as CALTRANS) to purchase a product through sole-source procurement when only one vendor exists. This scenario can present barriers to implementation.

Next, Milanes suggests considering the maintenance back-end. Field staff with minimal knowledge or expertise in communications technology would be responsible for maintaining state-of-the-art equipment. This needs to be factored into system design. Furthermore, it would be important to minimize field staff exposure to safety hazards of working near busy roadways and reduce the need for roadside equipment. Alternative energy also might be necessary as often there is no AC power on the roadside.

Milanes suggested partnering with the private sector where possible. California has a wireless right-of-way (ROW) agreement with several cellular companies, which offer CALTRANS space on their towers. The State DOT installs repeaters on the ROW instead of geographically challenging areas (like mountaintops). Finally, he asked if major telecoms like Nextel and Cingular might ask for the DSRC bands that the FCC has allocated (the 5.9 GHz ITS and 4.9 GHz public safety bands) if public entities opt not to use these.

Kirk Steudle (Michigan DOT) offered some background about Michigan’s two mesh networks which are fully functional. Steudle indicated he took part in a seminar with Shelly Rome and Richard Paul, and realized the importance of installing a mesh-based Wi-Fi network on a bridge. Thus, the DOT was challenged to install a system on the Mackinac Bridge in three weeks to monitor the structure during the annual bridge walk that has attracted up to 85,000 participants. The system is Wi-Fi based and functional. Steudle is willing to give others a demonstration of the system in Michigan, if interested. It is being replicated for other bridges and exemplifies how wireless communications can be used for purposes of interest to State DOTs beyond traffic safety and congestion management.

Lastly, in terms of practical considerations, Barry Einsig (Tyco) suggested that separating security-dependent safety applications from commercial applications can avoid legal issues. He suggests deploying experimental projects in confined areas (e.g., the island of Hawaii) in order to reduce intrusion or contamination by other users. The demonstration project could have two parts: one that features all safety applications and one that features all of the commercial applications.

Previous Section | Contents | Next Section