Urban Vehicle Networks: safe navigation and beyond - Welcome to Network …netlab.cs.ucla.edu/wiki/files/NSF_Vehicles_and_Internet_r…  · Web viewIn fact, the ubiquitous presence

NeTs-FIND Collaborative Research:Emerging Vehicle Networks: new roles for the Internet

Team: UCLA, Rutgers, UADeadline Jan 22, 2007

1 Summary Vehicle communications are becoming increasingly popular, propelled by navigation safety requirements and by the investments of car manufacturers and Public Transport Authorities. As a consequence many of the essential vehicle grid components (radios, Access Points, spectrum, standards, etc.) will soon be in place (and paid for) paving the way to unlimited opportunities for car-to-car applications. In this study, we take a visionary look at these emerging applications and examine the role of the Internet infrastructure in their support. The type and level of support will vary depending on the application. For instance, during e-mail downloading the vehicle is just a one hop wireless extension of the Internet. While, during Katrina type emergencies that knock out the infrastructure, the vehicular net will operate like a pure ad hoc network, yet, other applications such as location sensitive content sharing (eg, advertisements) rely on both Internet access and on Peer to Peer, “opportunistic” networking. It is conceivable that in the future most access will be from mobile (or at least portable) terminals. Thus, it is important to understand what new requirements are posed to the Internet by “mobile” applications. The main goal of this project is to identify the urban Internet infrastructure role in the support of the emerging vehicular applications. In fact, the ubiquitous presence of the infrastructure sets the vehicle grid apart from traditional, instantly deployed ad hoc nets, even when the vehicle network is operated in ad hoc mode. As the vehicular applications range from e-mail and voice over IP to emergency operations (natural disaster, terrorist attack, etc), the services requested from the infrastructure will vary. We investigate the impact of the Internet infrastructure in the following areas: (a) addressing – namely, geo addressing versus more traditional hierarchical addressing; (b) directory service support, service discovery, mobile resource monitoring, and mobility management; (c) congestion management; (d) path quality monitoring and QoS support; (e) privacy, anonymity, incentives; (f) smooth transition to fully independent, emergency mode operation (eg, fall back directory service in case the infrastructure fails). The required Internet services span the network and transport layers. The network services will be implemented via overlays or in Planet Lab. To achieve our goal, a critical task of this project will be to identify and characterize the vehicle applications, and in particular their Internet services needs. We will select a set of representative vehicular applications that have appeared in the literature and will analyze their performance requirements in the very challenged urban environment. These applications are supported by a variety of network and transport protocols that include network coding, epidemic dissemination, geographic routing, TCP, UDP, etc. The novel security issues posed by the mobile nature of the applications will also be considered, and representative solutions will be selected for this study. Since mobility plays a key role, a substantial effort will go into the development of realistic motion models. An overarching goal of this task is to define guidelines for the design of new applications so that they can best exploit the Internet infrastructure while being capable of fully autonomous operation in case of total Internet failure.A key component of this project is a Campus vehicular testbed that implements the above mentioned vehicle protocols. The testbed will support P2P vehicle communications in a pure ad hoc, multihop fashion. It will also interface to the Internet interconnection. The Campus testbed will interface with Internet overlays and with PlanetLab. In addition to the physical Campus tesdbed, we will use a simulation testbed and hybrid emulation testbed. The latter test facilities will heavily leverage existing equipment and tools developed at UCLA under the NSF sponsored WHYNET project.The intellectual merits of the project are: the identification, implementation and evaluation of critical Internet services required by the emerging vehicular environment, and; the definition of guidelines for the design of new applications that can smoothly interface with the Internet. The study will characterize the vehicle performance requirements and will generate benchmarks for evaluating and comparing new Internet services. This project will have an important educational impact. The Campus testbed experiments will be integrated into class projects, encouraging U/G individual study and graduate level research into vehicular communication, routing and security

issues. As for broader impact on Society, car manufacturers will use our findings to develop safer cars and more productive mobile-offices. Transport Authorities will manage vehicle traffic more efficiently with reduced transfer delays for the benefit of the urban population at large.

Prior Research resultsMario Gerla has been the PI on “Scalable Routing and Multicast in Mobile Ad Hoc Wireless Networks” (NSF 9814675, 2000-2002, $300K), which led to the design of the scalable routing protocol LANMAR and the very robust multicast protocol ODMRP. Both protocols have been presented at MANET IETF working group meetings. Project papers and results are found in www.cs.ucla.edu/NRL. OPTIONAL: Additional prior results by Ray, Marco , Giovanni, Xiao Yan

1. Introduction and Project Overview (2 pages)Safe navigation support through wireless car to car and car to curb communications has become an important priority for Car Manufacturers as well as Municipal Transportation Authorities and Communications Standards Organizations. New standards are merging for car to car communications (DSRC and more recently IEEE 802.11p). There have been several well publicized testbeds aimed at demonstrating the feasibility and effectiveness of car to car communication safety. For instance, the ability to rapidly propagate accident reports back to oncoming cars on the highway, the awareness of unsafe drivers in the proximity and the prevention of intersection crashes. The availability of powerful radios on board of vehicles, and of abundant spectrum (when not used for emergencies) will pave the way to a host of new applications for the “vehicle grid”. These emerging applications span many fields: extended office (e-mail, file transfers, group work), entertainment (mobile internet games, multimedia, news), e-commerce (mobile shopping), crime investigation, civic defense, etc. Some of these applications are just mobile extension of fixed internet applications (eg, e-mail, games). Others are location “aware”, ie, correlated to neighborhood resources and services (ex, restaurants, movie theaters, etc). Others yet involve not only “awareness” of the environment but also a close “cooperation” among cars, leading for instance to maintenance of distributed indices, creation, “temporary” storage and “epidemic” distribution of sharable content. Examples of the latter class include the collection of “sensor data” by cars acting as “mobile sensor platforms”; the sharing and streaming of files using “Car-torrent” P2P software, and; the creation/maintenance of massively distributed commercial, entertainment and culture information data bases. How can the Internet better support the vehicles? This is a legitimate question since the Internet was originally designed for fixed Hosts. Today, the great majority of Hosts are movable, if not moving. One expects that because of mobility some additional network services will be required. Let us consider for example the most basic vehicular applications that are a simple, one hop mobile extension of conventional fixed Internet applications, eg e-mail. In this case, the vehicle initiates the connection, thus simplifying the addressing and mobility management issue. An extension of the DHCP model can be used. If the urban cell is small and the vehicle speed is high, it may be necessary to use address tunneling (or IPV6) to maintain the connection during the handoff from one cell to another. Soft handoff becomes critical if the car is downloading real time video (eg. video conference, or news clip downloading, etc). The server may also wish to “probe” capacity to mobile, to determine the current allowed data rate. Moreover, the user may wish to conceal its location from the server, thus, there is the additional anonymity requirement. Consider now another application, this time inspired to civilian protection. Suppose there has been a bomb threat, and a number of police vehicles are dispatched to patrol an area of the city a few miles in diameter. Agents need to exchange with each other multimedia data picked up by their vehicles (video, sensor data, position, etc). Moreover, they need to contact possible witnesses and download multimedia data from them. This operation requires vehicle to vehicle connections over multiple hops. If the distance is significant, the Internet may used as a shortcut for part of the path, exploiting integrated routing via the ad hoc net and the infrastructure.From these simple examples we note that the Internet services must be extended to provide: (a) addressing – namely, geo addressing versus more traditional hierarchical addressing; (b) directory service support, service discovery, mobile resource monitoring, and mobility management; (c) path quality monitoring and QoS support; (d) security

support (anonymity). We will see later that other important requirement will emerge including: congestion management, and; smooth transition to fully independent, emergency mode operation (eg, fall back directory service in case the infrastructure fails).

In this project, we will address the interaction and synergy of the vehicle applications and protocols with the Internet infrastructure. In particular, we will investigate the issue of addressing (eg transparent geo-routing) across the Internet. Related to addressing is the maintenance of a vehicle location service both in the Internet and in the vehicular grid. Most of the Internet services in our study will be supported by a “vehicle overlay” that, among other functions, will offer the ability to estimate the (continuously changing) path quality to vehicles. These results will provide useful inputs to the design of the future Internet architecture based on vehicle needs. Our plan is to exploit our existing WHYNET testbed (separately funded by NSF), properly augmented with a physical Campus Vehicle Testbed (C-VET) in order to evaluate the vehicle P2P protocols and their interworking with the infrastructure Internet protocols. Likewise we will leverage WHYNET and on ongoing NSF and PATH funded projects for physical and MAC layer designs required for the vehicular testbed implementation. The collaboration with Rutgers will also open the opportunity to use the Rutgers ORBIT emulation platform as well as the Rutgers Vehicle Testbed currently under development.The rest of this proposal is organized as follows. Section 2 reviews related work and recent trends in vehicular network and transport protocols. Section 3 reviews vehicle applications and defines a representative set, to serve as starting point for our research. Section 4 describes the three main research directions of this project, namely: (1) vehicular protocol and architecture extensions that relate to Internet interfacing; (2) vehicle oriented Internet services, and; (3) vehicular test-bed and related experiments.

2. Related work in vehicle networks (1-2 pages)

Note: more work needed about previous research (including ours in WICON 06) on interaction of VANET with Internet Infrastructure; also, security, incentives etc. Uichin Lee, Claudio Palazzi, JS Park, Alex,

Much of the literature on Inter-Vehicle Communications (IVCs) is navigation safety related. A good survey on recent physical layer technologies for IVCs can be found in [Luo04]. At the network layer, the most common way to broadcast safety messages is via reliable, robust flooding. However, the efficiency of flooding quickly decreases with the number of nodes; thus, flooding must be scope-limited. For scalable delivery, researchers have proposed georouting and further, have focused on exploiting innate characteristics of vehicular networks such as high, but restricted mobility. For example, Urban Multi-hop Broadcast (UMB) [Korkmaz04] features a form of redundant flood suppression scheme where the furthest node in the broadcast direction from a sender is selected to forward and acknowledge the packet. The scheme alleviates broadcast storm and hidden terminal problems. In [Wischhof05], vehicles collect only the information relative to a given locality (i.e., a road segment). The paper further investigates the influence of broadcast rate on data propagation taking mobility into account and adapting the rate to traffic conditions. In this field, we have proposed a scheme by which each vehicle is able to estimate its transmission range and to put it to good use to reduce redundant transmissions and the number of hops a broadcast message has to traverse to cover a certain area of interest or reach a certain geographical location. As a result, broadcasting traffic and delays are reduced thus allowing efficient delivery of, for instance, alert messages for traffic safety applications and video triggering messages for entertainment and first responders operations [PFRPG07] [RGPFP07].

and in the and redundant transmissions the time required to broadcast a certain message over a certain area of interest. Inumber of hops For point to point unicast communications, geo-routing has been extensively investigated. A critical issue is the relaxing of beacon message requirements (and associated overhead). [Füßler03] proposed Contention-Based Forwarding (CBF). This scheme does not require proactive transmission of beacon messages for current location advertisements; instead, data packets are broadcast to all direct neighbors and the neighbors themselves decide if they should forward the packet based on a distributed timer-based contention process. A similar approach was proposed by Zorzi in GeRAF[Zorzi03] exploiting staggered MAC inter-segment intervals. In [LeBrun05], the

authors proposed a set of knowledge-based opportunistic forwarding protocols that use geographic information such as motion vectors. [Zhao06] proposes to reduce the delay to a known destination through mobility prediction. Finally, geocasting services were proposed to disseminate messages to all nodes within a target region. MDDV [Wu04] aims to support geocast by forwarding a packet along a predefined trajectory geographically. It works even with intermittent connection; intermediate vehicles must buffer and forward messages opportunistically, exploiting mobility. Abiding Geocast [Maihöfer05] features a lifetime constraint; namely, it restricts the delivery of messages to all the nodes that are in the geocast region “sometime” during the geocast lifetime. At the applications level, several cooperative peer to peer type schemes have been proposed for vehicular environments. TrafficView[Nadeem03] disseminates (through flooding) and gathers information about the vehicles on the road, thus providing real-time road traffic information to drivers. To alleviate a broadcasting storm problem, this work focused on data aggregation based on distance from the source. EZCab [Zhou05] is a cab booking application that discovers and books free cabs through vehicle multi-hopping. Free cabs are discovered with probabilistic flood search (static or decreasing probability as hop count increases). After discovery, georouting is used to negotiate the fare etc. [Xu04] proposed an opportunistic resource discovery protocol with a finite-buffer space model. The resource is a spatio-temporal resource, e.g., the availability of parking in a parking lot. A vehicle either “senses” the resources or obtains new resources from passing vehicles. Nodes exchange local databases and each keeps a fixed number (the size of buffer) of relevant resources. In PeopleNet [Motani05], a wireless virtual social network is used to support searching for spatio-temporal information, yet it exchanges resources by random swapping. Vehicular Information Transfer Protocol (VITP) [Dikaiakos05] provides on-demand, location-based, traffic-oriented services to drivers using information retrieved from vehicular sensors. A user “location-aware” query is forwarded to the target location where virtual ad hoc servers (VAHS), i.e., collection of private vehicles, resolve the query.

3. Proposed Research (7 pages)3.1 Vehicular Applications and internet requirements (2pages)

Note: this section must be reduced and better focused (Mario to take a crack at this Monday PM)

3. 1.1 Vehicle Protocol Requirements and challenges Vehicular networks provide a promising platform for future deployment of large-scale and highly mobile ad hoc network applications. With the increasing deployment of urban wireless access points, the application domain of traditional mobile and ad hoc networks (MANET) is giving way to wireless mesh networks that extend wireline connectivity to the Internet. Vehicular ad hoc networks (VANET), however, remain a largely unexplored platform for a class of compelling applications. Given the automobile’s role as a critical component in peoples’ lives, embedding software-based intelligence into them has the potential to drastically improve the user’s quality of life. This, along with significant market demand for more reliability, safety and entertainment value in automobiles, has resulted in significant commercial development and support into deployment of vehicular networks and applications. In this section, we outline key differences that distinguish the vehicular platform, introduce applications by their interactions with data, and describe a number of constraints and challenges for the vehicular application infrastructure.In designing protocols for the next generation vehicular network, we recognize that nodes in these networks have significantly different characteristics and demands from those in traditional wireless ad hoc networks deployed in infrastructureless environments (e.g. sensor field, battlefield, etc). We identify several key differences from traditional ad hoc networks with significant impact on application infrastructures. First, automobiles have much higher power reserves than a typical mobile computer. Power can be drawn from large on-board batteries, and recharged as needed from a gasoline or alternative fuel engine. Second, automobiles are orders of magnitude larger in size and weight compared to traditional wireless clients, and can therefore support significantly heavier computing (and sensorial) components. This combined with plentiful power means vehicular computers can be

larger, more powerful, and support components such as large capacity secondary storage (up to Terabytes of data), as well as powerful wireless transceivers capable of delivering wire-line capacities to mobile peers and satellite access points. Third, automobiles travel at speeds up to one hundred miles per hour, making sustained vehicle-to-vehicle communication difficult. However, existing statistics of vehicular motion, such as traffic patterns during commute hours, can be used to develop sophisticated mobility models much more realistic than the current random waypoint models. By accurately characterizing vehicles’ tendencies to travel together, these models can help maintain connectivity across mobile vehicular groups. In addition, vehicles’ extremely high rates of mobility reduce the reliability of hardware components. Networking and storage components are particularly vulnerable. The large storage capacities mean the loss of data availability is higher for each storage or network component failure. Finally, vehicles in a grid are always a few hops away from the Infrastructure (WiFi, cellular, satellite, etc). Thus, network protocol and application design must account for easy access to the Internet during normal operation. In our project, we also consider the value of the vehicle grid as emergency network when all else fails. We must therefore design protocols and applications that survive (with possible degraded performance) when isolated from the Internet.3.1.2 Innovative ApplicationsAnother important departure of vehicle networks from conventional ad hoc networks is the opportunity to deploy, in addition to traditional applications, a broad range of innovative content sharing applications (typically referred to as Peer-to-Peer applications). While their popularity has been well documented, they have been thus far confined to the fixed Internet (e.g., Bit Torrent, etc). The storage and processing capacity of VANET nodes make such applications feasible. Moreover, the fact that car passengers are a captive audience provides incentive for content distribution and sharing applications that would be unsuitable to other ad hoc network contexts. One of the key goals of this project is to understand the role of the vehicle in these applications, that is, to determine what VANET applications, both conventional and peer-to-peer, need from their infrastructure. Here, we describe a representative set of VANET P2P applications and classify them by the vehicle’s role in managing data: as a data source, data consumer, source and consumer, and intermediary. First, the vehicle provides an ideal platform for mobile data gathering especially in the context of monitoring urban environments. Each vehicle can sense events (e.g., images from streets or the presence of toxic chemicals), process sensed data (e.g., recognizing license plates), and route messages to other vehicles (e.g., forwarding notifications to other drivers or police officers). Because vehicular sensors have few constraints on processing power and storage capabilities, they can generate and handle data at a rate impossible for traditional sensor networks. We can also exploit mobility to opportunistically diffuse concise summaries or metadata of sensor data. These can be harvested by other agents to construct a low-cost, distributed and scalable data index. Finally, vehicular sensors augmented with remote control are ideal for monitoring inhospitable environments such as unexplored battlefields or scenes of disasters. These applications all require persistent and reliable storage of data for later retrieval. Second, the vehicles can be significant consumers of content. Their local resources are capable of supporting high fidelity data retrieval and playback. Since drivers and passengers are stationary for the duration of each trip, they make up a captive audience for large quantities of data. Examples include locality-aware information (map based directions) and content for entertainment (streaming movies, music and ads). These applications require high throughput network connectivity and fast access to desired data.In a third class of compelling applications, vehicles are both the producers and consumers of content. Examples include services that report on road conditions and accidents, traffic congestion monitoring, and emergency neighbor alerts, e.g. my brakes are malfunctioning. We note that their direct relevance to road safety makes them a high priority for commercial entities. These applications require real-time and location-aware data gathering and dissemination. Finally, all of the above applications will need to rely on vehicles in an intermediary role. Individual vehicles in a mobile group setting can cooperate to improve the quality of the applicant experience for the entire network. Specifically, vehicles will provide temporary storage (caching) for others, as well as forwarding of both data and queries for data. In this capacity, they require reliable storage as well as efficient location of and routing to data sources and consumers.The demands of these applications give us a list of requirements and challenges for vehicular applications. Note that we can leverage them to simplify the applications infrastructure.

Time sensitivity: Time-sensitive data must be retrieved or disseminated to the desired location within a given time window. Failure to do so renders the data useless. This mirrors the needs of multimedia streaming across traditional networks, and we can leverage relevant research results from the related areas.

Location awareness: Both data gathered from vehicles and data consumed by vehicles are highly location-dependent. This property has direct implications on the design of data management and security components. Data caching and indexing should focus on location as a first order property; while data dissemination must be location-aware in order to maintain privacy and prevent tampering.

A number of significant research challenges remain: Time-sensitive dissemination of data to and from vehicles Efficient data indexing and query mechanisms using location and secondary characteristics Reliable and persistent network-based storage in the presence of node churn (movement in and out of local

mobile groups) and unreliable hardware Reliable location-based communication in the presence of high vehicular mobility, intermittent connectivity

and lossy channelsAs we focus on addressing these challenges, we will integrate a number of existing tools, including geo-routing protocols, network coding and realistic mobility models. We will examine the problem from a number of different perspectives, including those of security, interactions with the infrastructure, and overall impact on the Internet architecture.

3.1.3 Representative VANET applicationsThe vehicle grid applications pose new challenges on the ad hoc vehicle network. They can benefit from the design and development of new protocols. This section covers the main research directions in the applications and protocols area

3.1.3.1 Content downloading

Please revise and update, with the network layer impact in mind JS Park, UichinAs we discussed in the previous section, several emerging applications involve peer to peer content distribution. Content is distributed in the vehicle grid in different ways and for different purposes: (a) the car explicitly requests (pulls) segments of a “popular” multimedia file, from an access point or from neighbors (eg, car torrent); (b) the access point and the cars “push” location relevant content using epidemic dissemination and “data muling” (ad torrent) ; drivers in turns “opportunistically” request specific data items when they need them (eg, ad torrent); (c) multimedia data is streamed from cars on the scene of an accident (eg, collision slide, flood, fire, etc) back to the oncoming cars, as a warning and to allow them to take diverse routes. The above applications are very different in nature. However, they share a common model. The data files are downloaded from one or more sources to one or more receivers, using intermediate store and forward nodes. In Car-Torrent for example [wons 05], a car decides that it can cooperatively assemble the desired file using P2P content sharing. One option is for the car to query the neighborhood (say up to K hops deep) for the missing segments. The car selects the “best” peer for download. Simulation experiments[NDPG05] shows that the best strategy combines closeness and rarity of the “piece” (while Bit Torrent generally selects the “rarest” piece). This however involves quite a bit of O/H. There is first the selection of the peer (one query and several response); then, the TCP transfer of the multi-piece segment (typically, with several retransmissions). Another way to download pieces exploits network coding [GKANMR06]. The K-hop query goes out as before. However, there is no selection. Each peer with ”pieces” of the requested file delivers a random XOR combination of its current pieces (we assume each piece is exactly one packet long). Intermediate nodes also participate in the “random” mixing of pieces. A prefix in the packet tells the weights of the linear combination. If the car requests 5 missing pieces, then peers will deliver as many as 5 packets (if they have all those missing pieces). The receiver then can recover the pieces by solving a linear system of equations. The network coding scheme offers several advantages. It drastically reduces the number of messages and thus the overhead. One request is sufficient to collect multiple pieces, as opposed to one request per

piece. No TCP is required; just UDP. In fact, TCP will not work with multiple simultaneous sources. If the receiver does not have enough packets to solve the equations (some packets may have been lost), it simply requests more random combinations, as opposed to requesting specific packets. Another application that features downloading from multiple neighbors is “Ad Torrent” [ mobiquitous 05]. Say, a driver needs to find out which movies are playing in a particular neighborhood, along with videoclips. Also, the driver wants to dine at an Italian restaurant after the movie. Videoclips, menus, restaurant reviews and address must be acquired within latency constraints. Driving to the access point each time is too time consuming. Multihop downloading from a remote access point may not be practical and may create excessive load on the system. As an alternative, in Ad Torrent the access point feeds passing cars with randomly selected “ad segments”. Next, each car probabilistically disseminates the pieces using an epidemic ( “gossip”) scheme (we later review the effectviness of epidemic dissemination in the vehicular context). As a result the neighborhood becomes populated with ads. Again, there are different neighbor download strategies. One method [Nandan06] is to query the neighborhood and selectively download pages using Bloom Filters. Another approach is to query the neighbors and solicit network coding downloads of whatever useful pieces they have. The main difference from Car Torrent is that Ad Torrent epidemically disseminates segments (to increase the hit ratio of even not so popular files); also, it downloads from”3rd party” peers who are not trying to assemble the information themselves. This may have impact on tit-for-tat bookkeeping.There is also content distribution related to navigation safety. Suppose that a critical traffic/safety situation occurs on a higway, say, major traffic congestion, weather condition, fire, act of war or natural disaster. In such cases, multimedia content, say, video, could be streamed from one or more lead cars to the vehicles following several miles behind – to “visually” inform them of the problem and allow them to decide if they should turn around. Conventional ad hoc broadcast (eg, via ODMRP or MAODV) may introduce excessive loss and severely impair video reception. This is a situation where network coding can enhance stream reliability. To this end, we have recently developed an ad hoc Network Coded broadcast scheme called CodeCast that improves reliability through localized neighbor recovery and path diversity. The following graphs show preliminary results of CodeCast in a 100 node ad hoc network with random way-point mobility. CodeCast yields 100% delivery ratio as compared to 98% by ODMRP. Such accomplishment is achieved with less overhead. The end-to-end delay is increased, but yet within acceptable stream quality constraints. In the proposed project, we will extend CodeCast to vehicle networks.We will evaluate and compare traditional distribution schemes versus Network Coding under different car density, speed, file popularity and motion pattern. Performance measures are the delay to assemble the file and the total traffic O/H. Both analytic and simulation models will be developed [Nandan06].

CodeCast vs ODMRP: Packet Delivery Ratio

0.9

0.92

0.94

0.96

0.98

1

1.02

0 10 20 30 40 50

Max Node Speed (m/sec)

Pack

et D

eliv

ery

Rat

io

CodeCast

ODMRP

CodeCast vs ODMRP: Normalized Overhead

0

0.5

1

1.5

2

2.5

0 10 20 30 40 50

Max Node Speed (m/sec)

Nor

mal

ized

Ove

rhea

d

CodeCast

ODMRP

3.1.3. 2 Vehicular sensor platforms and Epidemic dissemination

Please revise and update, with the network layer impact in mind Uichin

Vehicular networks are emerging as a new network paradigm of primary relevance, for example for proactive urban monitoring using sensors and for sharing and disseminating data of common interest. Each vehicle can sense one or more events (e.g., imaging from streets and detecting toxic chemicals), process sensed data (e.g., recognizing license plates), and route messages to other vehicles (e.g., diffusing relevant notification to drivers or police agents). Opposed to traditional wireless sensor platforms, vehicles can generate such a large volume of data that can hardly be handled by traditional sensor network approaches (eg, periodic reporting to sinks). In this proposal, building upon previously proposed techniques for epidemic information dissemination in mobile ad hoc networks with pedestrian mobility [PS01], [LW04], we explore alternative lightweight strategies for proactive urban monitoring. The basic idea is to exploit vehicle mobility and limited broadcast to opportunistically diffuse concise summaries (meta data) of the sensor inputs or some general kind of lookup information for data stored in cars. Note that such lookup information is typically only valid within a short time lapse and, thus, may well require some mechanisms for coherency. As a first application scenario, similar to [Xu04] parking lot model, consider the very time consuming problem of finding a parking spot in a large urban area. Suppose cars are equipped with street maps; moreover, parking slots at curbside advertise when they are free via beacons. Then, the cars that are driving through downtown (and most likely looking for parking) can share this information in a sort of distributed directory. As a second scenario, consider that police agents harvest sensor inputs and lookup information, and build a low-cost, distributed, scalable index [PerSense]. This forensic data can be very valuable in many ways: from identification of communing patterns, to rush hour traffic behavior, to crime investigations. To support commercial applications, a vehicular network must provide basic services widely deployed in the Internet. Examples include communication utilities like email and instant messaging as well as services like web caching and content delivery. In a vehicular network, it is not clear which node (a wired node, a wireless proxy or a redundant set of nodes) should provide server-like functionality due to intermittent connectivity. In many cases, the only feasible approach for implementing a client-server application consists in the distribution of the server functionality among all participating nodes. Namely, Internet client-server applications require P2P cooperation when moved to the vehicular network. This leads to the third application scenario, instant messaging (IM) for a vehicular network. As a difference from previous services, this is a “delay sensitive“ application. “Presence” technology enables users of an IM system to determine if their contacts are online, signed onto the IM application, and ready to communicate. The protocol design for disseminating presence information in the Internet has been matured and organizations such as the IETF and the Jabber software foundation have developed protocol standards. However, due to the dynamic network topology and the lack of fixed infrastructure the dissemination of presence information in vehicular network poses a challenging research problem.In the area of vehicular dissemination several research issues will be addressed. First, how should epidemic information dissemination be tailored to application constraints (ie delay sensitive or delay-insensitive, coherency required or not, etc.), and to network constraints such as node density and mobility pattern (group mobility or individual mobility). Secondly, we will compare the performance of pure epidemic methods with hybrid methods based on epidemic dissemination and controlled flooding. Thirdly, we will investigate hierarchical schemes combining epidemically generated indexes with higher level structured indices (eg GHTs and DHTs). Finally, building up on [LW05] we will develop models that characterize the dynamics of gathered data with their coherency and real-time constraints.3.1.3.2 Urban mobility models

Please revise and update, with the network layer impact in mind Kelvin

The Random Waypoint (RWP) mobility model with pauses [Johnson96], [LeBoudec05] is widely used in the literature. In RWP, a mobile device starts at a random position drawn from a uniform distribution and moves to a destination position also drawn from by a uniform distribution. The device speed is chosen uniformly from (0, vmax]. When the mobile device reaches the destination position, it holds for an amount of time chosen uniformly from (0,Thold], before choosing a new destination position and continuing the process. Unfortunately, the RWP model in its original definition [Johnson96] did not posses a steady-state node distribution and led to a non-uniform

distribution of mobile devices while they are moving [Bettstetter03], [LeBoudec05]. Even the improved definition of [LeBoudec05] posses the latter shortcoming. The Obstacle Mobility model proposed in [Jardosh03] extends the RWP model through the incorporation of obstacles restricting node movement as well as wireless transmissions. The Obstacle Mobility model also bears the intrinsic shortcomings of the RWP model. In the Manhattan mobility model proposed in [Bai03], the mobile node is allowed to move along the horizontal or vertical streets. At an intersection of a horizontal and a vertical street, the mobile node can turn left, right or go straight with certain probabilities. All these mobility are inadequate to model the motion correlation among vehicles. Nodes in vehicle networks tend to move in “convoys” along freeways/local streets. In such networks there is some random movement, but there are also factors that tend to introduce correlation between individual trajectories, for example: group merge/split, obstacles, traffic accidents, traffic lights, etc. To cope with these issues, we introduced r eference point group mobility (RPG, [Hong99]): In RGP, the mobile devices move in G groups that cover each a circular area with radius rg. Groups move according to the random waypoint model with Thold = 0. Each mobile device is associated with a reference point uniformly chosen from the area covered by the group. The mobile devices are placed at positions that are randomly chosen from a circular area with radius rn around their reference point.To capture the most representative features of a urban motion, building upon RGP, we propose a “track” based group motion model. The track model is based on a continuous-time Markov Chain. The tracks are represented by freeways and local streets. The nodes must move following the tracks. At each intersection (switch station) a group can be split into multiple smaller groups; or may be merged into a bigger group. The track model allows also individually moving nodes as well as static nodes. Such non-grouped nodes are not restricted by switch stations and by real tracks. Instead their movements are modeled as random moves in the whole field. An important research issue constitutes the derivation of a proper mathematical definition of the track model, so that the corresponding continuous-time Markov chain possesses a steady-state (ie is ergodic).The proposed track model will be tested with real freeway/street maps from the US census bureau. The results will be verified with real urban traffic data from sigalert.com which monitors live traffic for big cities in California. The model will then be integrated in our simulator and will be used in the experiments to investigate the impact of motion on routing, TCP, index disseminations, etc. Preliminary results obtained with an early Track Model version have shown orders of magnitude higher delay in epidemic dissemination convergence than the RWP model!

3.1.3.4. Robust transport (probably omit)

Please revise and update, with the network layer impact in mind JiWei

Most of the exchanges in the vehicle grid (for navigation safety or content sharing/dissemination) are broadcast or many-to-one cast. Thus, they run on UDP. There are however occasional unicast file (data or multimedia) transfers. The issue then arises of TCP efficiency over a mobile, multihop path. Unfortunately, performance and fairness of TCP is poor over multihop wireless networks [FZL03]. This because wireless networks possess several properties, which are different to the wired Internet for which the widely deployed TCP NewReno implementation has been optimized. [XGQ03], [EKL05] proposed modifications to TCP for improving performance and fairness of TCP over wireless multihop networks without mobility.For vehicular networks route breakage due to the high mobility constitutes an additional major problem for TCP. Fortunately, Geo-routings are more robust to highly dynamic route changes than conventional MANET routing protocols considered in [HV99], [Xu04]. For best performance, however, several georouting parameters must be carefully tuned (eg, hello message exchange rate, delay timer in TCP for out-of-order delivery, etc) [CG06].

We plan to apply the lessons learned from [CG06], [XGQ03], [EKL05] to the vehicle grid environment. To improve hello efficiency in Georouting, we propose an adaptive hello exchange scheme based on node mobility. Then, we propose to fix the out-of-order problem by using a receiver-side out-of-order detection and by properly calibrating the parameters. Building upon [XGQ03], [EKL05], we will integrate an adaptive scheme into the TCP sender for

achieving almost optimal fairness among competing TCP flows. Subsequently, we will evaluate the impact of these adjustments to vehicle unicast applications for various degrees of mobility.

3.2 Security considerations (1.5 pages)Note: this section must be refocused to privacy and anonymity guarantees, and the assistance from Internet (Xiao Yan)Alex to overview the editing

Xiao Yan, if you can generate say, one page based on the bullets below, it would be great!

Roles of Internet in Vehicular Network Security (Xiaoyan)

New roles in security and privacyProvide security infrastructure support OpportunisticPrivacy-preserving Provide VANET security surveillance Provide additional VANET anonymity and privacy

I will need comments here for the following reasons:(a) addressed privacy issues in the role 1,(b) may contradict to geo addressing and routing (c) if not (b), we achieve location privacy with sensitivity and performance tradeoff. (see the last slide)

Task 1

To address: new roles to provide security infrastructure support Opportunistic Privacy-preserving Proposed research: Loosely coupled authenticationDefend DDoS (false/stale data injection in content distribution applications)Enable security and privacy in network topology constructionDual-mode authentication architecture to support privacy-preserving authentication

task 2

To address: new roles in providing VANET security surveillanceProposed researchMobile DDoS attacker tracingScenarios: attackers surface at different APs to launch attack. Extend traditional traffic analysis techniques to the vanet

Task 3

To address: new roles to provide additional VANET anonymity and privacyProposed research:Builds on top of multi-resolution distributed location-service, perform fuzzy geo forwarding. Note: the location-service can be provided jointly by Internet and VANET. Note, the privacy is achieved with tradeoff of sensitivity (how fuzzy) and performance (delivery ratio)

General vehicular network security is supported by security primitives like key distribution and message authentication through Public Key Infrastructure [ZMTV02] [MBG05] [ABDF][HCL04] [D05]. Drivers can obtain and update (say, annually) their pubic key certifications from registration authorities like DMV[PP05][RH05][SHLP05]. Messages propagated within the network must be authenticated to prevent external attackers from injecting, altering and replaying old messages (messages in vehicular grid are stamped with location and time). One outstanding security problem is the location verification and bogus data detection[CH05][GGS04]. However, these general security measures do not solve many security issues regarding to the new applications foreseen in the proposal. Our research tasks are to identify the security requirements for various applications, to investigate the new security issues and to deal with the limitations of the vehicular networks in supporting security. Two very critical issues we discuss here are the security of network coding and the security of content distribution. Content distribution serves a variety of goals, ranging from traffic monitoring, hazard detection, to forensic investigation. One critical security problem is to prevent, identify and reject false data in the distribution process. More specifically we study the security of location and timestamp data that are very important parts of any record. Bogus location information and/or wrong time stamps can cause huge disruption to queries and to decision making over data collected from mobile sensor platforms. Sybil attack that counterfeits multiple identities [D02] to weaken defenses built based on majority rule is mostly impossible since a malicious node cannot counterfeit a valid certificate without the CA’s private key. Related secure data dissemination problem has been studied in sensor networks. Solutions include preventing through secure aggregation [PSP03], detecting and filtering when packets are en route[YLLZ04] [ZSJN04][V05]. All these schemes are based on static distributed symmetric secure key sharing, which is not applicable in the highly mobile vehicular networks. In vehicular networks, due to the lack of pair wise trustiness, secure locations and timestamps have to rely on a collection of witnesses and use majority rule. To elude such a security testing system, a group of malicious nodes must collude and generate witnesses for each other. In a dynamic vehicular network, excluding all good nodes from the witness database will be very hard to sustain over a period of time. Thus such a validation system will be very

effective. Specifically, location can be verified through multilateration among base stations given the distance bounding information [CH05]. However, the scheme relies on densely deployed roadside base stations. An alternative way is to construct a graph with all the observations and mark the links according to assertions calculated from the observations [GGS04]. A failed assertion could generate a label of “spoof” or “malicious”. This approach applies to timestamps as well with a certain threshold to tolerate drifts among different clocks. The success of the approach relies on rich inputs of observations. A third method is actually combining reputation methods [MGLB00] with the above approach. Using matrix operation as presented for p2p network ranking incentive algorithm Tit-for-Tat [LPYZ06], we will be able to evaluate both the data and the nodes. After all, these schemes have to be integrated with epidemic dissemination scheme or CodeCast scheme. Evaluations must pay attention to impact from possible insufficient observation inputs since the network is dynamic and highly mobile. We will also exploit locality in motion to improve the success rate of detection and filtering.

Also, a section on incentives to be produced by JS Park, using CS 218 project results as basis

3.3 Network and Internet services (3.5 pages)

3.3.1 Address conventions + GLS (1.5)

Mario to edit this section, using inputs from Claudio, with advise from Marco. A major challenge in the management of vehicular network mobility and interconnection to and through the Internet is addressing. Let us begin by defining: (a) Unique car name - Several options are possible here: license plate#; Vehicle-ID#; owner’s name. IP address can occasionally be used an stored as temporary “unique” ID (b) Routable car address: geo-coordinates; specific attribute (as in “attribute based” routing - eg, car torrent membership); unique ID (typically IP address) for some type of routing , eg AODV As earlier discussed, the most prominent routing in the vehicle grid is geo-routing. AODV may also be used over short paths (few hops). (Note: AODV currently uses IP addresses to set up/maintain on demand routes). Thus, geo-address is the dominant routable address in the vehicle grid. Internet type IP routing (ie, prefix routing etc) is meaningless in an ad hoc network such as the vehicle grid. Yet, the IP address will still be useful as identifier in AODV routing and in TCP connections. For this reason, we propose to maintain in the vehicle grid a “unique” IP address for cars. The car IP address can be initialized for example (after a long inactive period) by hashing the Vehic #, owner # and license #. Clearly, the result may not be unique. If address conflicts ever happen (during TCP connection set up, or AODV routing), the tie is broken by re-hashing the IP address (ie, the IP address of a car may change during lifetime). To avoid collisions in a proactive way, IP uniqueness within a local scope (say, 3-4 hops) may be constantly verified and enforced by an elected IP master node. We assume there will no DHCP for cars when they pass by Access Points. A DHCP address would be useless given the high car mobility.We envision that the addressing scheme should support the following services:

(a) a car must be able to efficiently address any other car in the urban grid; (b) an Internet server must be able to address any vehicles in the grid

One can easily show that geo-routing fits these requirements. More precisely, geo-routing takes a packet “in the neighborhood” of the target destination. Once the packet is within radio reach, any unique node identifies (eg IP address, or license number) can be used to deliver the packet to the node. A critical component of the geo-routing address structure is the Geo Location Service (GLS) - a distributed service that maps any car name to the set of most recent geo locations. We propose to implement two “parallel” version of GLS, namely: OLS (Overlay Location Service) and VLS (Vehicle Grid Location Service). OLS is maintained within the Internet infrastructure; VLS is maintained entirely in the vehicle grid. The two services are synchronized, but are independently maintained to provide fault tolerance (one of the tenets of our proposed architecture)Let us illustrate a possible OLS implementation. An overlay structure is established in the urban Internet. Each car, whenever it passes by an AP, registers its ID (license#, IP address(es), time, owner name, owner IP address billing address, etc) and the current geo-location. OLS maintains an index of IDs. Each ID is mapped to the geo

coordinates. The index is distributed. It may be managed via DHT (Distributed Hash Tables). Suppose now Host A (fixed or mobile) wants to establish a TCP connection to mobile B. Host A first queries OLS with: [email protected] starting from the nearest server of the OLS “overlay”; it gets back the “most recent” geo-locations, the IP address, etc of Car B. It predicts from recent locations the best current estimate for access point AP. Host A sends the msg to the AP closest to the geo-destination (the overlay can perform the mapping); the message is encapsulated in an IPv6 network envelope that contains the geo address in the extended header. Routing in the Overlay is based on geo address. Namely, the geo address determines the AP at the end of the Internet path. At destination, the AP geo-routes the packet to the ad hoc net; the car responds with own IP and directs the response (encapsulated in the overlay envelope) to the sender IP. Note that the encapsulation into a geo routed network envelope is identical regardless whether the sending Host A is fixed or mobile. In principle, one geo location server, say OLS, would be sufficient. However, since the urban Internet infrastructure (or wireless access to it) may fail, we maintain also VLS in the Urban Grid. Considerable amount of research has been done on VLS design. One must minimize registration overhead with minimizing at the same time index search – two conflicting requirements. We propose to explore a new solution, based, like the others [GLS ] on a hierarchy. At the lowest level, there is a unique (mobile) server – eg, a CalTran truck- that roams in a cell (1 km x 1 km), periodically advertising its coordinates. Cars register locally with the truck. Periodic advertising precludes the well known dead-ends of conventional geo routing. [Directional forwarding; Landmark assisted Geo Routing]. At the higher level of the hierarchy, there may be as many as 1000 cells in a large metropolitan area (say 33 km x 33 km). A car will geo-hash its license # in one of these cells (the permanent home cell, say). As the car moves from one cell to another, it must update the pointer in its home cell. This is quite a bit of overhead, due to the inefficiency of geo routing. During normal operations, when the infrastructure is up, home cell updating can be done through the Internet overlay, at the same time when the OLS updating takes place. Least cost routing (via vehicle grid or Internet) will also be explored.In this project, we will explore the effectiveness of coordinated VLS and OLS structures. We will investigate various hierarchic solutions for scalability. We will also implement OLS in the GRIDO testbed.

3.3.2Routing in the vehicle grid (1.5 pages)

Marco, can you please review the geo routing considerations?Kelvin, please look at this as wellMost of vehicle grid applications we described so far require exchange of unicast or broadcast messages between neighbors. For this type of proximity routing, on demand schemes such are AODV, ODMRP or even flooding are adequate [aodv], [odmrp]. Some applications require unicast routing to destinations several hops away. If geo-coordinates of the destination vehicle are obtained from a Geo Location Service, the “routable address” is the geo-address and the packet is routed using geo-routing (eg, greedy forwarding). If instead the destination (eg, a web server) is found using scoped flood search a la AODV, say, then the route to the server can be supported by routing table entries, like in AODV (geo routing may also be used once the server coordinates are learned). As another option, the destination might periodically advertise its presence (ie, proactive routing); then proactive table driven forwarding is used.

From the above we note that the vehicle grid must support many routing options, the selection depending on the name/address map scheme. The prominent scheme, especially to remote destinations, will be geo-routing. The georouting implementation in vehicular grids still poses research challenges. The first issue is vulnerability to “dead end” traps. Vehicle grids are full of such traps. Once GPSR falls in a trap, the recovery must be done with time consuming “perimeter routing” schemes. One research issue on our agenda is to investigate schemes that prevent/recover form traps. To this end, landmark assisted geo routing (GeoLanmar) can be used to warn nodes when the direction “as the crow flies” leads to a trap. This is done by comparing the “Euclidian” direction with the Landmark advertised direction. The advertised routes, however, may become stale if the refresh period is slow (to keep O/H low). To achieve more durable routes, we recently proposed to use not the advertised next node (on the route) but the advertised direction, ie “Direction Forwarding” [ad hoc net journal]. In this project, we plan to

examine the tradeoffs between increased route robustness and additional control traffic overhead for Landmark Assisted Geo-routing and for Direction Forwarding, in typical urban grid topologies, motion behavior and traffic patterns.

3.3.3 IP address auto-configuration (1.5 pages)

As already mentioned, any networking session (eg TCP) and application requires unique identifiers for peer communications nodes (eg Vehicle ID No). Needless to say, this remains true even in the considered scenario of urban vehicular grids. Historically, in the Internet, the IP address scheme was designed to represent both a unique ID and a routable address. However, at that time the Internet and its hosts where far from being mobile. Nowadays, instead, communication capabilities can be easily found on lightweight mobile devices and, very soon, even on cars. The very high and fast mobility that characterizes this emerging scenario changes the initial design assumption at the basis of the IP address. Indeed, major problems with maintaining sessions arise when routable address changes - ie during handoff. To this aim, specific solutions have been proposed such as Mobile IP, IPv6, tunneling, etc. The importance of the IP address with the new mobile scenario as not been weakened, as demonstrated also by its utilization in MANETs as “unique” ID (for TCP, UDP, and at times even for routing, eg, AODV). On the other hand, some tasks that have been successfully implemented in traditional networking systems require now the development of new solutions aimed at implementing them even in vehicular or, more in general, mobile scenarios. A prominent example of these tasks is represented by the IP address auto-configuration of nodes that leads to unique IP addresses and that has to be performed even when employing IPv6 in place of the traditional IPv4 [RFC2462]. Auto-configuration of IP addresses, such that the assignment results unique, requires specific investigation for the vehicle grid scenario. Indeed, the direct employment of solutions developed for regular ad hoc networks cannot be directly applied to a vehicular scenario due to the peculiar characteristics of the latter: high density of nodes (many cars in few meters on a highway or in town); high absolute speed (20-80mph) but low relative speed with respect to other cars traveling in the same direction (3-20 mph); and practically “infinite” network diameter (millions of cars could be present in a large metropolis). The problem we need to address is hence that of creating an auto-configuration service able to efficiently support vehicular networks guaranteeing the following properties:

high reliability (ie, low ID collision rate) of the address configuration; low signal overhead generated by the system; low configuration time (especially important traffic safety and real-time applications).

To this aim, typical decentralized approaches requires that all nodes in the network are involved in the address configuration task by maintaining and exchanging a list of addresses which are currently in use or that are going to be assigned to new nodes. [MP02], [NP02]. Obviously, distributing the information to all the nodes in the network represents a solution that does not scale as it would generate a very high volume of control traffic to keep the information updated and consistent if employed in large networks.As an alternative, best effort approaches provide correct routing with a limited control traffic, yet, without ensuring unique node addresses and thus generating serious delays when address duplications have to be solved among sessions (say at TCP level, for instance). Probabilistic algorithms such as [Weniger2005] belong to this class. Their aim is that of minimizing the address duplication probability and performing DAD (Duplicate Address Detection) procedure when two nodes with duplicate addresses start to communicate.A better solution is represented by leader based approaches. Solutions belonging to this hybrid class generally implement a hierarchical structure to configure nodes and perform the DAD procedure only within a cluster [Sun04], [Toner03]. Then, when leaders pass by the coverage area of an AP it could transfer to it information about the addresses utilized by vehicles within its control range. In this way, georouting could be exploited to deliver message to the right AP (when communications with the Internet are involved), whereas ongoing sessions (eg, TCP) between cars or between a car and the Internet will be feasible thanks to the presence of the unique IP address.We deem, and are aimed at demonstrating, that this kind of leader based approach represents an efficient scheme for IP address auto-configuration. In particular, we envision a scheme that will work with leaders proactively organized in a chain and work like DHCP servers to dispense (and manage) unique IP addresses to vehicles within the range

(scope) of that leader chain. Since relative speeds among vehicles are limited if compared to absolute ones, having leader vehicles managing IP addresses within the vehicle grid would ensure a longer longevity to those IP addresses with respect to utilizing traditional DHCP servers on fixed APs placed along the roadside. Moreover, our scheme would avoid that routing, TCP and thus ad-hoc networking services may fail if an area is not covered by at least one AP. We have already proven that this scheme is effective in a highway scenario [FPSG06a], [FPSG06b], [FPSG07]; we need now to adapted it to operate even in an urban context.Main tasks of the proposed approach that need to be addressed in an efficient way are:

The construction and maintenance of the leader chain . Due to different mobility patterns among vehicles, leaders may join/leave the chain or get too close/far to each other to appropriately cover all the vehicles within the scope. Therefore, fast reconfiguration of the leader chain (and of the regular vehicles relying on it) has to be provided.

The configuration of nodes’ addresses . This represents the core task of the scheme and is composed of a module for assigning/managing addresses and another one for performing the DAD procedure. Focusing on the former, the address space is partitioned in sets of addresses and each leader in a scope manages a different set. The synchronization of address information among leaders is performed by exploiting hello packets and addresses within each set are assigned to vehicles through a modified DHCP protocol. Instead, the DAD procedure is in charge of verifying whether an address among those in use within the considered scope has ceased to be unique due to vehicles’ mobility and restore the uniqueness of all addresses. As vehicles in a scope are proactively organized to be all covered by the leader chain, the DAD procedure is performed requiring only single-hop communications (between regular vehicles and leader ones).

The scheme will be evaluated to understand its effectiveness in a vehicular grid scenario in order to support all the possible applications that will be run in such a context. To this aim various metrics will be considered, eg, the configuration time of each node, the introduced control overhead, the longevity of an address assignment. The investigation could be run either via simulations (QualNet, NS2), or real testbed experiments (see Section “TESTBEDS”), or both. Achieved performances will be assessed and compared by considering different configurations of the scenario (eg, vehicle density, speeds, scope width).

3.3.4. Traffic & capacity management tools (.5)

Mario will work on thisTraffic and capacity measurement are fundamental for computer network management. In C2C scenarios, new challenges are posed by the management of mobile users and shared wireless communication channels. First, in C2C networks, the capacity of a path from Internet server to mobile can vary dynamically and rapidly due to changes in wireless hop count, interference and mobility; so, timely path capacity tracking is the key to efficient routing, traffic management and multimedia stream rate and format adaptation. We have developed a packet-pair based technique called AdHoc Probe that estimates end-to-end path capacity on a mixed Internet and ad hoc path [Chen:05]. It converges fast, thus proving adequate for mobile, rapidly changing scenarios such as vehicular grids. We propose to evaluate AdHoc Probe to support representative vehicular network applications (eg, seamless handoff from one media to another).Residual Capacity or equivalently, the load on a path is also of great interest. The challenge stems from the fluctuation of the load on a path. Residual capacity is less stable and much more difficult to estimate than path capacity. Existing methods such as pathload [Jain:03] rely on increasing a probing rate until a link is saturated. They requires relatively long time to converge. Besides, they only work for the Internet, while, in vehicle networks the wireless ad hoc portion of a path is more likely to become the bottleneck for residual capacity. This calls for research in “fast” residual capacity estimation in multi-hop wireless paths. Related to path characteristics estimation is motion and location prediction. Even a coarse mobility prediction could help estimate future load patterns and capacity changes and apply such knowledge to anticipatory routing and congestion control actions.The relationship of buffer size at the forwarding nodes to the bandwidth delay product of the path, has a significant impact on TCP performance such as delay and packet loss, and particularly impact how different TCP and other protocols share a path. Buffer size estimation is especially needed for C2C networks, since there can be multiple

bottlenecks on both Internet portion and on C2C portion of a path. Buffer size estimation also enables a more accurate estimation of congestion, which was put to good use in TCP Westwood BBE [Shimonishi:05]Path persistence is an important cross traffic characteristic. The cross traffic is called path persistent if it shares much of the same end-to-end path with a flow of interest. When the cross traffic exist only on a small portion of the path, it is called path non-persistent. Path persistence affects the accuracy of residual capacity estimation tools (e.g., Spruce [Strauss:03]) as well as fairness among TCP flows. We propose to investigate cross traffic path persistence behavior and its efficient estimation in the hybrid wire/wireless context. Responsiveness refers to the ability of a flow to adjust its rate based adapting to network condition. TCP traffic is responsive in that it will reduce its rate in the presence of congestion, while UDP traffic is non-responsive. Knowing traffic responsiveness in vehicular networks will assist when downloading continuous media from a server. The path can be chosen based on residual capacity and cross traffic responsiveness.

3.4 TESTBEDS – testing the applications and the Internet services (3 pages)

Marco, can you please add ORBIT and Rutgers vehicle testbed contribution to this research?

We plan to study the interaction of vehicular applications with the Internet through the Campus Vehicular Testbed at UCLA (C-VeT) and the GRIDO overlay implementation [DNPP05][CVET]. The C-VeT testbed has been designed to study the behavior of network protocols and applications for various urban scenarios and mobility patterns. In particular, C-VeT exploits the mobility of the UCLA facility management vehicles and the regional mobility of the UCLA VAN commuter fleet to build a networked facility that can be shared by other users and can provide realistic motion benchmarks. Each C-VeT node consists of a vehicle equipped with a compact size industrial PC, 2 IEEE 802.11n interfaces, a GPS, and a ham radio interface (to be used as control channel). In the future we will also introduce more advanced radios as they are made available from other programs such as the MIMO radios designed in the DARPA MNM program at UCLA [REF] and the cognitive radios produced by the current NSF FIND program at Kansas University [REF]. At steady state, C-VeT will include 30 facility management cars and 30 regional VANs thus providing 60 mobile nodes. The facility management cars run mainly in campus. Their mobility is driven by the daily maintenance schedule and by the on-demand department and campus logistic requests. The UCLA regional VAN-RIDE provides a door-to-door shared ride for UCLA students and employees covering the whole greater Los Angeles area. The Van mobility is driven by the needs of its users and can change every day according to the user requests. The C-VeT vehicular testbed has been designed to be shared through the Internet in a “Planetlab” fashion. Users will be able to book testbed resources in advance or share them through virtualization. Each node, indeed, is equipped with the XEN virtual machine technology developed at University of Cambridge, UK thus offering the users an insulated virtual sandbox for their experiments [XEN]. A specifically designed web interface will allow researchers to interact with the whole or part of the network. We will also develop and expose a set of API suitable to interconnect the C-VeT testbed to other wireless testbeds such the Orbit testbed developed at Rutgers University or, in the future, to the GENI infrastructure [ORBIT]. In its initial phase the C-VeT vehicular testbed will be interconnected to the campus wide UCLA Wireless LAN infrastructure thus allowing the vehicles to access the Internet directly or through an optimized overlay such as the GRIDO testbed developed at UCLA [CDKL04].

The C-VeT Snbapshot of the Control Center showing an area of the campus experiments. The links represent the OLSR routing table.

The UCLA Van fleet

In addition to enabling protocol prototyping and performance evaluation in a realistic enviroment the C-VeT fleets will allow the collection of a large amount of data about the vehicular traffic in the Los Angeles area thus enabling the creation of realistic mobility models that can be compared with those generated by microscopic traffic simulators such CorSim or TransitSim. The C-VeT testbed is connected to a control center. The control center reports statistics on each vehicle such position, routing information, network load information etc. Additionally it manages and provides the access to the testbed and to the wide area connectivity through GRIDO or the Internet.

Node Equipment: an industrial PC, a GPS unit, and a MIMO IEEE802.11n card by Linksys.

The UCLA facility management carts.

The GRIDO testbed has been designed to provide the backbone overlay for the WHYNET project. It is suited to serve as the Vehicle Overlay to support the Vehicle Grid. It can be used to store the DHT based Geo Location Server. GRIDO tools can also be leveraged to provide a bandwidth-latency optimized overlay for data streams of varying kinds: multimedia, file transfer, content replication etc. GRIDO features a WS-Agreement based negotiation interface complying with the current Global Grid Forum (GGF) standards. In GRIDO we use a virtual-coordinates-assisted overlay construction and maintenance protocol to construct and maintain an optimal backbone structure [CDKL04].

The GRIDO/C-VET compound will enable realistic in field studies of a new generation of mobile applications and protocols including scalable location services and wireless content distribution. For example, a provider of wireless connectivity to various freeways in the US wants to set up a certain number of popular feeds to some freeway segments in Florida, in New Jersey and in California, respectively. GRIDO automatically geo-locates the segment addresses and sets up the content delivery points near those locations. In concert with car-to-car optimized data-delivery schemes like CarTorrent [NDPS05] and AdTorrent [NDZP05], GRIDO distributes pervasive content and applications close to where it will be used, reducing the load on servers and increasing the response time as perceived by the users.

GRIDO overlay on Planetlab Grido and CarTorrent at work

We plan to study the interaction of vehicular applications with the Internet through C-VeT and GRIDO [DNPP05]. GRIDO has been developed on top of the “Click Modular Router Project” thus allowing fast implementation, flexibility and reconfigurability [KMCJ00]. One plan currently coordinated via WHYNET is to use the GRIDO infrastructure to interconnect vehicular testbed islands that are emerging on the WHYNET project and beyond (see the schematic overlay in the figure above). We already have collaboration plans with the ORBIT testbed at the Rutgers University and with international facilities in Nijgata Japan and Leipzig, Germany. This vision will lead to a unique geographically distributed vehicular testbed that will include a wide variety of vehicle grid technologies, architectures, and mobility models. It will be and ideal observatory for heterogeneous vehicle technology interconnection.

[ORBIT] http://www.orbit-lab.org/[CVET] http://www.C-VeT.org/[XEN] http://www.cl.cam.ac.uk/research/srg/netos/xen/

4 Plan of Work (.5)More work needed after the research section is finalized (Mario)Year 1: extend target applications (Car Torrent, Ad Torrent, etc) with Network Coding; begin development of the Campus testbed C-VET; extend current simulator and emulator testbeds (Qualnet, PeerKit ); develop realistic mobility models; begin GRIDO overlay implementation; protocol/applications testing via simulation; verification schemes for content distribution, performance evaluation through simulation. Year 2: develop, simulate and test epidemic dissemination protocols (both flat and hierarchical); test robust georouting and robust TCP protocols in simulation. Implementation of network coding security scheme in testbed.Year 3: Implementation of GRIDO overlay and of OLS and VLS geo-servers in the testbed; demonstrate transparent geo-addressing in the infrastructure; demostrate simple C2C applications (Car Torrent; messaging) with

real vehicles on the UCLA campus; carry out vehicle grid interconnection experiments with WHYNET partners and with our international collaborators in Japan and Germany.

2 Broader Impact and Educational Activities (.5)More work needed (Mario)

Integration into Educational Activities: All of the member institutions offer regular graduate courses on Advanced Networks and Mobile and Wireless Networking. The artifacts produced by this research will be integrated into class projects, and encourage further research into vehicular communication and security issues. Simulations and test-bed experiments will offer hands-on experience in protocol engineering and network systems. This project will foster the collaboration and student exchange between international research institutions in Germany and Japan.Support for under-represented groups and states: The University of Alabama (UoA) is an EPSCoR state with 75% of the undergraduate students coming from the state of Alabama. Among them, 13% are African-American, 1% Hispanic-American, 1% Asian-American and 53% women students. The project will provide opportunities for both senior undergraduate and graduate students to participate in research activities including simulation and testbed experiments during summers. Separate NSF REU grants are in our plan for additional support for exchange students in the summers. Through this opportunity, we will be able to encourage students from Alabama to engage in science and technology. Broader Impact: We anticipate the proposed work to advance the state of the art in several research areas, including vehicular networking, peer-to-peer networks, and wireless security. The application infrastructure developed can encourage rapid prototyping of novel and innovative applications. Our strong ties to industrial partners can lead to adoption and distillation of these infrastructure designs into real software products. There will be also significant society impacts. Car manufacturers will exploit our findings to develop safer cars and more productive mobile-offices, with enormous impact on social welfare. Transport Authorities will be able to manage vehicle traffic more efficiently with reduced transfer delays and again substantial benefits to society. Police Departments and Civilian Defense Agencies will have at their disposal new extremely powerful forensic investigation, prediction and terrorist attack prevention toolsThe intellectual merits of the project are: the identification, implementation and evaluation of critical Internet services required by the emerging vehicular environment, and; the definition of guidelines for the design of new applications that can smoothly interface with the Internet. The study will characterize the vehicle performance requirements and will generate benchmarks for evaluating and comparing new Internet services. Original contributions will be made in the areas of Network Coding, Geo-routing, robust TCP and Epidemic Dissemination, both in terms of performance models and implementations. Within the C-VET testbed, the synergy of vehicle grid with the Internet infrastructure will be enabled by a robust overlay and redundant Geo Location Servers.

3 Budget (Giovanni)

4 Current and pending support (Giovanni)

5 References

(Alex and Sewook in charge)Rule: [First Author Initial, Second Auth Init, Third Init, year, year, version (A toZ)] ie a Max of 6 characters.[ACL00] R. Ahlswede, N. Cai, S. Li, R. Yeung, Network Information Flow, IEEE Transactions on Information Theory, Vol. 46(4), 2000.[Ai05] A. Aijaz, B. Bochow, F. Dötzer, A. Festag, M. Gerlach, R. Kroh and T. Leinmüller. “Attacks on Inter-Vehicle Communication Systems - An Analysis,” Network on Wheels Project, http://www.network-on-wheels.de/documents.html, 2005.[Ba03] F. Bai, N. Sadagopan, A. Helmy. “IMPORTANT: A framework to systematically analyze the Impact of Mobility on Performance of RouTing protocols for Adhoc NeTworks,” In IEEE INFOCOM’03, San Francisco, CA, April 2003.[Bettstetter02] C. Bettstetter. “On the Minimum Node Degree and Connectivity of a Wireless Multihop Network,” In ACM MOBIHOC’02, Lausanne, Switzerland, 2002. [Bettstetter03] C. Bettstetter, G Resta, P Santi. The Node Distribution of the Random Waypoint Mobility Model for Wireless Ad Hoc Networks, IEEE Trans. on Mobile Computing, 2003. [BettstetterRS03] C. Bettstetter, G Resta, P Santi. “The Node Distribution of the Random Waypoint Mobility Model for Wireless Ad Hoc Networks.” IEEE Trans. on Mobile Computing, vol. 02, no. 3, pp. 257-269, Jul-Sept, 2003.[BettstetterW02] C. Bettstetter and C. Wagner. “The Spatial Node Distribution of the Random Waypoint Mobility Model.” In German Workshop on Mobile Ad-Hoc Networks (WMAN), Ulm, Germany, GI Lecture Notes in Informatics, no. P-11, pp. 41-58, Mar 25-26, 2002.[Biswas05] S. Biswas, R. Morris. “ExOR: Opportunistic Multi-Hop Routing for Wireless Networks,” In ACM SIGCOMM’05, Phila, PA, August 2005[Blough02] D. M. Blough, G. Resta and P. Santi. “A statistical analysis of the long-run node spatial distribution in mobile ad hoc networks.” In ACM International Workshop on Modeling, Analysis and Simulation of Wireless and Mobile Systems(MSWiM), Atlanta, GA, Sep. 2002.[Blum03] B. Blum and T. He and S. Son and J. Stankovic. “IGF: A State-Free Robust Communication Protocol for Wireless Sensor Networks,” Tech Report, University of Virginia, 2003.[CapkunH05] S. Capkun and J.-P. Hubaux. “Secure Positioning of Wireless Devices with Application to Sensor Networks.” In IEEE INFOCOM’05, Miami, FL, Mar., 2005.[Chen05] L.-J. Chen, T. Sun, G. Yang, M. Y. Sanadidi, and M. Gerla. “AdHoc Probe: Path Capacity Probing in Wireless Ad Hoc Networks.” In IEEE International Conference on Wireless Internet (WICON’05), Budapest, Hungary, 2005.[ChenDV05] D. Chen, J. Deng and P. K. Varshney, “A State-Free Data Delivery Protocol for Multihop Wireless Sensor Networks,” In WCNC’05, New Orleans, LA,March, 2005.[Cox04] R. Cox, F. Dabek, F. Kaashoek, J. Li, R. Morris. “Practical, Distributed Network Coordinates.” In Workshop on Hot Topics in Networks (Hotnets’03), Cambridge, MA, November 2003.[Das05] S. Das, A. Nandan, M. G. Parker, G. Pau and M. Gerla, "Grido- An Architecture for a Grid-based Overlay Network." In International Conference on Quality of Service in Heterogeneous Wired/Wireless Networks (QShine 2005), Orlando, FL, USA, August 22 - 24, 2005.

[Dikaiakos05] M. Dikaiakos, S. Iqbal, T. Nadeem, and L. Iftode “VITP: An Information Transfer Protocol Vehicular Computing,” In ACM workshop on Vehicular ad hoc networks (VANET’05), Cologne, Germany, August, 2005. [Dotzer05] F. Dötzer, “Privacy Issues in Vehicular Ad Hoc Networks,” In Workshop on Privacy Enhancing Technologies (PET 2005) , Croatia , May-June 2005.[Douceur02] J. R. Douceur. “The Sybil Attack,” In International Workshop on Peer-to-Peer Systems (IPTPS '02), Cambridge, MA, USA, Mar. 2002.[Dovrolis04] C. Dovrolis, P. Ramanathan, and D. Moore. “Packet Dispersion Techniques and Capacity Estimation,” IEEE/ACM Transactions on Networking, Vol. 12, No. 6, December 2004, 963-977. [DSRC] 5.9GHZ DSRC http://grouper.ieee.org/groups/scc32/dsrc[ElRakabawy05] S. ElRakabawy, A. Klemm, and C. Lindemann. “TCP with Adaptive Pacing for Multihop Wireless Networks,” In ACM MobiHoc’05, Urbana-Champaign, IL, USA, May 2005.[EKL05] S. ElRakabawy, A. Klemm, and C. Lindemann, TCP with Adaptive Pacing for Multihop Wireless Networks, Proc. ACM MobiHoc 2005, Urbana-Champaign, IL, USA, May 2005.[FPDG06a] M. Fazio, C. E. Palazzi, S. Das, and M. Gerla. “Automatic IP Address Configuration in VANETs”, in Proc. of the Third ACM International Workshop on Vehicular Ad Hoc Networks (VANET 2006), MOBICOM 2006, Marina del Rey, Los Angeles, CA, USA, September 2006.[FPDG06b] M. Fazio, C. E. Palazzi, S. Das, and M. Gerla. “Vehicular Address Configuration”, in Proc. of the 1st IEEE Workshop on Automotive Networking and Applications (AutoNet), GLOBECOM 2006, San Francisco, CA, USA, December 2006.[FPDG07] M. Fazio, C. E. Palazzi, S. Das, and M. Gerla. “Facilitating Real-time Applications in VANETs through Fast Address Auto-configuration”, in Proc. of the 3rd IEEE CCNC International Workshop on Networking Issues in Multimedia Entertainment (CCNC/NIME 2007), Las Vegas, NV, USA, IEEE Communications Society, January 2007.[Fu03] Z. Fu, P. Zerfos, H. Luo, S. Lu, L. Zhang, and M. Gerla. “The Impact of Multihop Wireless Channel on TCP Throughput and Loss,” In IEEE INFOCOM 03, San Francisco CA, 2003.[Füßler03] H. Füßler, J. Widmer, M. Käsemann, M. Mauve, H. Hartenstein. “Contention-Based Forwarding for Mobile Ad-Hoc Networks.” Elsevier's Ad Hoc Networks 1 (4), pp. 351-369, November 2003.[FZL03] Z. Fu, P. Zerfos, H. Luo, S. Lu, L. Zhang, and M. Gerla, The Impact of Multihop Wireless Channel on TCP Throughput and Loss, Proc. IEEE INFOCOM 03, San Francisco CA, 2003.[Gkan05] C. Gkantsidis, P. Rodriguez. "Network Coding for Large Scale Content Distribution", In IEEE INFOCOM, Miami. March 2005.[Gkan06] C. Gkantsidis, P. Rodriguez, "Cooperative Security for Network Coding File Distribution", In IEEE INFOCOM’06, Barcelona, April 2006.[GkanMR06] C. Gkantsidis, J. Miller, P. Rodriguez. “Anatomy of a P2P Content Distribution system with Network Coding,” In International Workshop on Peer-to-Peer Systems (IPTPS '06), Santa Barbara, CA, Feb 2006.[GolleGS04] P. Golle, D. Greene, and J. Staddon. “Detecting and correcting malicious data in VANETs.” In ACM workshop on Vehicular ad hoc networks (VANET’04), Philadelphia, PA, USA, October 2004. [Groenevelt05] R. Groenevelt, P. Nain, and G. Koole, “The Message Delay in Mobile Ad Hoc Networks,” In Performance 2005, Juan-les-Pins, France, 2005.[Guerin87] R. A. Guerin. “Channel Occupancy Time Distribution in a Cellular Radio System.” IEEE Transactions on Vehicular Technology, 36(3): 89–99, 1987.[Ho03] T. Ho, R. Koetter, M. Médard, D. R. Karger and M. Effros, The Benefits of Coding over Routing in a Randomized Setting. In Proc. of IEEE International Symposium on Information Theory, 2003.[HV99] H. Holland and N. Vaidya, Analysis of TCP Performance over Mobile Ad Hoc Networks, Proc. ACM MOBICOM 99, Seattle, WA, 1999.[HongGP99] X. Hong, M. Gerla, G. Pei, and C.-C. Chiang. “A Group Mobility Model for Ad hoc Wireless Networks.” In ACM/IEEE MSWIM’99, pp.53-60, Seattle, WA, August 1999.

[HubauxC04] J.-P. Hubaux, S. Capkun and J. Luo. “The Security and Privacy of Smart Vehicles,” IEEE Security & Privacy Magazine, Vol. 2, No. 3, pp 49--55, May-June 2004[Jain03] M. Jain and C. Dovrolis. “End-to-End Available Bandwidth: Measurement Methodology, Dynamics, and Relation with TCP Throughput.” IEEE/ACM Transactions on Networking, Vol. 11, No. 4, August 2003.[Jardosh03] A. Jardosh, E. M. Belding-Royer, K. C. Almeroth, S. Suri. “Towards Realistic Mobility Models For Mobile Ad hoc Networks.” In ACM MobiCom’03, San Diego, CA, September 2003.[Johnson96] D.B. Johnson and D.A. Maltz. “Dynamic Source Routing in Ad Hoc Wireless Networks.” In Mobile Computing, Chapter 5, pp. 153-181, Kluwer Publishing Company, 1996.[Kohler00] E. Kohler, R. Morris, B. Chen , J. Jannotti and M. F. Kaashoek, “The Click Modular Router.” ACM Transactions on Computer Systems, Vol. 18, N. 3, 2000[Korkmaz04] G. Korkmaz, E. Ekici, F. Ozguner, and U. Ozguner, “Urban Multi-Hop Broadcast Protocols for Inter-Vehicle Communication Systems,” In ACM workshop on Vehicular ad hoc networks (VANET’04), Philadelphia, PA, USA, October 2004.[LeBoudec05] J.-Y. Le Boudec and M. Vojnovic, Perfect Simulation and Stationarity of a Class of Mobility Models, In IEEE INFOCOM’05, Miami, FL, 2005.[LeBrun05] J. LeBrun, C-N. Chuah and D. Ghosal. “Knowledge-Based Opportunistic Forwarding in Vehicular Wireless Ad Hoc Networks.” In IEEE Vehicular Technology Conference, May/June 2005.[Lee06] U. Lee, E. Magistretti, B. Zhou, M. Gerla, P. Bellavista, A. Corradi "Efficient Data Harvesting in Mobile Sensor Platforms," In IEEE International Workshop on Sensor Networks and Systems for Pervasive Computing (PerSeNS 2006), Pisa, Italy, March, 2006.[Levis03] P. Levis, N. Lee, M. Welsh, and D. Culler. "TOSSIM: Accurate and Scalable Simulation of Entire TinyOS Applications." In Proceedings of the First ACM Conference on Embedded Networked Sensor Systems (SenSys 2003), L.A., CA, November 2003.[Levis05] P. Levis, S. Madden, D. Gay, J. Polastre, R. Szewczyk, K. Whitehouse, A. Woo, J. Hill, M. Welsh, E. Brewer, and D. Culler. "TinyOS: An Operating System for Wireless Sensor Networks.” Ambient Intelligence, 2005. [Lian06] Q. Lian, Y. Peng, M. Yang, Z. Zhang, Y. Dai, X. Li, “Robust Incentives via Multi-level Tit-for-tat,” In International Workshop on Peer-to-Peer Systems (IPTPS '06), Santa Barbara, CA, Feb 2006.[Luo04] J. Luo and J.-P. Hubaux. “A Survey of Inter-Vehicle Communication.” Technical Report IC/2004/24, School of Computer and Communication Sciences, EPFL, 2004.[LW04] C. Lindemann and O. Waldhorst, Exploiting Epidemic Data Dissemination for Consistent Lookup Operations in Mobile Applications, ACM Mobile Computing and Communication Review, 8, 44-56, 2004.[LW05] C. Lindemann and O. Waldhorst, Modeling Epidemic Information Dissemination on Mobile Devices with Finite Buffers, Proc. ACM Int. Conf. on Measurement and Modeling of Computer Systems (SIGMETRICS 2005), Banff, Canada, June 2005.[Maihöfer05] C. Maihöfer, T. Leinmüller, and E. Schoch. “Abiding Geocast: Time-Stable Geocast for Ad Hoc Networks.” In ACM workshop on Vehicular ad hoc networks (VANET’05), Cologne, Germany, September 2005.[Marti00] S. Marti, T.J. Giuli, K. Lai and M. Baker, "Mitigating Routing Misbehavior in Mobile Ad Hoc Networks." In ACM MobiCom’00, Boston, NY, August 2000.[Motani05] M. Motani, V. Srinivasan, P. Nuggehalli. “PeopleNet: Engineering A Wireless Virtual Social Network” In ACM MobiCom’05, Cologne, Germany, Aug.-Sept. 2005.[Moustafa05] H. Moustafa, G. Boudron, and Y. Gourhand. “AAA in Vehicular Communication on Highways with Ad Hoc Networking Support: A Proposed Architecture,” in Proc. of ACM workshop on Vehicular ad hoc networks (VANET’05), Cologne Germany, Sept. 2005. [MP02] M. Mohsin and R. Prakash. “Ip address assignment in mobile ad hoc networks”. In Proc. of IEEE MILCOM 2002, September 2002.[Nadeem03] T. Nadeem, S. Dashtinezhad, C. Liao and L. Iftode. “TrafficView: Traffic Data Dissemination using Car-to-Car Communication,” In ACM Mobile Computing and Communications Review (MC2R), Vol. 8, No. 3, pp. 6-19, July 2003.

[Nandan05] A. Nandan, S. Das, B. Zhou, G. Pau, and M. Gerla, "AdTorrent: Digital BillBoards for Vehicular Networks." In Vehicle-to-Vehicle Communications Workshop (V2VCOMM 2005), San Diego, CA, USA, July 2005.[Nandan06] A. Nandan, S. Das, S. Tewari, M. Gerla, L. Klienrock. “AdTorrent: Delivering Location Cognizant Advertisements to Car Networks,” In The Third International Conference on Wireless On Demand Network Systems and Services(WONS 2006), Les Menuires, France, January, 2006[NandanD05] A. Nandan, S. Das, G. Pau, M.Y. Sanadidi, and M. Gerla. "Cooperative downloading in Vehicular Ad Hoc Networks." In Wireless On-Demand Network Systems and Services (WONS 2005), St Moritz, Switzerland, 2005. [NoWURL] NoW project. http://www.network-on-wheels.de[NP02] S. Nesargi and R. Prakash. “MANETconf: Configuration of Hosts in a Mobile Ad Hoc Network”, In Proc. of INFOCOM 2002, New York, USA, June 2002.[P1556] http://www.leearmstrong.com/DSRC/DSRCHomeset.htm[PFRPG07] C. E. Palazzi, S. Ferretti, M. Roccetti, G. Pau, and M. Gerla. “How Do You Quickly Choreograph Inter-Vehicular Communications? A Fast Vehicle-to-Vehicle Multi-Hop Broadcast Algorithm, Explained”, in Proc. of the 3rd IEEE CCNC International Workshop on Networking Issues in Multimedia Entertainment (CCNC/NIME 2007), Las Vegas, NV, USA, IEEE Communications Society, Jan 2007.[PS01] M. Papadopouli and H. Schulzrinne, Effects of Power Conservation, Wireless Coverage and Cooperation on Data Dissemination among Mobile Devices, Proc. 2nd ACM MobiHoc 2001, Long Beach, NY, 117-127, 2001.[Parno05] B. Parno and A. Perrig. “Challenges in Securing Vehicular Networks,” In Workshop on Hot Topics in Networks (HotNets-IV), MD, November 2005.[Przydatek03] B. Przydatek, D. Song, and A. Perrig. “SIA: Secure Information Aggregation in Sensor Networks”, In ACM SenSys’03 (Conference on Embedded Networked Sensor Systems), November 2003.[Raya05] M. Raya and J.-P. Hubaux. “The Security of Vehicular Ad Hoc Networks,” In ACM Workshop on Security of Ad hoc and Sensor Network 2005 (SASN'05), November 2005.[RFC2462] S. Thomson and T. Narten. “IPv6 Stateless Address Autoconfiguration”, December 1998.[RGPFP07] M. Roccetti, M. Gerla, C. E. Palazzi, S. Ferretti, and G. Pau. “First Responders' Crystal Ball: How to Scry the Emergency from a Remote Vehicle”, in Proc. the 1st IEEE International Workshop on Research Challenges in Next Generation Networks for First Responders and Critical Infrastructures (NetCri 07) - 26th IEEE International Performance Computing and Communications Conference (IPCCC 2007), New Orleans, LA, IEEE Computer Society, Apr 2007.[Royer01] E. M. Royer, P. M. Melliar-Smith, and L. E. Moser. “An Analysis of the Optimum Node Density for Ad hoc Mobile Networks.” In IEEE International Conference on Communications, Helsinki, Finland, June 2001.[Sampige05] R. Sampigethaya, L. Huang, M. Li, R. Poovendran, K. Matsuura, and K. Sezaki. “CARAVAN: Providing Location Privacy for VANET,” In Embedded Security in Cars (ESCAR), November 2005.[Shimonishi05] H. Shimonishi, M. Y. Sanadidi, and M. Gerla. “Improving Efficiency-Friendliness Tradeoffs of TCP in Wire-Wireless Combined Networks,” In ICC’05, Seoul, Korea, 2005[Stojmenovic02] I. Stojmenovic. “Position-based routing in ad hoc networks.” IEEE Communications Magazine, Volume: 40, Issue: 7, pp128-134, July 2002. [Strauss03] J. Strauss, D. Katabi and F. Kaashoek. “A Measurement Study of Available Bandwidth Estimation Tools.” In Internet Measurement Conference (IMC’03), Miami, Florida, October 2003.[Sun04] Y. Sun and E. M. Belding-Royer. “A Study of Dynamic Addressing Techniques in Mobile Ad hoc Networks”, In Wireless Communications and Mobile Computing, pages 315–329, April 2004.[Toner03] S. Toner and D. OMahony. “Self-Organising Node Address Management in Ad-hoc Networks”. In Proc. of Personal Wireless Communications (PWC 2003), Venice, Italy, September 2003.[Vogt05] Harald Vogt, "Increasing Attack Resiliency of Wireless Ad Hoc and Sensor Networks,", In International Workshop on Security in Distributed Computing Systems (SDCS) (ICDCSW'05), Columbus, OH, 2005.

[VSC] US Vehicle Safety Communication Consortium, http://www-nrd.nhtsa.dot.gov/pdf/nrd-12/CAMP3/pages/VSCC.htm[Weniger05] K. Weniger. “PACMAN: Passive Autoconfiguration for Mobile Ad Hoc Networks. IEEE Journal On Selected Areas In Communications”, 23(3), March 2005.[Wischhof05] L. Wischhof, A. Ebner, and H. Rohling. “Information Dissemination in Self-organizing Intervehicle Networks.” IEEE Transactions on Intelligent Transportation Systems, 6:90–101, March 2005.[Wu04] H. Wu, R. Fujimoto, R. Guensler and M. Hunter. "MDDV: A Mobility-Centric Data Dissemination Algorithm for Vehicular Networks," In ACM workshop on Vehicular ad hoc networks (VANET’04), Philadelphia, PA, USA, October 2004.[XGQ03] K. Xu, M. Gerla, L. Qi, and Y. Shu, Enhancing TCP Fairness in Ad Hoc Wireless Networks using Neighborhood RED, Proc. ACM MOBICOM 03, San Diego CA, 2003.[Xu03] K. Xu, M. Gerla, L. Qi, and Y. Shu. “Enhancing TCP Fairness in Ad Hoc Wireless Networks using Neighborhood RED,” In ACM MOBICOM 03, San Diego CA, 2003.[Xu04] B. Xu, A. Ouksel, O. Wolfson, "Opportunistic Resource Exchange in Inter-Vehicle Ad-Hoc Networks,” In IEEE International Conference on Mobile Data Management (MDM'04), Berkeley, CA, Jan. 2004.[YeLLZ04] F. Ye, H. Luo, S. Lu, L. Zhang, “Statistical En-route Filtering of Injected False Data in Sensor Networks”, In IEEE INFOCOM’04, Hong Kong, March 2004.[Yoon03] J. Yoon, M. Liu, and B. Noble. “Random Waypoint Considered Harmful.” In IEEE INFOCOM’03, San Franciso, CA, March 2003. [Yu04] X. Yu, “Improving TCP Performance over Mobile Ad Hoc Networks by Exploiting Cross-Layer Information Awareness,” Proc. ACM MOBICOM 04, Philadelphia, PA, 2004.[Zarki02] M. El Zarki, S. Mehrotra, G. Tsudik and N. Venkatasubramanian. “Security Issues in a Future Vehicular Network.” In EuroWireless, February 2002[Zhao06] J. Zhao and G. Cao. “VADD: Vehicle-Assisted Data Delivery in Vehicular Ad Hoc Networks.” In IEEE INFOCOM’06, Barcelona, Spain, April 2006.[Zhou05] P. Zhou, T. Nadeem, P. Kang, C. Borcea, and L. Iftode. “EZCab: A Cab Booking Application Using Short-Range Wireless Communication” In IEEE PerCom 2005, Kauai Island, Hawaii, March 2005.[ZhuSJN04] S. Zhu, S. Setia, S. Jajodia, and P. Ning. “An Interleaved Hop-by-Hop Authentication Scheme for Filtering False Data Injection in Sensor Networks,” In IEEE Symposium on Security and Privacy, Oakland, CA, May 2004. [GLS] Yinzhe Yu∗, Guor-Huar Lu†, and Zhi-Li Zhang∗ Enhancing Location Service Scalability with HIGH-GRADE, MASS 2004, Florida, Nov 2004[directional forwarding] Yeng-Zhong Lee, Mario Gerla, Jason Chen, Jiwei Chen,Biao Zhou and Antonio Caruso, “Direction” Forward Routing for Highly Mobile Ad Hoc Networks, Ad Hoc & Sensor Wireless Networks Journal, January, 2006[GPSR] Brad Karp and H.T. Kung “GPSR: Greedy Perimeter Stateless Routing for Wireless Networks”, Mobicom 2000[DSR] D. B. Johnson and D. A. Maltz, "Dynamic Source Routing in Ad-Hoc WirelessNetworks," Mobile Computing, 1994.[AODV] Charles E. Perkins and Elizabeth M. Royer. "Ad hoc On-Demand Distance VectorRouting." Proceedings of the 2nd IEEE Workshop on Mobile Computing Systemsand Applications, New Orleans, LA, February 1999, pp. 90-100.[ODMRP] S.-J. Lee, M. Gerla, and C.-C. Chiang, "On-Demand Multicast Routing Protocol," Proceedings of IEEE WCNC'99, New Orleans, LA, Sep. 1999, pp. 1298-1302.[Broadnets] F. de Rango, Mario Gerla, Biao Zhou, S. Marano, Geo-LANMAR Routing: Asymptotic Analysis of a Scalable Routing Scheme with Group Motion Support,The IEEE Second International Conference on Broadband Networks (BROADNETS 2005), Boston, USA, Oct. 3-7, 2005

[Zhou05] B. Zhou, F. D. Rango, M.Gerla, S. Marano. "GeoLANMAR: Geo Assisted Landmark Routing for Scalable, Group Motion Wireless Ad Hoc Networks," In IEEE 61st Semiannual Vehicular Technology Conference (VTC2005-Spring), Stockholm, Sweden, May 30-June 1, 2005.[Zorzi03] M. Zorzi, R. R. Rao, "Geographic Random Forwarding (GeRaF) for Ad Hoc and Sensor Networks: Energy and Latency Performance," IEEE Transactions on Mobile Computing, vol. 02, no. 4, pp. 349-365, Oct-Dec, 2003