Content uploaded by Sergio Martínez Tornell
Author content
All content in this area was uploaded by Sergio Martínez Tornell on Oct 01, 2016
Content may be subject to copyright.
1
DTN Protocols for Vehicular Networks:
an Application Oriented Overview
Sergio M. Tornell, Carlos T. Calafate, Juan-Carlos Cano and Pietro Manzoni
Universitat Polit`
ecnica de Val`
encia
Camino de Vera, s/n, 46022 Valencia, Spain
sermarto@upv.es, {calafate, jucano, pmanzoni}@disca.upv.es
Abstract
This survey provides an in-depth analysis of the different proposals for Vehicular Delay Tolerant Networks
(VDTNs). We introduce the DTN architecture and classify VDTN proposals according to the type of knowledge
needed to route messages. This classification also includes some Delay Tolerant Network (DTN) protocols originally
designed for Opportunistic Networks to illustrate the evolution from Opportunistic DTN protocols to VDTN protocols.
We also identify a set of common mechanisms that can be applied to almost all the VDTN protocols, heavily
influencing their performance. Finally, we present some applications where VDTNs can be applicable and evaluate
the suitability of the different proposals for each specific application. Moreover, this survey is not only limited to
describing the different protocols but also focuses on the reproducibility and repeatability of experiments. With this
in mind, we also review the evaluation methods used by VDTN researchers. We identify a lack of realism in most
of the simulation models used by the VDTN research community, providing certain guidelines to address this issue.
Index Terms
Delay tolerant networks, Vehicular networks, VANET, survey
I. INTRODUCTION
Wireless networks have evolved at a very fast rate and are applicable to several contexts and different communica-
tion solutions. In the automobile industry, many wireless solutions have been proposed to improve safety-related and
data communication among vehicles and between vehicles and infrastructure. These proposals form the Intelligent
Transport Systems (ITSs) field, which aims to improve the efficiency and security of transportation using Vehicular
Networks (VNs).
Although, VNs make use of Vehicular Ad-Hoc Networks (VANETs) for Vehicle to Vehicle (V2V) communication,
the concept of VN expands VANETs by adding Vehicle to Infrastructure (V2I) as well as cellular communication.
Sometimes, VANETs are considered a subset of Mobile Ad-hoc NETworks (MANETs). However, the high speed of
October 22, 2015 DRAFT
the nodes in a VANET, and the presence of obstacles like buildings, produce a highly variable network topology, as
well as more frequent partitions in the network. Therefore, typical MANET protocols [1] do not adapt very well to
VANETs since a complete connected path between sender and receiver is usually missing. Under these conditions,
Delay Tolerant Networks (DTNs) [2] are an alternative able to deal with VANET characteristics, and are applicable
to VN for ITS.
DTNs originated as a proposal for InterPlanetary Networks (IPNs) to provide communication between satellites,
and base stations. DTNs allow for information to be shared between nodes even in the presence of high delays,
which are typical in spatial communications. In DTNs, when a message cannot be routed to its destination, it
is not immediately dropped but is instead, stored and carried until a new route becomes available. Messages are
removed from the buffer when their lifetime expires or for buffer management reasons. This mechanism cannot
only be applied to IPNs but also to VNs, taking advantage of their high degrees of mobility [3][4]. DTNs have been
standardized by the Delay Tolerant Network Research Group (DTNRG) [5] to ensure network interoperability.
The research community has been very active over recent years, proposing new protocols and applications for
Vehicular Delay Tolerant Networks (VDTNs). This diversity may overwhelm the inexperienced researcher. Our aim
in this survey is to provide the reader with a broad view of the different proposals for VDTNs. We classify them
according to their main routing metric, showing their relationships and evolution. We also present the applications
where VDTNs can be applied, and evaluate the suitability of different protocols for each application.
There have been other works to survey DTN routing proposals and opportunistic routing for VNs, but, as far
as we know, this is the first survey to specifically focus on VDTNs and how Opportunistic DTN protocols have
evolved into VDTN protocols.
Before VDTN became a hot research topic, in [6], the authors developed a framework to classify DTN routing
algorithms and protocols. Their framework described routing protocols based on i) routing objective, ii) proactive
routing vs reactive routing, iii) source routing vs per-hop routing, and iv) message splitting. To classify routing
algorithms they defined several knowledge oracles, called Contacts summary Oracle, Contacts Oracle, Queuing
Oracle and Traffic Demand Oracle, which gradually increase the knowledge available at the nodes. Based on the
knowledge of the nodes they mathematically formulated the DTN routing problem as several resource management
problems and proposed mathematical algorithms to solve them.
In [7], the authors presented a survey of the most representative DTN protocols for MANETs to date (2006).
They distinguished between i) deterministic routing, ii) epidemic and random routing, iii) link forwarding probability
estimation, and iv) the model based approach. Most of the modern routing VDTN protocols we survey in this paper
may have been included in the last category. They also included “node movement control based” algorithms, which
allow the routing protocol to control the movement of certain nodes, and “network coding” methods. The earlier
types of algorithm clearly do not apply to vehicular networks where vehicles move freely.
In a more recent work [8], the authors presented a survey on VANET routing protocols that included a small
2
section devoted to DTN protocols. This section was insufficient and only summarized some of the characteristics
of VAAD [9] and GeOpps [10].
In [11], the authors performed an extensive survey of DTN architectures, analyzing the bundle protocols and its
advantages and disadvantages. They did not classify DTN routing protocols, but instead presented some mechanisms
generally applicable to any DTN routing protocol and listed several protocols that use them. Their work provides
a broad view of the DTN routing problem, without considering the special characteristics of VDTNs.
Position-based routing surveys have been previously published [12]. Although some works referred to in this
paper match the definition of “position-based routing”, our analysis focuses on the DTN characteristics of the
protocol, while previous papers focused mainly on their pure geographic characteristics.
In [13], the authors performed an analysis of certain DTN routing protocols in vehicular networks. We consider
the scope of their work to be limited, as they consider only a dozen protocols while in this survey we consider 41
different contributions.
More recently, in [14], the authors presented a detailed DTN survey with more than 140 referred papers. However,
their impressive work focused solely on Opportunistic DTN protocols and, therefore they did not cover most of
the protocols we analyze in this survey. Their classification of DTN routing protocols was one of the bases of this
work and we really encourage the reader to read their article in order to obtain a broader perspective of the DTN
routing protocols universe.
As far as we know, this is the first survey to have focused on VDTNs and their applications. Moreover, this survey
is not limited to protocol descriptions, focusing on reproducibility and repeatability of experiments, we include a
review of the evaluation methods used by VDTN researchers.
The rest of this article is organized as follows: Section II introduces the DTN paradigm and its standards,
discussing their suitability to VNs. Section IIIanalyzes and classifies VDTN protocols. Section IV surveys the
methods used by researchers to evaluate VDTN protocols. Later, Section V introduces applications proposed by the
research community which depend on the use of VDTNs and, finally, Section VI concludes the paper and provides
some insights on future trends.
II. OVE RVIEW OF DELAY TOLERANT NETWORKS
The DTN paradigm was initially proposed to enable communication between satellites, surface rovers, and other
devices within the IPN [16] [17]. Space communication may suffer high delays and frequent disconnections. The
DTN concept was also adapted for wildlife monitoring [18] and remote village communication [19], [20]. However,
DTN solutions used their own protocols and were unable to intercommunicate. To enable intercommunication
between different DTNs, regardless of the network technology, the DTNRG [5] started to work towards its stan-
dardization [3]. Figure 1 represents a heterogeneous DTN, which interconnects the IPN with terrestrial DTN nodes.
3
As a result of these efforts, in 2007 two RFCs were published in 2007 that defined the DTN architecture [2] and
the Bundle protocol [21]. The following subsection describes the DTN architecture.
A. Architecture and Standards
To support the heterogeneity of different networks the DTN architecture is designed to run as an overlay network
over the network layer (IP in the case of the Internet). To do so, two new layers are added: The bundle layer, and the
convergence layer [21]. The bundle layer encapsulates application data units into bundles, which are then forwarded
by DTN nodes following the bundle protocol. The convergence layer abstracts the characteristics of lower layers
to the bundle layer. The convergence layer does not need to run over the internet protocol stack, thus allowing for
the implementation of DTNs over any type of network.
1) The Bundle Protocol: The Bundle Protocol stores and forwards bundles between DTN nodes. Instead of
end-to-end forwarding, the Bundle Protocol performs hop-by-hop forwarding. To deal with network disruption, the
Bundle Protocol can store bundles in permanent storage devices until a new transmission opportunity appears. The
concept of reliable custody transfer ensures that a DTN node will not remove a bundle from its buffer until another
node has taken custody of it.
The Bundle Protocol operation depends on contacts. A contact occurs when a connection between two DTN
nodes can be established. The contact type depends on the type of operating network: it may be deterministic, as
in Interplanetary networks, opportunistic, as in VN, or persistent, as in the Internet.
Fig. 1. Heterogeneous Delay Tolerant Network Example [15].
4
UDP
LLC
PHY
MAC
IPv6
PHY
MAC
TCP
Wave Short Message Protocol
(WSMP)
802.11p
Multi Channel
Link and Physical
Layers
Fig. 2. WAVE Architecture.
Where the size of a bundle exceeds the maximum transferred data of contacts, the bundle protocol must perform
fragmentation. Fragmentation is supported in two different schemes: proactive, where a DTN node may fragment
an application message into different bundles and forwards every bundle independently, and reactive, where bundles
are fragmented during transmissions between nodes.
2) The Convergence Layer: The convergence layer abstracts the characteristics of lower layers to the bundle
protocol and is in charge of sending and receiving bundles on behalf of the bundle protocol. The convergence layer
allows for any set of lower protocols to be used to reliably transfer a bundle between two DTN nodes. For example
the TCP/IP convergence layer [22] uses a TCP connection between two DTN nodes to transfer bundles. That TCP
connection can be established via the Internet. To implement a DTN over other technologies, new convergence
layers are needed. Convergence layers must provide the bundle protocol with a reliable delivery and reception
mechanism.
3) The Generic Opportunistic Routing Framework (GORF): After the standardization of DTN architecture, the
DTNRG focused on the routing protocols, releasing GORF [23]. GORF architecture specifies all necessary basic
functionalities common for utility-based routing protocols, and provides a framework to easily define and implement
any opportunistic routing protocol for DTNs. To date, only the Epidemic protocol [24] and the PRoPHET protocol
[25] have been standardized [26] [27].
The GORF assumes that nodes are able to detect their neighbors using a service running independently. When
a neighbor has been detected the protocol sets up a link between the current carrying node, called custodian, and
the detected neighbor, called candidate. Once a link is established, nodes exchange routing information on other
nodes in the network. Afterwards, the custodian sends a bundle offer that contains a list of the bundles in its buffer.
Then, the candidate responds with a list of requested bundles, that will be forwarded to it.
B. Does the Standard Apply to Vehicular Networks?
Before evaluating the suitability of DTN standards to VN, it is worth briefly introducing the currently approved
standard in the USA for ITS: the Wireless Access for Vehicular Environment (WAVE) standard [28]. This standard
uses the 5.9 GHz band by relying on the 802.11p protocol for medium access [29]. Figure 2 shows its architec-
5
Application
Bundle Layer
Convergence
Layer
Bundle Layer
Convergence
Layer
Bundle Layer
Convergence
Layer
Bundle Layer
Convergence
Layer
Application
Source Destination
Intermediary
Vehicle 1
Intermediary
Vehicle 2
Application
Bundle Layer
Convergence
Layer
TCP/UDP
IPv6
Legacy Internet
DTN Stack
DTN
Layers
Layers
Replaced
Link
PHY
WSMP WSMP WSMP WSMP
802.11p 802.11p 802.11p 802.11p
WAVE
LAYERS
Fig. 3. Comparison between Internet DTN stack and VDTN stack. Scheme of a message transmission in a VDTN
ture. WAVE architecture includes two different transport/network layers: one compatible with IPv6 and its own
network/transport layer based on the WAVE Short Message Protocol (WSMP), which reduces the overhead.
The standard DTN protocol stack can be used directly in VDTNs through the IPv6 compatible stack. To implement
a pure VDTN directly over the WSMP, which introduces less overhead and more flexibility, the only requirement
is the implementation of a convergence layer between the bundle layer and the WSMP. Figure 3 compares the pure
VDTN stack against the legacy internet DTN stack. Few researchers have tried to adopt the standard DTN stack for
VDTNs. Among the papers reviewed in this survey, only those proposals which were tested on the UMassDieselNet
testbed [30] implemented the standard DTN stack.
With regards to GORF architecture as it is proposed at present it may be applicable to all of the unicast protocols
surveyed in this article. However, due to its newness, none of the protocols exactly match the functions and phases
defined by the GORF. The main difference arises in the node that performs the routing decision process. Most
proposals consider that the custodian node must decide whether or not to forward a bundle, according to its
neighbors’ characteristics; whereas GORF architecture assigns the routing decision process to the candidate node,
which requests bundles stored in the custodian buffer. Since the candidate node may have a different local view of
the network status, decisions may be different, and the routing information exchange phase should be appropriately
adapted.
III. DTN PROTO CO LS F OR VANETS: TAXONOMY
In this section, DTN protocols are classified according to different parameters. Firstly, they must be grouped
together according to the objective of the protocol: a) protocols whose objective is to disseminate messages to all
the nodes in the network (Dissemination) and b) protocols whose messages have a specific destination that can
either be a vehicle or an Road Side Unit (RSU) (Unicast). Secondly, they are grouped together according to the
amount of control information required by each protocol. Inside the dissemination protocols group, we distinguish
between the epidemic approach and a group of protocols that uses geographic information to estimate connectivity
of nodes (geo-connectivity). Inside the unicast group, we distinguish between zero knowledge protocols, those
6
that do not require any knowledge about the vehicles status or the environment and utility based protocols. Utility
based protocols try to estimate the benefit of each transmission (i.e. how a transmission improves the probability of
reaching the destination) to determine the best forwarding node among neighbors. Each protocol estimates this utility
using a pre-defined metric. We have divided these utility-based proposals into five different categories, according to
the type of knowledge they need: i) contact history & social relationships,ii) geographic location,iii) road map,
iv) hybrid protocols and v) online protocols. The “online” subcategory includes protocols that, besides combining
several simpler protocols, require information on the current state of the road network or use sophisticated metrics
that do not fit into any other category. Figure 4 summarizes this classification, while Figure 5 orders and classifies
the protocols collected in this survey chronologically. For each category, we first list the different protocols forming
part of it before describing those protocols and, finally, explaining their advantages and disadvantages.
Fig. 4. DTN Protocol Taxonomy.
Table I summarizes the characteristics of the different proposals. The second column indicates whether the
protocols were originally proposed for VN or not. The third column contains the objective application of each
protocol as it is stated in its original publication. The fourth column classifies protocols according to the classification
explained previously. The fifth column offers a quick and simple description of the routing metric used by each
proposal. Finally, the columns under the “Optimizations” label indicate whether the mechanisms described in
Subsection III-D are considered in the proposal.
A. Dissemination Protocols
The objective of the dissemination protocols is to inform as many nodes as possible of an event. The most
obvious solution is the simple flooding scheme, where nodes rebroadcast every message received [31]. However,
this scheme generates some well-known problems, such as the broadcast storm [32], or infinite rebroadcasting loops
that waste resources. To limit the impact of these problems, some modifications to the simple flooding scheme have
been proposed [33]. Simple flooding and its modifications are limited by the connectivity of the network: they will
only propagate messages as long as the network is connected. In this section we present proposals that add DTN
7
support to dissemination protocols. Since DTN dissemination protocols are not limited by the connectivity of the
network, the dissemination process must be limited in time or space to avoid collapsing the network.
8
TABLE I
CHA RACT ER IST IC S OF DI FFE REN T PROT OC OLS .
Protocol Year VN specific Application Group Routing Metric Optimizations
Reliability Redundancy Messages Priority
Epidemic [24] 2000 No Dissemination Zero Knowledge – – Multicopy No
ProPHET [25] 2003 No P2P Contact History Contact Rate No Open No
MoVe [57] 2005 Yes Collect Geographic Loc. Direction No No No
Spawn [76] 2005 Yes Cooperative Download Geographic Loc. Distance No Fragment Multicoy Neighbor Dist
Spray&Wait [44] 2005 No P2P Zero Knowledge – No Multicopy No
MaxProp [30] 2006 Yes P2P Zero Knowledge Conctac Rate End-End ACK Multicopy PRoPHET
RAPID [?] 2007 No P2P Zero Knowledge Contact Rate No Multicopy Contact Rate
SimBet [55] 2007 No P2P Social Social Graph No No No
GeOpps [10] 2007 Yes P2P/V2I Road Map Nearest Point, ETA No No No
[77] 2008 Yes Collect Zero Knowledge Direct No No No
POR [48] 2008 No P2P Zero Knowledge Distance No Multicopy Distance
DAER [49] 2008 Yes P2P/V2I Zero Knowledge Distance No Multicopy Distance
VADD [9] 2008 Yes P2P/V2I Online Loc+Density+Speed No No No
DSCF [35] 2009 Yes Dissemination Geographic Loc. Loc+Connectivity No No No
FFRDV [36] 2009 Yes Dissemination Geographic Loc. Speed Hop ACK No No
Infocast [34] 2009 Yes Dissemination Zero Knowledge – – Rateless Coding No
ADPBSW [50] 2009 No P2P Contact History Contact Rate No No No
Adv. ProPHET [53] 2009 No P2P Contact History Contact Rate No No No
Extended GeOpps [61] 2010 Yes Cooperative Download Road Map GeOpps+Estimated Route No No No
C-DTN [78] 2010 Yes Dissemination No Data No Data Open Open –
DvCast [39] 2010 Yes Dissemination Connectivity Loc+Connectivity No No No
ROD [37] 2010 Yes Dissemination Connectivity Loc+Connectivity No No No
Uv-Cast [38] 2010 Yes Dissemination Connectivity Loc+Connectivity No No No
ProPHET+ [?] 2010 No P2P Contact History Buffer+Power+Contact Rate No No No
DRTAR [69] 2010 Yes P2P/V2I Road Map Loc+Density+Speed No No No
GeoDTN+NAV [63] 2010 Yes P2P/V2I Geographic Loc. GPCR+GeOpps No No No
[62] 2010 Yes P2P/V2I Road Map Nearest Point No No No
SADV [68] 2010 Yes P2P/V2I Online Loc+Density+Speed No No No
D-Greedy D-MinCost [70] 2011 Yes Collect Online Nearest Point/VADD like Hop ACK No No
SERVUS [?] 2011 Yes Dissemination Geographic Loc. Loc+Connectivity Hop ACK Multicopy No
DTFR [65] 2011 Yes P2P Hybrid S&W+Location No Multicopy No
Orion [64] 2011 No P2P Hybrid Distance+Contact Rate No No No
RENA [66] 2011 Yes P2P Hybrid Contact rate+S&W No No No
GeoSpray [51] 2011 Yes P2P/V2I Hybrid S&W+GeOpps No Multicopy No
DSRelay [79] 2012 Yes Cooperative Download Distance Direction on Highway No No No
[71] 2012 Yes Cooperative Download Online Predicted Contacts No Fragment Multicoy No
CAN DELIVER [72] 2012 Yes P2P Online Nearest Point End-End ACK Multicopy No
RWR [67] 2012 No P2P Hybrid Distance+RSU No No No
CSM [80] 2013 Yes Collect Data Agregation Geographic Location No Yes No
MSDP [60] 2013 Yes Collect Road Map Nearest Point+Buffer Hop ACK Fragment Redundancy No
ZOOM [54] 2013 Yes P2P Social Social Graph + Contact Rate No No No
9
2004 Randomized
Fig. 5. Protocols ordered chronologically, grouped by knowledge required. Protocols at the end of the arrows are an evolution of the protocol
at the beginning of the arrow.
1) Epidemic Protocol: The simplest DTN dissemination protocol is the Epidemic protocol [24], which consists
of sharing all the messages in the nodes’ buffers every time a contact occurs. The Epidemic protocol needs a
negotiation phase to determine which messages to share, increasing the delay and generating more overhead than
the non-DTN proposals. In dense networks this negotiation traffic may be even bigger than data traffic. Moreover,
the Epidemic protocol neglects the opportunity of a node overhearing a message from broadcast transmissions
between neighbors. The Infocast protocol [34] extends the Epidemic protocol with fragmentation and coding, to
give better performance.
10
2) Geographic & Connectivity Protocols: Within this category we include DTN dissemination protocols that
need information on node location. This information can be used to limit the number of messages exchanged by
nodes and to estimate the connectivity of the network in order to choose the best possible candidate as the new
carrier. This carrier will bring the message to the next cluster. The protocols matching this definition are: Directional
Store-Carry-Forward (DSCF) [35], Fastest Ferry Routing in DTN-enabled Vehicular Ad-Hoc (FFRDV) [36], Road
Oriented Dissemination (ROD) [37], Urban Vehicular BroadCast (UV-CAST) [38], Distributed Vehicular BroadCast
(DV-CAST) [39], and SERVUS [40].
•The DSCF protocol [35] requires every node to have 2 different antennas. It works by following three simple
rules; i) messages received from one direction are transmitted in the opposite direction, ii) if there are no
vehicles in the propagation direction, the message is stored in the buffer until a new neighbor appears, and
iii) any duplicated message is ignored. Apart from the requirement of having two interfaces, which is not
considered in the WAVE standard, when several nodes rebroadcast they will probably collide when accessing
the channel. Moreover, it is limited to highways, where the propagation direction is clearly defined.
Block 1 Block 2 Block 3
AB
C
D
E
(a) T1: An event is detected and node Abecomes a carrier.
Block 1 Block 2 Block 3
HELLO
AB
C
D
E
(b) T2: Node Aenters a new block and broadcasts a beacon.
Block 1 Block 2 Block 3
85k/h
120k/h
AB
C
D
E
(c) T3: Only nodes moving away from the event answer the
beacon.
Block 1 Block 2 Block 3
Message
AB
C
D
E
(d) T4: Node Bis chosen as the new carrier, it will broadcast
a beacon as soon as it enters block 3.
Fig. 6. FFRDV example
•The FFRDV protocol [36] assumes that vehicles are moving on a highway. It divides the road into small
blocks, and, when an event occurs, the first vehicle passing by generates a message and becomes its carrier.
11
The carrier broadcasts a beacon message every time it enters a new block. Neighbors inside the same block
answer the beacon message with information on their speed and moving direction. Then, the fastest vehicle
moving towards the propagation direction is chosen to become the new carrier, while the remaining nodes
overhear the message. If no neighbor answers to the beacon, the carrier keeps it in its buffer until the next
block. It is clear that, besides the connectivity of the network, the propagation delay depends on the size of
the blocks. Moreover, since the FFRDV is invalid for city environments, it must be complemented by other
dissemination protocols. Figure 6 depicts the behavior of this protocol.
As long as the network is connected, multi-hop forwarding protocols disseminate information faster than store
and carry protocols. To take advantage of this characteristic, several protocols use the multi-hop forwarding scheme
until they detect a disconnected network. Then, they use geographic information to choose several carriers that will
carry the message further.
•The ROD protocol [37] does not need nodes to periodically send beacon messages. When a node receives a
message from another node, it decides whether to retransmit it according to its relative position with respect
to the sender. This phase of the protocol is similar to the Distance Defer Transmission (DDT) protocol [41].
If a node detects that none of its neighbors rebroadcasted a message, it switches to store-carry and forward
mode. In this mode, the node periodically rebroadcasts the message until it detects that another node has also
received and rebroadcasted the message.
A
S
N
N1
2
θ
θ
2
1
(a)
A
S
N1
N
2
θ
θ2
1
(b)
Fig. 7. UV-CAST example: in a) node Aswitch to SCF mode while in b) does not.
•The UV-CAST protocol [38] defines a Region of Interest (ROI) where the message must be disseminated.
The main difference between UV-CAST and ROD lies in how they choose the carrier nodes. While in ROD
the selection is only based on overhearing messages from neighbors, UV-CAST nodes use their geographic
location information to determine wheter they are boundary nodes for the source node’s connected region. To
determine if a node must switch to store-carry and forward mode UV-CAST follows this process. Suppose
node Areceives a message from the source (S), with Nneighbors (Ni) (Fig. 7): i) it calculates the angle
θibetween −→
AS and −−→
ANi,ii) if the sum of the smallest and largest angles is less than π,Amust switch to
store-carry and forward mode. Once in store-carry and forward mode, the node will rebroadcast the message
and switch to normal mode as soon as a beacon from a new neighbor is received.
12
•The DV-CAST protocol [39] is another example of a highway-limited protocol. As in ROD, nodes are
grouped into clusters, and they switch between normal and store-carry and forward modes according to
the estimated connectivity. DV-CAST defines three different operation modes, well connected neighborhood,
sparsely connected neighborhood, and totally disconnected neighborhood. In the first mode, nodes work in
normal mode; in the second mode, nodes switch to store-carry and forward when they move contrarily to the
message source and, finally, in the third mode, nodes always switch to store-carry and forward mode.
•The SERVUS protocol [40] follows a similar approach, where nodes modify their behavior according to the
location of their neighbors. In SERVUS, nodes detect whether they are the last node of a group of connected
nodes, called a cluster, and then rebroadcast previous messages when they contact a new node from outside
the cluster. In SERVUS, cluster detection is only based on the geographic location of neighbors obtained from
periodic beacons.
To conclude, in all these protocols, to choose the next carrier node the algorithms assume that all the nodes in
the neighborhood have the same information and, therefore, they depend greatly on the correctness of the neighbors
list, which can be easily compromised by a high loaded channel and high mobility. Moreover, the calculation of
angles and relative locations may be affected by the variability of heterogeneous Global Positioning System (GPS)
devices.
B. Unicast Protocols
Besides pure unicast protocols, we have included in this category those anycast protocols where the destination is
any of the RSUs present in the VN, since they are reduced to unicast by choosing the closest RSU as the destination.
The first subgroup inside the unicast protocols category is formed of protocols that do not need any external source
of information; we call these Zero Knowledge protocols. A much larger group includes protocols that estimate
the utility of each transmission, i.e. how a transmission improves the probability of reaching the destination to
determine the best forwarding node among neighbors. For the sake of clarity, we will discuss the Utility Based
Protocols in a separate subsection (III-C).
Under the Zero Knowledge category we have included protocols that do not need any external source of
information, or to collect information while they are running. As a result of this limitation, their performance
is usually surpassed by utility based protocols. Most of them were designed for intermittently connected MANETs
[1], but are usually used as a reference for comparison with VDTN protocols. The protocols included in this category
are: Direct [42], Randomized Routing [43], Epidemic [24], and Spray&Wait [44].
•The Direct is the simplest possible protocol [42]. It works as follows: a node Aforwards a message to a node
Bonly if Bis the destination. This case presents an unbounded delay but it has the advantage of performing
only a single transmission per message. It represents an upper bound for delay and a lower bound for delivery
ratio.
13
•The Randomized Routing protocol was presented in [43]. It works as follows; node Aforwards a message
to another node B, which Afinds with a given probability p. In its work, authors showed that random routing
behaves better than direct routing.
•The Epidemic protocol [24] has also been applied to the unicast problem. As long as enough resources are
available, the Epidemic protocol guarantees that messages will eventually arrive at their destination along the
shortest path. Therefore, under ideal conditions, the Epidemic protocol provides a lower bound for delay and
an upper bound for delivery probability. The main problem of the epidemic protocol is that it wastes resources
by propagating copies of messages that have already been delivered, and along paths that will never reach
the destination. In order to limit this resource wastage, researchers have proposed several modifications to the
original Epidemic protocol. In [45], authors presented four different mechanisms to block the propagation of
already-delivered messages. In [46], nodes exchange a copy of the messages with a probability smaller than 1,
which reduces the number of copies in the network. Protocols such as MaxProp [30], RAPID [47], POR [48],
and DAER [49] add message priority management techniques to make the most of every contact. We will go
into detail about these techniques in Section III-D.
D
#Copies
S
A
B
C
S:4 A:0 B:0 C:0
(a) T1: S generates a message for D.
D
#Copies
S
A
B
C
S:2 A:2 B:0 C:0
(b) T2: S sends a copy to A.
D
#Copies
S
A
B
C
S:1 A:2 B:1 C:0
(c) T3: S sends a copy to B.
D
#Copies
S
A
B
C
S:1 A:1 B:1 C:1
(d) T4: A sends a copy to C, which will finally deliver it
to D.
Fig. 8. Binary Spray and Wait example
14
•The Spray&Wait protocol [44] divides the propagation of messages into two different phases. Initially it
disseminates a certain number of copies of a message to neighbor nodes and then it waits until any of the
carrier nodes moves and reaches the destination of the message. Several spraying mechanisms were presented
and studied in [44], where the Binary Spray & Wait (BS&W) protocol offered the best results. In the BS&W
protocol, the source of a message initially starts with Lcopies. Any node Athat has n > 1message copies
(source or carrier) and encounters another node B(with no copies) hands bn/2ccopies over to Band keeps
dn/2efor itself. When only one copy is left, it switches to direct transmission. Figure 8 shows this behavior.
In the following subsections we will go into detail as to how some authors adapted these zero knowledge protocols
to turn them into utility based protocols, as seen for example in [50] and [51].
Since the protocols included in this category do not consider any type of external information, they are suitable
for environments where we cannot make any assumption about mobility models, road maps, or social relationships.
However, in VDTNs we typically find better alternatives because mobility is restricted to the road network, vehicles
are driven following certain rules and people usually live in communities.
C. Utility Based Protocols
We define the utility function as a function that combines several parameters to obtain an index that estimates
how a transmission would increase the probability of reaching the destination of a message (hereafter called the
Utility Index). In some protocols the utility function can be as simple as the distance to the destination, while in
others it may combine several parameters from different sources of information. In this section we classify utility-
based protocols into five different categories according to the type of knowledge they need to obtain the required
parameters to calculate the utility index: i) contact history & social relationships,ii) geographic location,iii)
road map,iv) hybrid protocols and v) online protocols.
1) Contact History & Social Relationship Protocols: The protocols included in this category work under the
assumption that the probability of a node meeting the destination node of a message can be estimated based on
the history of previous contacts. Although most of them were developed for MANETs, and are mainly applicable
to wildlife tracking systems [18] or pedestrian communities [52] (where this frequent contacts paradigm seems to
clearly apply), these protocols have been extensively used for comparison with VDTN protocols. In this category,
we find the following protocols: PRoPHET [25], APRoPHET [53], PRoPHET+ [52], ZOOM [54], and SimBet [55].
•PRoPHET, which was the first contact history based protocol, was presented in [25]. This protocol relies
on a self-defined delivery predictability metric,P∈[0,1], which is updated according to Equation 1, where
P(a,b)is the delivery predictability that node ahas for node b, and Pinit is an initialization constant. Note
that, nodes experiencing frequent encounters have a higher delivery predictability.
P(a,b)=P(a,b)old + (1 −P(a,b)old )×Pinit (1)
15
The defined delivery predictability ages (decreases its value) when two vehicles do not meet for a while.
PRoPHET also defined the transitivity property for the delivery predictability,i.e. if node afrequently encoun-
ters node b, and node boften encounters node c, node ais a good node to forward messages to c. To grasp
this behavior, the delivery predictability metric is updated in line with Equation 2, where βis a constant that
quantifies the impact of the transitivity on the delivery predictability metric.
P(a,c)=P(a,c)old + (1 −P(a,b)old )×P(a,b)×P(b,c)×β(2)
•The Advanced PRoPHET protocol was presented in [53]. It modifies the PRoPHET protocol’s metric to
smooth its variability. The main effect of the smoothed variability is that the protocol needs more time to react
to changes in the network.
•In [52], authors presented PRoPHET+, another improved version of the PRoPHET protocol that adds four
new parameters related to i) buffer (VB), ii) power (VP), iii) popularity (VO), and iv) bandwidth (VA). Using
Simple Additive Weighting, the utility function is defined as follows:
Vd=WB(VB) + WP(VP) + WA(VA) + WO(VO) + WP RoP HE T (VP RoP H ET )(3)
Where Wirefers to weight factors that must be configured by the users and whose optimal value depends
on the scenario. Their results showed that, by considering more variables and not only the contacts history,
the performance of PRoPHET is improved. They also proved that a misconfiguration of weight factors may
degrade the performance of the protocol.
A
B
C
D
Fig. 9. Social graph: nodes inside cluster C are connected better than nodes in others clusters.
•In [54] and [55], authors presented ZOOM and SimBet, which use social metrics, such as the node’s number
of links in the social graph or their centrality, to choose the next forwarding node. They complement the
delivery predictability by estimating the centrality of the node within the social graph formed by the nodes
16
inside the network. Figure 9 shows an example of the relationships inside a community. Nodes from cluster
C are better connected than nodes from other clusters, meaning that, those nodes are better carriers.
These routing schemes require a nearly-closed community to be effective: new nodes, which do not have previous
contacts, seem to be isolated, and nodes that left the network, which had a long contacts-history, seem to still
belong to it for a long time after leaving. Since a high mobility and a highly changing membership are among
of the main characteristics of VNs, the protocols studied in this section may tend to select old routes. Moreover,
several authors have shown that the average inter-contact time, when applied to VNs, is in the order of several
hours or even days [54], [56]. Since the inter-contact time is closely correlated to the expected delay to destination,
the applications running on top of one of these protocols should expect an end-to-end delay in the order of hours.
Finally, when using social metrics, the relationships between the nodes need to be carefully analyzed before full
deployment, which presents scalability and privacy issues.
2) Geographic Location Protocols: Protocols included in this subsection assume that each node is aware of its
location and its moving direction. Although we found only two examples of protocols related to VNs that match this
exact definition, we decided to create a new category since these can be considered the ancestors of more advanced
protocols that, beside location and direction, use other sources of information. Those protocols are Greedy-DTN
and MoVe [57].
•The Greedy-DTN protocol is closely related to the most referenced geo-assisted routing protocols in literature,
GPSR [58] and GPCR [59], which are not delay-tolerant protocols. In GPCR/GPSR messages are forwarded
greedily towards the destination, i.e., the best forwarding neighbor is the one closest to the destination. When
a message reaches a local minimum, where no neighbor is closer to the destination, it is routed in perimeter
mode in an attempt to find a new route. GPSR is generally adapted to DTN omitting the perimeter mode and
carrying the message inside the buffer until a better forwarding node to forward the messages appears. From
now on we will refer to this adapted version of GPSR as Greedy-DTN. Greedy-DTN has been widely used
as a reference for comparison with more sophisticated DTN protocols [10][60].
•MoVe [57] is a protocol that estimates the future location of the nodes using their current direction of movement.
In MoVe, the node whose estimated trajectory is the closest to destination becomes the best forwarding node.
A modification of MoVe, called MoVe-Lookahead, uses the location of the next waypoint (authors assumed
the random waypoint mobility model) to predict the mobility of the nodes and avoid forwarding messages to
nodes that will change their direction before arriving at the closest point to destination. Figure 10 shows an
example where node Bis the best forwarding node when using MoVe, while Cwould be the best forwarding
node when using MoVe-Lookahead.
These approaches are suitable for unrestricted mobility models, but ignore the fact that mobility in vehicular
networks, despite its high variability, is constrained to roads. Therefore, these proposals are prone to inducing
suboptimal routing decisions. For example, the Greedy-DTN protocol may get blocked when a constant flow of
17
S
D
A
B
C
WP
WP
WP
dMb
dMLbdMc
dMLc
=
Fig. 10. In MoVe, only the direction is taken into account to choose the next forwarding node, while in MoVe-Lookahead the way-points are
also considered. Therefore, when using MoVe, node S will choose node B to forward a message to D, while it will choose node C when using
MoVe-Lookahead.
Local Minimum
A
Flow of Cars
D
Fig. 11. When using geo-routing, if a constant flow of vehicles exists, messages for D could get stuck in A because there is a permanent local
minimum.
vehicles generates a permanent local minimum, as illustrated in Fig. 11. Besides, loops occur when two vehicles
moving in opposite directions meet. In the case of MoVe, it assumes a random-waypoint model for node movements,
ignoring the fact that the current direction of vehicles, especially in downtown or rough rural areas, may change
frequently and that it may not match the long term direction of movement.
Another problem of geo-assisted protocols is that they require a location service to obtain the destination’s
location. However, authors usually ignore this requirement. Without a location service, protocols are limited to
V2I communication. This problem also affects Road-Map protocols, which are covered in the next subsection. The
design of a location service is far from trivial and is outside the scope of this survey.
3) Road Map Protocols: Since vehicular mobility is always restricted to roads, the use of pure geographic
protocols, such as Greedy-DTN or MoVe, can lead to messages being forwarded to vehicles whose long term
destination is far from the destination of the message. The long term destination is important in the case of sparse
18
networks, where vehicles rarely meet. Protocols included in this section assume that vehicles have a Navigation
System (NS) that provides information on the road layout and the vehicle’s future route, besides an accurate
geographic location. The protocols included in this category are: GeOpps [10], and its extension [61], the protocol
presented in [62] and the Map-based Sensor-data Delivery Protocol (MSDP) [60].
D
A
B
C
NPa
NPc
Fig. 12. Calculation of the NP in GeOpps. Although NPBis closer to the destination than N PAand N PC, A and C nodes are probably
better forwarding nodes, since they will reach their NPs faster than B.
•The GeOpps protocol chooses the next-forwarding nodes based on the Minimum Estimated Time of Delivery
(METD) metric, which is the sum of: i) the estimated time that a vehicle would need to reach the nearest
point (NP) of its route to the destination, plus ii) the time a vehicle would need to travel from the NP to the
final destination. If the latter factor cannot be calculated, an estimation based on the straight distance can be
used [10]. Fig. 12 shows an example of the NP calculation.
GeOpps was extended to support traffic from RSU to vehicles [61] by routing the reply message to a point
inside the vehicle’s route, and then backwards on the vehicle’s path until the destination is reached.
•The protocol presented in [62] simplifies GeOpps by ignoring the speed of the vehicle, and selecting the vehicle
whose route passes closest to the destination as next carrier.
•MSDP is another protocol that uses the programmed route and the road layout to estimate the time required
to reach the destination [60]. It also takes into account the reliability of the programmed route, giving priority
to reliable nodes with fixed routes like buses or taxis over private vehicles which might modify their routes.
These protocols emphasize the store-carry phase, missing multi-hop communication opportunities, which increases
the delay. Moreover, they depend heavily on the reliability of programmed routes, which may present vulnerability.
4) Hybrid Protocols: In this category we include protocols that combine the behavior of several protocols
from those expounded previously. The protocols we find are: Average Delivery Probability Binary Spray and Wait
(ADPBSW) [50], GeoDTN+Nav [63], Orion [64], GeoSpray [51], Delay Tolerant Firework Routing (DTFR) [65],
REgioN-bAsed (RENA) [66], and RWR [67].
19
•ADPBSW [50] combines the PRoPHET protocol with the BS&W protocol. It was originally designed for
MANETs and is the first proposed hybrid protocol for DTNs. It complements the Spray & Wait protocol by
using the delivery probability calculated by PRoPHET to propagate copies only to vehicles experiencing a
delivery probability higher than the current carrier.
•GeoDTN+Nav is a protocol that divides the process of delivering a message into two different phases [63].
During the first phase it uses GPCR to forward the message near to the destination. Once a local maximum
is reached, the protocol switches to perimeter mode. Contrarily to GPCR, in GeoDTN+Nav, after a certain
number of hops in perimeter mode, the protocol switches to DTN mode, and the message is delivered using
the GeOpps protocol. The vehicle switchs back to GPCR phase if it finds a neighbor closer to the destination
than the previous local maximum that triggered the switch to DTN mode.
•Similar to GeoDTN+Nav, the Orion routing protocol [64] combines the Greedy-DTN protocol with a con-
tact history based protocol. Therefore, messages are forwarded greedily until a local maximum is reached.
Afterwards, the message is scheduled to be forwarded to the vehicle with the highest delivery probability.
•The GeoSpray routing protocol [51] combines the S&W multi-copy scheme with the GeOpps protocol.
Similarly to S&W, Lcopies of every message are distributed through the network. Then, instead of waiting
until carriers arrive to the destination, the copies are propagated using GeOpps.
•The DTFR protocol [65] forwards a message to the destination greedily. However, the target destination differs
from the actual destination of the message. The target destination depends on the phase of the protocol and
changes step-by-step, combining phases similar to S&W with pure Greedy-DTN phases. If, at any time during
any of the phases, a vehicle finds a path to the destination, it uses that path to deliver the message.
S
D
Dest Reg 1
Dest Reg 2
A
BC
Fig. 13. RENA: To send a message from Sto D, two copies will be sent to Aand B, which will later distribute them in the destination
regions.
•The RENA protocol [66] combines the Spray & Wait and Epidemic protocols. It divides the map into
regions and calculates the probability of moving between them for every vehicle. Additionally, it estimates the
probability of being inside a given region for every node. Then, the routing process is divided in four phases.
20
When a message is generated, i) it distributes ncopies to vehicles that will probably travel to regions where
the destination vehicle is likely to be located, ii) those copies are forwarded to vehicles that have a better
probability of reaching the destination region than the current carrier, iii) once the message has arrived at the
destination regions, mcopies are distributed to vehicles with a low probability of leaving the region, and iv)
these copies are forwarded to vehicles with a smaller return time to the destination region until the destination
is found. The main advantage of RENA, when compared with other replication mechanisms, is that RENA
limits the replication of the messages to the destination regions. Figure 13 illustrates this behavior.
•Finally, RWR [67] combines pure multi-hop geo-routing with an alternative where messages are delegated
to an RSU. It estimates the expected delay of a message using GPCR and using RSU delegation to choose
between these two alternatives.
The protocols included in this section have the advantages and some of the disadvantages of the combined
protocols. For example, GeoDTN+Nav benefits from the typical low delay of GPCR for multi-hop routing and
the low message loss ratio of DTN, but consumes more resources than GeOpps; GeoSpray, probably performs
better than its predecessors, the original GeOpps protocol and BS&W, at the cost of consuming more resources;
and RENA is clearly expected to waste fewer resources than epidemic routing. However, their implementation is
complicated and depends on many user-defined parameters, which may lead to incompatibilities.
5) Online Protocols: Under the name of Online Protocols we have included protocols that need information on
the current state of the road network, for example, number of nodes, average speed of the nodes, and congestion of
every road. Some of them are also hybrid protocols that combine these new metrics with modifications of protocols
we have previously reviewed. The list of protocols included in this section is: Vehicle-Assisted Data Delivery
(VADD) [9], Static-Node-Assisted Adaptive Data Dissemination in Vehicular Networks (SADV) [68], Distributed
Real-time Data Traffic Statistics Assisted Routing (DRTAR) [69], D-Greedy and D-MinCost [70], the protocol
presented in [71], and CAN DELIVER [72].
•In [9], authors presented VADD, which allows vehicles to send messages to an RSU. The routing process in
VADD is divided in four steps; i) it estimates the travel time of a message for each road taking into account
the vehicles density of the road, its length and the duration of traffic lights. Then ii) it calculates the shortest
path to the destination using Dijkstra, iii) it routes messages between road intersections using the Greedy
DTN protocol and finally iv), when a threshold distance to the destination is reached, it routes messages
using GPSR [58]. Every node traversed by the message recalculates steps iand ii. To obtain information on
road density, duration of traffic lights, maximum speed of roads, etc, required in step i, a database containing
this information is preloaded. The authors presented three variations of VADD that differ on how they route
messages inside crossroads. L-VADD routes messages based on location, while D-VADD uses the direction
of the vehicles. H-VADD combines L-VADD and D-VADD, switching from the first to the second when a
loop is detected. The main problem of VADD is that it clearly tends to use the most heavily populated roads,
21
which may congest the network.
•The SADV protocol [68] complements the VADD protocol by installing static nodes at intersections. It routes
packets like VADD, although inside intersections, when no vehicle in the shortest path is found, the message
is stored in static nodes until a vehicle in the shortest path appears. A more general, but similar, architecture,
where the routing protocol between static nodes is not specified, was proposed in [73]. From our point of
view, inside cities, the increase in cost from backbone-disconnected static nodes to fully-connected RSUs is
negligible compared with the deployment cost. Therefore, we believe that once static nodes are deployed, it
is a better option to connect them to the backbone than to simply use them as static relays.
•The DRTAR protocol [69] is similar to VADD, but it uses a distributed data traffic statistics service to obtain
information on road status. In addition, in DRTAR, the shortest path is only calculated by the first node, which
attaches it to the message. The shortest path is then only recalculated when the current carrier cannot find a
neighbor inside the attached shortest path. Other authors have also proposed different distributed data traffic
statistics services [74], which show the feasibility of this approach.
•In [70], authors presented D-Greedy and D-MinCost, two DTN protocols for traffic-monitoring in vehicular
networks. As far as we know, this is the first paper to introduce a routing protocol that does not try to
minimize the delay from source to destination, but minimizes the consumed resources while ensuring that
the collected information meets certain maximum delay requirements. Authors defined two operation modes,
multi-hop forwarding (MF) mode and the DTN mode (DM). During MF mode messages are forwarded using
Greedy-DTN through the shortest path to destination, while in DM mode messages are only forwarded at
intersections to keep them inside the shortest path when the current carrier moves away. The only difference
between D-Greedy and D-MinCost is that, in the former, only local and map layout information is available,
while in the latter the current road status information is also available. Therefore, D-Greedy calculates the
shortest path to destination based solely on road lengths, while D-MinCost also takes into account the road’s
vehicle density, like VADD. Once the shortest path is calculated, both protocols estimate the delay of the
message using MF, as well as DM. Afterwards, it uses the DM as long as its estimated delay is less than the
Time To Live (TTL), switching to MF in all other cases. Since MF mode is much faster than DM, both modes
tend to alternate, thereby minimizing the number of hops to the destination.
•In [71], authors presented a routing protocol for delivering data from RSUs to vehicles. In their protocol
vehicles make requests while they are connected to an RSU. The answers are usually larger than the requests
and, therefore, cannot be downloaded during the period while they are connected. Authors proposed the use
of other vehicles to deliver the answers to the destination vehicles. They assumed that all RSUs are connected
via a backbone network, and that, based on empirical data, contacts between vehicles can be predicted. With
this information, their protocol uses other vehicles as carriers for answers.
•The CAN DELIVER [72] protocol allows for the routing of messages from vehicles to RSUs and vice-
22
versa. In the former case, the vehicle calculates the shortest road-path to the RSU and attaches it, together
with information on its own route and speed to the message. Then, the message is forwarded between the
intersections using the Greedy DTN protocol. In the latter case, RSUs try to estimates the vehicle’s location
using the information from the vehicle previously attached to the message. Once the future location is estimated,
an area around it is defined, and the reply message is forwarded to it using a scheme that combines the S&W
multi-copy scheme with the Greedy-DTN forwarding metric. When the message reaches the estimated area,
vehicles switch to a limited epidemic mode and broadcast the message inside that area. To avoid broadcast
storms, vehicles only broadcast each message once. If a vehicle outside the estimated area receives a message
from inside it, it must be dropped.
These protocols require a complex platform formed by RSUs, information servers, databases, etc., increasing the
implementation and deployment cost. Moreover, they depend on real-time information, which is easily available in
simulations but can be difficult to obtain in real implementations.
D. Common Basic Mechanisms
Our DTN protocol taxonomy is based on the criteria used to select the next forwarding node, also called the
routing metric. However, this is not the only element that can make a difference in the performance of different
protocols. A set of mechanisms that define the hop-by-hop and the end-to-end communication schemes can heavily
influence the delivery ratio, the delay or other performance metrics. Generally, these mechanisms can be applied to
any utility based protocol. In this section we cover the most representative mechanisms available in the bibliography
addressing: reliability, redundancy, path diversity and message priority. We introduce these concepts, provide some
examples of protocols that use them and measure their impact on performance.
1) Reliability: The reliability of a protocol is the degree of guarantee that the protocol provides to the sender
with respect to the delivery of messages. The typical mechanism used to provide end-to-end reliability in non-DTN
networks is the use of ACK messages to confirm that messages are correctly received. VDTN protocols use hop-by-
hop reliable mechanisms. By using hop-by-hop ACK messages, the protocols ensure that a message will be kept in
the buffer of the vehicle until another vehicle confirms its reception. This mechanism does not explicitly ensure the
reception of the message by its destination, but it does ensure that a message will eventually reach its destination
if no node failure occurs (node shutdown or buffer overflow). Most of the protocols covered in this survey simply
ignore the impact of reliability. Those that consider and use it are: DTFR [65], D-Greedy and D-MinCost [70],
CAN DELIVER [72] and MSDP [60].
The impact of hop-by-hop ACKs increases with the number of hops. For example, if a message traverses 6
hops and the Packet Error Rate (PER) is 10−1(quite optimistic in wireless communications [75]), the end-to-end
PER would be 1−(1 −10−1)6= 0.46, which is an unacceptable value. Since the PER increases with the distance
between transmitters, protocols that tend to select the furthest node as the forwarding node, face higher transmission
23
losses and can heavily benefit from the use of hop-by-hop ACKs. Obviously, the use of ACKs increases both the
load of the channel and the delay experienced by the messages but, from our point of view, it is a small price to
pay compared with its advantages.
Since there is not a specific destination in dissemination protocols, the concept of reliability changes. In dissem-
ination protocols we consider reliability as the capability of the protocol to guarantee that at least one of the nodes
inside the ROI will disseminate the message until it expires. This feature is usually implemented as follows: i) the
current carrier broadcasts the message, ii) after broadcasting the carrier keeps sniffing the channel to check if a
neighbor has rebroadcasted it, iii) the transmission confirmation is implicit when a neighbor has rebroadcasted the
message. All of the dissemination protocols included in this survey implement this mechanism.
2) Fragmentation and Redundancy: The objective of fragmentation is to provide flexibility to routing. In VDTN,
the duration of the contacts limits the amount of data that two nodes can exchange. When a connection between
two nodes breaks, the message being transmitted has to be discarded by the receiver and enqueued again by the
transmitter, thus wasting the resources used for that transmission to date. In the case of messages of a large size,
the amount of wasted resources can be high.
The use of fragmentation allows for redundancy fragments to be added using Forward Error Correction (FEC)
techniques. This means that, if a message needs Nfragments, N∗αfragments will be sent, where the redundancy
factor alpha is greater than 1 and depends on the configuration. At the destination, only Nfragments are needed
to reassemble the original message. This type of redundancy is usually called coding, and it reduces the impact of
possible losses. The cost of coding depends on the amount of extra fragments sent. Fragmentation and coding only
appear in two of the protocols we have reviewed, which are CAN DELIVER [72] and MSDP [60].
A more aggressive type of redundancy consists of sending multiple copies of the same message. This mechanism
is much simpler than coding, but it also consumes more resources. Moreover, it does not solve the problems
arising from large-sized messages. This redundancy mechanism is much more common and is used in the following
protocols: Epidemic [24], PRoPHET [25], Spray&Wait [44], MaxProp [30], RAPID [47], DAER [49], POR [48],
ADPBSW [50], DTFR [65], GeoSpray [51], RENA [66] and CAN DELIVER [72].
3) Message Priority: By message priority we refer to the order in which messages are forwarded to another node
when a contact occurs. This is important, as the duration of contacts is limited. In the bibliography, some protocols
have extended the Epidemic protocol to consider message priority: MaxProp [30] and RAPID [47] prioritize those
messages with a better transmission delivery probability according to PRoPHET, while POR [48] and DAER [49]
prioritize those messages that will get closer to their destination. Although we were unable to find more examples
of this mechanism, it may be implemented to complement and improve the performance of any protocol.
IV. EVALUATING DTN PROT OC OL PERFORMANCE IN VANETS
Since developing and conducting real implementation and tests for VNs is an expensive task in terms of time,
personnel and money, researchers have focused on simulations to evaluate and compare the performance of different
24
protocols. However, on analizing the reviewed articles, we have found a balanced mix of different simulation models
that complicates the comparison of results. Moreover, very rarely do works evaluate the same metrics under the
same scenarios, which totally invalidates any comparison among results from different papers.
Table II summarizes the contents of this section. The second column shows the different metrics measured
during the evaluation of each proposal. The third column specifies the simulator used for this evaluation. The fourth
and fifth columns contain the MAC and radio channel models they used. The sixth column briefly describes the
simulated scenario. Finally, the last column shows the number of DTN protocols compared to justify every new
proposal. As stated in Table II, we found that most researchers did not compare their proposal against any other
DTN protocol (14 out of 41 papers) and that a large group of researchers compared their proposal against only
one previously proposed protocol (12 out of 41 papers). This unfortunate situation is a consequence of the mix of
available simulation models, as well as the commonly vague description on low-level protocol details, as already
explained in Section III-D. Moreover, researchers do not usually offer the source code of their proposals, which
complicates the replication and validation of their experiments.
In this section, we first list the metrics evaluated by researchers discussing their relevance. Second, we provide
an overview of the models and tools used by the research community to evaluate VDTN protocols and identify the
most advanced solutions.
A. Evaluated Metrics
When introducing a new proposal, researchers need to justify the performance improvement by comparing metrics
among different protocols. We have found that the most commonly evaluated metrics are:
•The Delivery Ratio (DR), which is given by the ratio of the number of successfully received messages and the
number of sent messages. Since delivering messages to their destination is the task of a routing protocol, the
DR is the most important metric when evaluating such a protocol. However, researchers must find a trade-off
between resource consumption and effectiveness.
•The Average Delay (AD), which is given by the average time needed to deliver a message. In DTNs, this
metric may be heavily influenced by a small number of high delay measurements and, therefore, its value is
not representative of the general behavior of a protocol.
•The Delay Cumulative Distribution Function (DC), which illustrates the distribution of the delay experienced
by messages. Since the average delay is heavily influenced by messages experiencing long delays, this
measurement gives a better idea of the performance of a protocol.
•The Overhead (O), which measures the amount of extra bytes needed per delivered byte. This is a very
important metric when evaluating VDTN protocols because part of the network may become easily saturated.
•The Average Number of Hops (H) traversed by a message. This measurement provides an idea of resource
consumption. As a general rule, more hops means more consumed resources. However, fewer hops usually
25
TABLE II
EVALUATI ON OF DI FFE REN T PROTO CO LS.
Measurements Simulator Medium Access Radio & Channel Simulation Scenario #Compared
Epidemic [24] DR AD DC Ns2 No, only contacts Fix Distance Random 0
ProPHET [25] No evaluation
MoVe [57] DR AC DC O Custom No Data No Data Limited random in city map
and traces
1
Spawn [76] O H Nab CSMA/CA Model No Data One Direction Highway 0
Spray&Wait [44] DR AD O Custom Sloted Collision Detection No Data Random Way Point 2
MaxProp [30] DR AC H Custom No, only contacts No Data Real & Synthetic Traces 0
RAPID [47] DR AD O Custom No, only contacts Fix Distance Real Traces, Contact Model 1
SimBet [55] DR DC O H Custom No, only contacts Contacts Traces Real Traces 2
GeOpps [10] DR AD DC O H Omnet++, MF CSMA/CA Model Nakagami, No Obst, Inter-
ferences
Synthetic Realistic Traces 2
Direct [77] DR AD DC Custom No Data Fix Distance RandomWay Point in Grid
& Implemented
0
POR [48] DR AD DC Custom No, only contacts Fix Distance Real Traces 4
DAER [49] DR DC H Custom No, only contacts Fix Distance Real Traces, SUMO 0
VADD [9] DR AD O Ns2 CSMA/CA Model Fix Distance, Interferences Limited Random in city
map
1
DSCF [35] DC Custom No, only contacts Fix Distance Grid Map, Random mobil-
ity
0
FFRDV [36] DC Ns2 No, Only contacts Fix Distance Highway, Car Following
Model
1
Infocast [34] DR Ns2 No Data Fix Distance Only 1 road 0
ADPBSW [50] DR AD ONE No, Only contacts Fix Distance Limited random in city map 1
Adv. ProPHET [53] DR AD ONE No, Only contacts Fix Distance Limited random in city map 1
Extended GeOpps [61] No Evaluation
C-DTN [78] DR AD QualNet CSMA/CA Model Fix Distance, Interferences Car Following Model in a
Grid
0
DvCast [39] DR DC O Ns2 No Data Ricean Fading Circular High Way 0
ROD [37] DR Airplug-ns CSMA/CA Model Fix Distance, Interferences VehicleMobiGen 2
Uv-Cast [38] DR O Ns2 CSMA/CA Model Fix Distance, Interferences Real City, SUMO 0
ProPHET+ [52] DR DC ONE No, Only contacts Contacts Traces Real Traces 1
DRTAR [69] AD Custom CSMA/CA Model Fix Distance, Interferences Limited random in city map 0
GeoDTN+NAV [63] DR AD H QualNet CSMA/CA Model Fix Distance, Interferences VanetMobiSim 0
[62] DR O Custom CSMA/CA Model Nakagami, No Obst, Inter-
ferences
NETSTREAM 1
SADV [68] AD O MatLab No, Only contacts Fix Distance Limited random in city map 1
D-Greedy D-MinCost [70] DR AD DC O Custom No, Only contacts Fix Distance Synthetic Realistic Traces 2
SERVUS [40] DR O Ns2 CSMA/CA Model Fix Distance, Interferences Grid, Random 0
DTFR [65] DR AD Custom Slotted mac Fix Distance, Interferences Limited random in city map
and traces
3
Orion [64] DR AD H Omnet++ No Data No Data Random 1
RENA [66] DR AD ONE No, Only contacts Fix Distance Limited random in city map
and traces
4
GeoSpray [51] DR AD O ONE* No, Only contacts Fix Distance Limited random in city map 4
DSRelay [79] DR Custom No, Only contacts Fix Distance Highway, Random Speed 1
[71] DR OM Custom Slotted mac Fix Distance Synthetic Realistic Traces 0
CAN DELIVER [72] DR AD DC O Ns2 CSMA/CA Model Nakagami, No Obst, Inter-
ferences
Real Map, SUMO 3
RWR [67] DR AD DC Custom No, Only contacts Fix Distance Real Traces 2
CSM [80] Error Estimation Custom No, Only contacts Fix Distance Real Traces 1
MSDP [60] DR AD DC O Omnet++,INET CSMA/CA Model Nakagami, W. Obst, Inter-
ferences
Real Map, SUMO 2
ZOOM [54] DR AD O Custom No, Only contacts Contacts Traces Real Traces 3
DR=Delivery Ratio; AD=Average Delay; DC=DelayCDF; O=Overhead; H=Number of Hops
26
implies longer carrying phases, increasing the average delay of the messages.
Table II includes a column that shows the different metrics evaluated in each paper.
B. Simulators and Models
The choice of a certain simulator does not influence the results of simulation studies, but it commonly implies
the use of a certain set of models and default values. Through the reviewed papers, we clearly identify a worrying
trend: 18 works out of 41 used a custom simulator. The use of a custom simulator complicates or almost prevents
proper comparison among different proposals. Moreover, it also complicates the peer reviewing system and code
reutilization, slowing the developing pace. On the other hand, we have found four different event-driven simulators
that have been previously validated and are long-established in the networking community: Ns2 (8 times), The ONE
(5 times), OMNeT++ (3 times) and Qualnet (2 times). Below we briefly describe the characteristics of different
simulators.
•The Ns2 simulator integrates advanced propagation and channel models (Nakagami fading and shared channel),
medium access (CSMA/CA) and mobility models (traces generated using SUMO) [81]. However, only one
of the reviewed proposals used the most advanced features of Ns2 [72]. Three of the articles that used Ns2
neglect the effects of propagation and interferences, while remaining articles used a deterministic propagation
model combined with an interference model.
•The ONE is a contact-oriented simulator[82]. As far as we know, it is the only simulator specifically designed
for DTN, speeding up the development and implementation of new protocols. At present, it does not support
propagation or channel models and the mobility model is limited to map-constrained random mobility or
real traces, although it is easily extendable. Due to its simplicity, The ONE is significantly faster than other
simulators. We would recommend it for early research stages, to evaluate the logic of different proposals and
to test whether they have major drawbacks, such as local minimums where messages get stuck. We believe
that The ONE may be easily extended to implement car following mobility models and a non-deterministic
propagation model.
•Veins [84], for OMNeT++ [83], is currently the most advanced simulation framework for VN simulation. It
implements a complex propagation and interference model and a fully featured medium-access model based
on the 802.11p standard, with support for advanced driving models provided by SUMO. However, none of
the reviewed works used this framework. In [60], authors used the INET framework [85], whose medium
access model is limited to 802.11a/b/g. In [10], authors used a framework that was later integrated in the
Inet framework. Because of the fine-grain simulation provided by OMNeT++ and Veins, it consumes a lot of
resources in terms of memory, CPU and time, making unaffordable simulations with thousands of nodes.
•QualNet is a non-opensource simulator. Therefore, the correctness of its models cannot be verified. It imple-
ments a 802.11 medium-access model and a complex propagation model, as well as an interference channel
27
model. It supports the use of trace-based mobility models, which can be obtained from mobility generators
such as SUMO or VanetMobiSim. The models implemented in QualNet are less advanced than the ones
implemented in Veins.
When simulating network protocols, models are more important than the simulators [86], [87], [49]. In the
following subsections, we go through the models used by researchers to evaluate their proposals. The following
subsections does not seek to be a survey on Inter-Vehicle Communication (IVC) simulation models, which can be
found in [88].
C. Low level models
Radio propagation models for VN must reflect the effects of path loss, shadowing, and multipath fading. The path
loss defines the average received power at certain distance from the transmitter, while shadowing and multipath
fading add a random component related to obstacles between the transmitter and the receiver, and the multiple
delayed replicas of the signal received. A more extended discussion of these effects is not included in the scope of
this survey, and can be found in [89].
Only considering the effects of path loss results in a deterministic propagation distance, which is far from a realistic
scenario, we have found that 26 out of 41 reviewed papers use a deterministic propagation model. Considering a
deterministic communication range between neighbors has overly optimistic effects on the performance evaluation
of the protocols.
More recent works have incorporated the effects of fading into their propagation models [10], [39], [62], [72],
which is closer to propagation behavior in real environments. However, only one of the reviewed papers [60]
considers the effects of buildings and obstacles when simulating urban scenarios.
In terms of interference models, we have found that only 8 works considered the effects of interference between
neighbor nodes.
In this survey we have found that some papers (5 out of 41) ignore or do not specify the radio propagation and
channel models used. We firmly believe that the VDTN research community should make an effort to improve the
quality of the propagation and channel model used to evaluate protocols.
Besides the propagation and channel model, as shown in [90], it is also important to use a fully featured IEEE
802.11p model. However, none of the reviewed papers used such an advanced model. The most advanced models
were limited to a CSMA/CA model, used in 12 out of 41 papers, while 3 papers used a simplified slotted MAC.
As a negative trend, 19 of the 41 reviewed papers ignore the necessity of a medium-access model, and assume that
nodes within the communication range can always communicate. This assumption only holds true in very sparse
networks, where the probability of interfering neighbors is negligible. Moreover, 5 papers did not define the MAC
model they used, which clearly compromises the reproducibility of their simulations.
28
Although the medium-access model may seem less important than the propagation and channel models, from our
point of view, the minimum required medium-access model is a slotted mac, where only a connection between 2
nodes in a certain area can be established. It is clear that researchers must improve the average detail of medium-
access models used in VDTNs.
D. Mobility Model and Simulated Scenario
Authors such as Joerer et al. [88] have shown their concerns about mobility model specifications in VN.
Fortunately, only two of the reviewed papers used the Random Way Point mobility model [91]. The majority
of the papers used a limited random mobility model, i.e. nodes move randomly but their movements are limited by
the road network topology. This model is better than pure random mobility, but it does not capture the characteristics
of vehicular mobility; for example, two vehicles may occupy the same location at the same time. We also found a
group of papers that implemented their own car following mobility model [36], [78]. Given the complexity of the
models, a self-implemented car following model also compromises the reproducibility of the experiments. Finally,
in only 8 papers, we found what we consider the best practice: the use of a validated micro mobility simulator. In
the papers we reviewed, researchers used VanetMobiSim [92], SUMO [93] and NETSTREAM [94] as the mobility
generator. SUMO is the most advanced mobility simulator, implementing a car following model and real maps and
enabling researchers to run mobility and network simulation concurrently, thus allowing events in the network to
influence the mobility of the nodes. It is also worth noticing that 11 of the reviewed articles used traces obtained
from real vehicles to simulate the mobility of the nodes. Real traces are a good option but they lack flexibility
when varying network parameters such as number of nodes, road topology, etc.
Concerning the simulated scenario, it is important to evaluate VDTN protocols in both city and highway scenarios.
We found that only 3 papers considered the highway scenario, while 22 used the city scenario. Inside the city
scenario there is a huge variety of configurations ranging from urban grids to low-building-density suburban areas.
Once again, this diversity complicates the comparison of different proposals. The mobility model and the simulated
scenario can significally affect the performance of protocols, especially VDTN protocols, where nodes tend to carry
information in buffers and protocols tend to make decisions based on node mobility.
Table II summarizes our findings when analyzing the tools and models used by researchers. As previously
explained, the diversity of models and simulators makes it impossible to compare different proposals without re-
implementing every proposal.
E. Testbeds and Implementations
Over recent years, some researchers have pointed out the need for real tests prior to VN deployment [95]. Within
the set of papers reviewed in this survey only [96], [77] and [30] test their proposals in a real environment. In
[96], authors run a test of the Cartorrent system, which is based on the Spawn protocol. In [77], authors extended
29
TABLE III
GRA DE OF SUITABILITY OF PROTO COL S TO DI FFER EN T APLICATIONS
Application
Group Zero Knowledge Contacts History & Social Geographic Location Road Map Online
P2P & I2V 1 5 2 3 4
V2I & Sensors Collecting 1 2 3 4 5
Cooperative Downloads 1 5 4 3 5
Dissemination 3 5 4 5 3
less suitable 1-5 more suitable
the Controller Area Network (CAN) bus of vehicles to send its data to a base station using the DTN reference
implementation [97]. In [30], authors used a testbed formed by buses inside the University of Massachusetts called
UMassDieselNet.
Others authors have presented their testbed for VNs where VDTN protocols could easily be tested. In [98], authors
presented Cabernet, a VN deployed over 10 taxis of Boston area. In 2010, researchers from UCLA presented C-VeT,
an advanced testbed for vehicular networking and urban sensing, which combined a VANET formed by management
vehicles and buses with a mesh network based on Open-WRT. In [99], an implementation of a warning protocol for
VDTNs was presented and tested. In [100], authors presented a Creative Testbed that combines simulations with
testbed results to maximize flexibility while minimizing deployment cost.
We clearly identify a positive trend towards more advanced testbeds, closer to real deployment. We would like
to support and encourage researchers to use these new testbeds whenever possible, since such initiatives are vital
for promoting the full deployment of VDTNs.
V. D TN BASED APPLICATIONS IN VEHICULAR NETWORKS
In this section we introduce applications proposed by the research community that depend on the use of DTNs.
We describe them presenting some of the problems and challenges they must face. We start this classification with
the most frequent application in the reviewed articles, Peer-to-Peer (P2P) communication. Secondly we present what
we call environment-sensing applications, which consider the use of DTN protocols in order to collect information
using vehicles as sensors. The third group includes dissemination applications; beside broadcast dissemination, we
also consider context-based dissemination. Finally, we explore collaborative content-downloading applications
and new proposals such as cellular offloading. Table I classifies each of the protocols analyzed previously in each
of these categories. For each application described, we provide some examples of its utilization and discuss which
group of protocols best adapts to it.
Table III quantifies the suitability of each group of protocols for each application according to the criteria explained
in this section.
30
A. P2P Applications
The most obvious application of any communications system involves allowing users to exchange messages and
information between them. Hence, it is not surprising that the majority of the analyzed articles focus on “P2P”
communication.
As stated in previous sections, when using geographic protocols for P2P communication we need a Location
Service to obtain the location of the destination of a message. Table I shows that 22 of 41 works are labeled as “P2P”
or “P2P/V2I”. The second label includes protocols that are presented as a “P2P” protocol, but obviate the complexity
of the required location service, which makes the communication between vehicles impossible, thereby reducing
them to V2I communication protocols. We have grouped V2I applications with environment-sensing applications,
due to their similarities.
The typical example of a P2P application is a kind of e-mail system, where users can exchange personal messages.
Obviously, this scenario application assumes that the sender and the receiver have met previously. We can also
assume that the number of users of this application is relatively small (dozens of individuals), compared to the
number of vehicles that typically form a VN (thousands of nodes). Given these assumptions, we believe that contact
rate and social relationship-based protocols are the best alternative for this application.
If the cost of infrastructure deployment is affordable, it is probably a better option to deploy a set of RSUs
connected by a backbone network and then use them to slice the source-to-destination routing problem into two
smaller problems: routing from source to an RSU, and routing from another RSU to the destination. This scheme
is similar to the one described in [72].
B. V2I and Sensing Applications
In V2I applications the objective is to send information from a vehicle to an RSU. In environment-sensing
applications, the main objective is the same, but it can be assumed that the information is typically correlated to
the geographic location of the source.
An example of a V2I application is the scenario where a user wants to order a large number of goods in a shop.
Using VDTNs, the user can send a message to the shop, which will be able to prepare the order in advance. In the
second case, we envision a scenario where traffic management and road security authorities collect information on
speed, road status or weather from vehicles. This information can be used to optimize emergency vehicle routes,
monitor pollution inside cities, plan taxi routes, etc.
Since RSUs have a fixed location that can be stored in a quasi-static database, geographic, road map, and online
protocols do not require a location service to route messages to its destination. This feature is used in protocols such
as GeOpps [10], GeoDTN+Nav [63] or MSDP [60]. In [80], authors introduce a new scheme where messages from
different nearby sources are combined to compress their information and reduce the channel load. As stated before,
one of the key issues of zero knowledge protocols is that node mobility increases the probability of reaching the
31
destination of a message. Since RSUs are static, zero knowledge protocols are not suitable for these applications. A
similar problem applies to contact history and social based protocols. Since they require nearly-closed communities,
they tend to ignore nodes that pass by a region.
C. Dissemination
Dissemination applications aim to quickly deliver information to as many nodes as possible. In this scope,
the adoption of delay-tolerant protocols may seem counter-intuitive since the expectable delay is rather high.
Nevertheless, in sparse networks where the degree of node connectivity is low, the store-carry-and-forwarding
paradigm may be the only method capable of guaranteeing a high message delivery ratio.
Accidents occurring on highways represent a typical scenario where quick message dissemination may be useful,
for example by notifying drivers approaching the accident area and thereby avoiding cascading car crashes.
When disseminating information, an ROI where a message must be disseminated is typically defined. The ROI
is usually related to geographic or road network restrictions, being mostly useful to vehicles moving towards an
accident, vehicles moving on streets adjacent to a traffic jam or vehicles ahead of an ambulance route, for example.
The strong relationship between the ROI and the actual characteristics of the road environment makes geographic
and map based protocols the most suitable alternatives for this application.
Other cases, for example when disseminating non-geographically correlated information (e.g. advertisements),
the best socially-connected nodes would probably be the best carriers.
D. Cooperative Download
In cooperative download applications, the main data flow occurs from RSUs to vehicles. Typically, a user requests
data that is too large to be transferred during a single contact with an RSU. To solve this problem RSUs which are
connected to a backbone inject fragments of the responses into the network. Once fragments are injected, there are
two main alternatives: to distribute these fragments between every interested node [76] or to deliver them only to
its specific destination [71], [61], [79].
When distributing fragments to every interested node, it is usually easy to identify social relationships between
interested nodes and this information can be used to maximize the protocol performance. On the other hand,
geographic information can also be useful to select the best contact, as in [76].
When delivering a message to its specific destination, it can be seen as a P2P communication between an RSU
and a mobile node and, therefore, we can apply the same methods as for P2P applications.
VI. CONCLUSIONS
In this survey, we provided the reader with a broad view of the different proposals for VDTNs. We classified
them according to their utility index, showing the relationships between different protocols and their evolution. We
32
identified a set of common mechanisms that can be applicable to almost all VDTN protocols, and that may heavily
influence their performance. We also presented some applications where VDTNs can be used and evaluated the
suitability of the different proposals for each application.
Moreover, this survey is not limited to a mere description of protocols, since it also addresses critical issues such
as the reproducibility and repeatability of experiments and reviews the evaluation methods used by the different
VDTN researchers. We pointed out a lack of realism in most of the simulation models used by the VDTN research
community.
Tables I and II summarize the contents of this survey, and offer important information at a glance.
Based on the extensive survey presented in this paper, we can conclude that no VDTN protocol is suitable for all
possible target applications. From our point of view, researchers must focus on providing services/applications,
and VDTN protocols should be flexible enough to adapt their behavior to the characteristics of the running
application. We also believe that routing metrics should adapt to current network characteristics by making the
most of opportunistic contacts or taking advantage of vehicular mobility according to available resources.
ACK NOW LE DG ME NT S
This work was partially supported by the Ministerio de Econom´
ıa y Competitividad, Spain, under Grants TIN2011-
27543-C03-01 and BES-2012-052673.
REFERENCES
[1] G. V. Kumar, Y. V. Reddyr, and M. Nagendra, “Current Research Work on Routing Protocols for MANET : A Literature Survey,” (IJCSE)
International Journal on Computer Science and Engineering, vol. 02, no. 03, pp. 706–713, 2010.
[2] V. Cerf, S. Burleigh, A. Hooke, L. Torgerson, R. Durst, K. Scott, K. Fall, and H. Weiss, “Delay-Tolerant Networking Architecture,”
RFC 4838 (Informational), Internet Engineering Task Force, Apr. 2007. [Online]. Available: http://www.ietf.org/rfc/rfc4838.txt
[3] K. Fall, “A delay-tolerant network architecture for challenged internets,” in Proceedings of the 2003 conference on Applications,
technologies, architectures, and protocols for computer communications - SIGCOMM ’03, ser. SIGCOMM ’03. New York, New
York, USA: ACM Press, 2003, p. 27.
[4] J. Kolodziej, S. U. Khan, L. Wang, N. Min-Allah, S. A. Madani, N. Ghani, and H. Li, “An Application of Markov Jump Process Model
for Activity-Based Indoor Mobility Prediction in Wireless Networks,” in Frontiers of Information Technology (FIT), 2011, Dec. 2011,
pp. 51–56.
[5] “Delay tolerant networks research group,” http://www.dtnrg.org/wiki/Home, Nov 2013.
[6] S. Jain, K. Fall, and R. Patra, “Routing in a delay tolerant network,” in Proceedings of the 2004 conference on Applications, technologies,
architectures, and protocols for computer communications - SIGCOMM ’04, New York, USA, 2004, p. 145.
[7] Z. Zhang, “Routing in intermittently connected mobile ad hoc networks and delay tolerant networks: overview and challenges,” IEEE
Communications Surveys & Tutorials, vol. 8, no. 1, pp. 24–37, Mar. 2006.
[8] B. Paul, M. Ibrahim, M. Bikas, and A. Naser, “VANET Routing Protocols: Pros and Cons,” International Journal of Computer Applications,
vol. 20, no. 3, pp. 28–34, 2011.
[9] J. Zhao and G. Cao, “VADD: Vehicle-assisted data delivery in vehicular ad hoc networks,” IEEE Transactions on Vehicular Technology,
vol. 57, no. 3, pp. 1910–1922, 2008.
[10] I. Leontiadis and C. Mascolo, “GeOpps: Geographical Opportunistic Routing for Vehicular Networks,” in World of Wireless, Mobile and
Multimedia Networks, 2007. WoWMoM 2007. IEEE International Symposium on a, Helsinki, Finland, Jun. 2007, pp. 1–6.
33
[11] M. Khabbaz, C. Assi, and W. Fawaz, “Disruption-tolerant networking: A comprehensive survey on recent developments and persisting
challenges,” IEEE Communications Surveys & Tutorials, vol. 14, no. 2, pp. 607 – 640, 2012.
[12] S. M. Bilal, C. J. Bernardos, and C. Guerrero, “Position-based routing in vehicular networks: A survey,” Journal of Network and Computer
Applications, vol. 36, no. 2, pp. 685–697, 2013.
[13] N. Benamar, M. Benamar, and J. M. Bonnin, “Routing protocols for DTN in vehicular environment,” 2012 International Conference on
Multimedia Computing and Systems, pp. 589–593, May 2012.
[14] Y. Cao and Z. Sun, “Routing in Delay/Disruption Tolerant Networks: A Taxonomy, Survey and Challenges,” IEEE Communications
Surveys & Tutorials, vol. 15, no. 2, pp. 654–677, 2013.
[15] F. Warthman, “Delay tolerant networking- a tutorial,” 2003.
[16] “Interplanetary networks project,” http://www.ipnsig.org, Nov 2013.
[17] S. Burleigh, A. Hooke, and L. Torgerson, “Delay-tolerant networking: an approach to interplanetary internet,” Communications Magazine,
IEEE, vol. 41, no. 6, pp. 128–136, 2003.
[18] P. Juang, H. Oki, Y. Wang, and M. Martonosi, “Energy-efficient computing for wildlife tracking: Design tradeoffs and early experiences
with ZebraNet,” SIGPLAN Not., vol. 37, no. 10, pp. 96–107, 2002.
[19] “N4c project,” http://www.n4c.eu, Nov 2013.
[20] R. Shah, S. Roy, S. Jain, and W. Brunette, “Data MULEs: modeling a three-tier architecture for sparse sensor networks,” in Proceedings
of the First IEEE International Workshop on Sensor Network Protocols and Applications, 2003. Ieee, 2003, pp. 30–41.
[21] K. Scott and S. Burleigh, “Bundle Protocol Specification,” RFC 5050 (Experimental), Internet Engineering Task Force, Nov. 2007.
[Online]. Available: http://www.ietf.org/rfc/rfc5050.txt
[22] M. Demmer, J. Ott, and S. Perreault, “Delay Tolerant Networking TCP Convergence Layer Protocol,” RFC 7242, Internet Research
Task Force, June 2014. [Online]. Available: https://tools.ietf.org/rfc/rfc7242.txt
[23] A. Lindgren, E. Davies, and A. Doria, “Generic Opportunistic Routing Framework,” Internet Draft (Experimental), Internet Engineering
Task Force, Jul. 2013. [Online]. Available: https://tools.ietf.org/id/draft-lindgren-dtnrg-gorf-epidemic-00.txt
[24] A. Vahdat and D. Becker, “Epidemic Routing for Partially Connected Ad Hoc Networks,” in Technical Report CS-200006, vol. CS-200006,
no. CS-200006. Duke University, Apr. 2000, pp. CS–2000–06.
[25] A. Lindgren, A. Doria, O. Schel´
en, and O. Schelen, “Probabilistic Routing in Intermittently Connected Networks,” SIGMOBILE Mob.
Comput. Commun. Rev., vol. 7, no. 3, pp. 19–20, Jul. 2003.
[26] A. Lindgren, E. Davies, and A. Doria, “Epidemic Routing Module for Generic Opportunistic Routing Framework,” Internet Draft
(Experimental), Internet Engineering Task Force, Jul. 2013. [Online]. Available: http://www.ietf.org/id/draft-lindgren-dtnrg-gorf-epidemic-
00.txt
[27] M. Yevstifeyev, “Moving RFC 4693 to Historic,” RFC 6393 (Informational), Internet Engineering Task Force, Sep. 2011. [Online].
Available: http://www.ietf.org/rfc/rfc6393.txt
[28] D. Jiang and L. Delgrossi, “IEEE 802.11p: Towards an International Standard for Wireless Access in Vehicular Environments,” in VTC
Spring 2008 IEEE Vehicular Technology Conference. Marina Bay, Singapore: Ieee, 2008, pp. 2036–2040.
[29] “IEEE Standard for Information technology– Local and metropolitan area networks– Specific requirements– Part 11: Wireless LAN
Medium Access Control (MAC) and Physical Layer (PHY) Specifications Amendment 6: Wireless Access in Vehicular Environments,”
IEEE Std 802.11p-2010 (Amendment to IEEE Std 802.11-2007 as amended by IEEE Std 802.11k-2008, IEEE Std 802.11r-2008, IEEE
Std 802.11y-2008, IEEE Std 802.11n-2009, and IEEE Std 802.11w-2009), pp. 1–51, Jul. 2010.
[30] J. Burgess, B. Gallagher, D. Jensen, and B. N. Levine, “MaxProp: Routing for Vehicle-Based Disruption-Tolerant Networks,” in
Proceedings IEEE INFOCOM 2006. 25TH IEEE International Conference on Computer Communications, vol. 6, no. c, Barcelona,
Spain, Apr. 2006, pp. 1–11.
[31] K. Obraczka, K. Viswanath, and G. Tsudik, “Flooding for reliable multicast in multi-hop ad hoc networks,” Wireless Networks, vol. 7,
no. 6, pp. 627–634, 2001.
[32] S.-Y. Ni, Y.-C. Tseng, Y.-S. Chen, and J.-P. Sheu, “The broadcast storm problem in a mobile ad hoc network,” Proceedings of the 5th
annual ACM/IEEE international conference on Mobile computing and networking - MobiCom ’99, pp. 151–162, 1999.
34
[33] Y. Tseng, S. Ni, and E. Shih, “Adaptive approaches to relieving broadcast storms in a wireless multihop mobile ad hoc network,”
Computers, IEEE Transactions on, vol. 52, no. 5, pp. 545–557, 2003.
[34] M. Sardari, F. Hendessi, and F. Fekri, “Infocast: A new paradigm for collaborative content distribution from roadside units to vehicular
networks,” in Sensor, Mesh and Ad Hoc Communications and Networks, 2009. SECON ’09. 6th Annual IEEE Communications Society
Conference on, no. Section III, 2009.
[35] S. Kuribayashi, Y. Sakumoto, S. Hasegawa, H. Ohsaki, and M. Imase, “Performance Evaluation of Broadcast Communication Protocol
DSCF (Directional Store-Carry-Forward) for VANETs with Two-Dimensional Road Model,” 2009 10th International Symposium on
Pervasive Systems Algorithms and Networks, pp. 615–619, 2009.
[36] D. Yu and Y.-B. Ko, “FFRDV: fastest-ferry routing in DTN-enabled vehicular ad hoc networks,” in Advanced Communication Technology,
2009. ICACT 2009. 11th International Conference on, vol. 02, Phoenix Park, Feb. 2009, pp. 1410–1414.
[37] M. Cherif, S.-M. Secouci, and B. Ducourthial, “How to disseminate vehicular data efficiently in both highway and urban environments?”
in Wireless and Mobile Computing, Networking and Communications (WiMob), 2010 IEEE 6th International Conference on, Niagara
Falls, ON, 2010, pp. 165–171.
[38] W. Viriyasitavat, F. Bai, and O. Tonguz, “Uv-cast: An urban vehicular broadcast protocol,” in Vehicular Networking Conference (VNC),
2010 IEEE, vol. 49, no. 11, Jersey City, NJ, 2010, pp. 25–32.
[39] O. Tonguz, N. Wisitpongphan, and B. Fan, “DV-CAST: A distributed vehicular broadcast protocol for vehicular ad hoc networks,” Wireless
Communications, IEEE, vol. 17, no. 2, pp. 47–57, 2010.
[40] R. Gorcitz, P. Spathis, M. Dias de Amorim, R. Wakikawa, and S. Fdida, “SERVUS: Reliable low-cost and disconnection-aware broadcasting
in VANETs,” in Wireless Communications and Mobile Computing Conference (IWCMC), 2011 7th International, Istanbul, 2011, pp. 1760–
1765.
[41] M.-T. Sun, W.-c. Feng, T.-H. Lai, K. Yamada, H. Okada, and K. Fujimura, “GPS-based message broadcast for adaptive inter-vehicle
communications,” in Vehicular Technology Conference, 2000. IEEE-VTS Fall VTC 2000. 52nd, Boston, MA, 2000, pp. 2685–2692 vol.6.
[42] M. Grossglauser and D. N. C. Tse, “Mobility increases the capacity of ad hoc wireless networks,” IEEE/ACM Transactions on Networking,
vol. 10, no. 4, pp. 477–486, 2002.
[43] T. Spyropoulos, K. Psounis, and C. Raghavendra, “Single-copy routing in intermittently connected mobile networks,” in 2004 First Annual
IEEE Communications Society Conference on Sensor and Ad Hoc Communications and Networks, 2004. IEEE SECON 2004., 2004, pp.
235–244.
[44] T. Spyropoulos, K. Psounis, and C. S. Raghavendra, “Spray and wait: an efficient routing scheme for intermittently connected mobile
networks,” in Proceeding of the 2005 ACM SIGCOMM workshop on Delay-tolerant networking - WDTN ’05, ser. WDTN ’05. New
York, New York, USA: ACM Press, 2005, pp. 252–259.
[45] T. Small and Z. J. Haas, “Resource and performance tradeoffs in delay-tolerant wireless networks,” in Proceeding of the 2005 ACM
SIGCOMM Workshop on Delay-tolerant networking - WDTN ’05. New York, New York, USA: ACM Press, 2005, pp. 260–267.
[46] X. Zhang, G. Neglia, J. Kurose, and D. Towsley, “Performance modeling of epidemic routing,” Computer Networks, vol. 51, no. 10, pp.
2867–2891, Jul. 2007.
[47] B. Balasubramanian, Aruna Levine and A. Venkataramani, “DTN routing as a resource allocation problem,” in Proceedings of the 2007
Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications, New York, NY, USA, 2007, pp.
373–384.
[48] X. Li, W. Shu, M. Li, H. Huang, and M.-Y. Wu, “DTN Routing in Vehicular Sensor Networks,” in IEEE GLOBECOM 2008 - 2008 IEEE
Global Telecommunications Conference. New Orleans, LO: Ieee, Nov. 2008, pp. 1–5.
[49] P. Luo, H. Huang, W. Shu, M. Li, and M.-Y. Wu, “Performance Evaluation of Vehicular DTN Routing under Realistic Mobility Models,”
in Wireless Communications and Networking Conference, 2008. WCNC 2008. IEEE, vol. 2, Las Vegas, NV, Mar. 2008, pp. 2206–2211.
[50] J. Xue, X. Fan, Y. Cao, J. Fang, and J. Li, “Spray and Wait Routing Based on Average Delivery Probability in Delay Tolerant Network,”
in 2009 International Conference on Networks Security, Wireless Communications and Trusted Computing, vol. 2, Wuhan, Hubei, Apr.
2009, pp. 500–502.
35
[51] F. Soares, Vasco N. G. J. Rodrigues, Joel J. P. C. Farahmand, “GeoSpray: A geographic routing protocol for vehicular delay-tolerant
networks,” Inf. Fusion, vol. 15, pp. 102–113, Jan. 2011.
[52] T.-K. Huang, C.-K. Lee, and L.-J. Chen, “PRoPHET+: An Adaptive PRoPHET-Based Routing Protocol for Opportunistic Network,” in
2010 24th IEEE International Conference on Advanced Information Networking and Applications, Apr. 2010, pp. 112–119.
[53] J. Xue, J. Li, Y. Cao, and J. Fang, “Advanced prophet routing in delay tolerant network,” in Communication Software and Networks,
2009. ICCSN ’09. International Conference on, no. C, Macau, 2009, pp. 411–413.
[54] H. Zhu, M. Dong, S. Chang, Y. Zhu, M. Li, and X. Sherman Shen, “ZOOM: Scaling the mobility for fast opportunistic forwarding in
vehicular networks,” 2013 Proceedings IEEE INFOCOM, pp. 2832–2840, Apr. 2013.
[55] E. M. Daly and M. Haahr, “Social network analysis for routing in disconnected delay-tolerant MANETs,” Proceedings of the 8th ACM
international symposium on Mobile ad hoc networking and computing - MobiHoc ’07, p. 32, 2007.
[56] D. Zhang, H. Huang, M. Chen, and X. Liao, “Empirical study on taxi GPS traces for Vehicular Ad Hoc Networks,” 2012 IEEE International
Conference on Communications (ICC), pp. 581–585, Jun. 2012.
[57] J. LeBrun, C.-N. Chuah, D. Ghosal, and M. Zhang, “Knowledge-based opportunistic forwarding in vehicular wireless ad hoc networks,”
in Vehicular Technology Conference, 2005. VTC 2005-Spring. 2005 IEEE 61st, vol. 4, Dallas, Texas, 2005, pp. 2289–2293 Vol. 4.
[58] B. Karp and H. T. Kung, “GPSR: greedy perimeter stateless routing for wireless networks,” in Proceedings of the 6th Annual International
Conference on Mobile Computing and Networking, ser. MobiCom ’00, vol. pages, no. MobiCom, New York, NY, USA, 2000, pp. 243–254.
[59] C. Lochert, M. Mauve, H. F¨
uß ler, and H. Hartenstein, “Geographic routing in city scenarios,” SIGMOBILE Mob. Comput. Commun.
Rev., vol. 9, no. 1, pp. 69–72, 2005.
[60] S. M. Tornell, C. T. Calafate, J.-C. Cano, and P. Manzoni, “Assessing the Effectiveness of DTN Techniques Under Realistic Urban
Environments,” in 38th Annual IEEE Conference on Local Computer Networks (LCN 2013), Sydney, Australia, Oct. 2013.
[61] I. Leontiadis, P. Costa, and C. Mascolo, “Extending Access Point Connectivity through Opportunistic Routing in Vehicular Networks,”
in Proceedings of the 29th Conference on Information Communications (INFOCOM), Piscataway, NJ, USA, Mar. 2010, pp. 486–490.
[62] M. Nakamura, T. Kitani, N. Shibata, K. Yasumoto, M. Ito, and W. Sun, “A method for improving data delivery efficiency in delay tolerant
vanet with scheduled routes of cars,” in Consumer Communications and Networking Conference (CCNC), 2010 7th IEEE, vol. 9, no. 5,
Las Vegas, NV, Jan. 2010, pp. 1–5.
[63] P.-C. C. Cheng, K. C. Lee, M. Gerla, and J. H¨
arri, “GeoDTN+Nav: Geographic DTN Routing with Navigator Prediction for Urban
Vehicular Environments,” Mobile Networks and Applications, vol. 15, no. 1, pp. 61–82, Jun. 2009.
[64] S. Medjiah and T. Ahmed, “Orion Routing Protocol for Delay Tolerant Networks,” in 2011 IEEE International Conference on
Communications (ICC), Kyoto, Japan, Jun. 2011, pp. 1–6.
[65] A. Sidera and S. Toumpis, “DTFR: A geographic routing protocol for wireless delay tolerant networks,” in Ad Hoc Networking Workshop
(Med-Hoc-Net), 2011 The 10th IFIP Annual Mediterranean, Favignana Island, Sicily, 2011, pp. 33–40.
[66] H. Wen, F. Ren, J. Liu, and C. Lin, “A Storage-Friendly Routing Scheme in Intermittently Connected Mobile Network,” Vehicular
Technology, IEEE Transactions on, vol. 60, no. 3, pp. 1138–1149, 2011.
[67] Y. Zhu, Y. Qiu, Y. Wu, and B. Li, “On adaptive routing in urban vehicular networks,” Global Telecommunications Conference, 2012.
IEEE GLOBECOM 2012. IEEE, pp. 1611–1616, 2012.
[68] Y. Ding and L. Xiao, “SADV: Static-node-assisted adaptive data dissemination in vehicular networks,” Vehicular Technology, IEEE
Transactions on, vol. 59, no. 5, pp. 2445–2455, 2010.
[69] W. Xiaomin and C. Song, “Distributed Real-Time Data Traffic Statistics Assisted Routing Protocol for Vehicular Networks,” in Parallel
and Distributed Systems ICPADS 2010 IEEE 16th International Conference on. Ieee, Dec. 2010, pp. 863–867.
[70] A. Skordylis and N. Trigoni, “Efficient data propagation in traffic-monitoring vehicular networks,” Intelligent Transportation Systems,
IEEE Transactions on, vol. 12, no. 3, pp. 680–694, Sep. 2011.
[71] O. Trullols-Cruces, M. Fiore, J. Barcelo-Ordinas, and S. Member, “Cooperative Download in Vehicular Environments,” IEEE Transactions
on Mobile Computing, vol. 11, no. 4, pp. 663–678, Apr. 2012.
[72] K. Mershad, H. Artail, and M. Gerla, “We Can Deliver Messages to Far Vehicles,” Transportation Systems, IEEE, vol. 13, no. 3, pp.
1099–1115, 2012.
36
[73] O. Khalid, S. U. Khan, J. Kolodziej, L. Zhang, J. Li, K. Hayat, S. A. Madani, L. Wang, and D. Chen, “A Checkpoint Based Message
Forwarding Approach For Opportunistic Communication,” in ECMS, 2012, pp. 512–518.
[74] S. M. Bilal, A. U. R. Khan, S. U. Khan, S. a. Madani, B. Nazir, and M. Othman, “Road Oriented Traffic Information System for Vehicular
Ad hoc Networks,” Wireless Personal Communications, Feb. 2014.
[75] R. Meireles, M. Boban, P. Steenkiste, O. Tonguz, and J. Barros, “Experimental study on the impact of vehicular obstructions in VANETs,”
2010 IEEE Vehicular Networking Conference, pp. 338–345, Dec. 2010.
[76] A. Nandan, S. Das, G. Pau, M. Gerla, and M. Sanadidi, “Co-operative Downloading in Vehicular Ad-Hoc Wireless Networks,” in Second
Annual Conference on Wireless On-demand Network Systems and Services, St. Moritz, Switzerland, 2005, pp. 32–41.
[77] F. Gil-Castineira, F. Gonzalez-Castano, and L. Franck, “Extending vehicular CAN fieldbuses with delay-tolerant networks,” Industrial
Electronics, IEEE Transactions on, vol. 55, no. 9, pp. 3307–3314, 2008.
[78] W. Chen, R. Guha, J. Chennikara-Varghese, M. Pang, R. Vuyyuru, and J. Fukuyama, “Context-driven disruption tolerant networking for
vehicular applications,” in 2010 IEEE Vehicular Networking Conference, Dec. 2010, pp. 33–40.
[79] J. Liu, J. Bi, Y. Bian, X. Liu, and Z. Li, “DSRelay: A scheme of cooperative downloading based on dynamic slot,” in 2012 IEEE 12th
International Conference on Computer and Information Technology, no. 2011, Ottawa, ON, 2012, pp. 381–386.
[80] H. Wang, Y. Zhu, and Q. Zhang, “Compressive sensing based monitoring with vehicular networks,” 2013 Proceedings IEEE INFOCOM,
pp. 2823–2831, Apr. 2013.
[81] “Ns2 website,” Nov 2013. [Online]. Available: http://nsnam.isi.edu/nsnam/index.php/Main Page
[82] A. Ker¨
anen, J. Ott, and T. K¨
arkk¨
ainen, “The ONE simulator for DTN protocol evaluation,” in Proceedings of the 2Nd International
Conference on Simulation Tools and Techniques, Brussels, Belgium, 2009, pp. 55:1–55:10.
[83] “Omnet++,” http://www.omnetpp.org/, April 2014.
[84] C. Sommer, R. German, and F. Dressler, “Bidirectionally Coupled Network and Road Traffic Simulation for Improved IVC Analysis,”
Mobile Computing, IEEE Transactions on, vol. 10, no. 1, pp. 3–15, 2011.
[85] “Inet framwork for OMNeT++,” http://inet.omnetpp.org/, November 2013.
[86] R. Protzmann, B. Schuunemann, and I. Radusch, “The influences of communication models on the simulated effectiveness of V2X
applications,” Communications Magazine, IEEE, vol. 49, no. 11, pp. 149–155, 2011.
[87] W. Alasmary and W. Z. W. Zhuang, “The Mobility Impact in IEEE 802.11p Infrastructureless Vehicular Networks,” in Vehicular Technology
Conference Fall VTC 2010 IEEE 72nd, vol. 10, no. 2, Ottawa, ON, 2010, pp. 222–230.
[88] S. Joerer, C. Sommer, and F. Dressler, “Toward Reproducibility and Comparability of IVC Simulation Studies: A Literature Survey,”
IEEE Communications Magazine, vol. 50, no. 10, pp. 82–88, Oct. 2012.
[89] J. Gozalvez, M. Sepulcre, and R. Bauza, “Impact of the radio channel modelling on the performance of VANET communication protocols,”
Telecommunication Systems, vol. 50, no. 3, pp. 149–167, Dec. 2010.
[90] D. Eckhoff, C. Sommer, and F. Dressler, “On the Necessity of Accurate IEEE 802.11 p Models for IVC Protocol Simulation,” in Vehicular
Technology Conference Spring, no. Ivc, Yokohama, Japan, 2012, pp. 1–5.
[91] J. Yoon, M. Liu, and B. Noble, “Random waypoint considered harmful,” in INFOCOM 2003. Twenty-Second Annual Joint Conference
of the IEEE Computer and Communications. IEEE Societies, vol. 2, no. C, 2003, pp. 1312 – 1321 vol.2.
[92] J. H¨
arri, F. Filali, C. Bonnet, and M. Fiore, “VanetMobiSim: generating realistic mobility patterns for VANETs,” in Proceedings of the
3rd international workshop on Vehicular ad hoc networks - VANET ’06, ser. VANET ’06, New York, NY, USA, 2006, pp. 96–97.
[93] M. Behrisch, L. Bieker, J. Erdmann, and D. Krajzewicz, “SUMO - Simulation of Urban MObility: An Overview,” in SIMUL 2011, The
Third International Conference on Advances in System Simulation, Barcelona, Spain, Oct. 2011, pp. 63–68.
[94] E. Teramoto, M. Baba, and H. Mori, “Netstream: traffic simulator for evaluating traffic information systems,” in Intelligent Transportation
System, 1997. ITSC ’97., IEEE Conference on, Boston, 1997, pp. 484–489.
[95] M. Gerla, J.-T. Weng, E. Giordano, and G. Pau, “Vehicular testbeds; Validating models and protocols before large scale deployment,” in
Computing, Networking and Communications (ICNC), 2012 International Conference on, vol. 7, no. 6, Maui, HI, 2012, pp. 665–669.
[96] K. Lee, S.-h. Lee, R. Cheung, U. Lee, and M. Gerla, “First Experience with CarTorrent in a Real Vehicular Ad Hoc Network Testbed,”
2007 Mobile Networking for Vehicular Environments, pp. 109–114, 2007.
37
[97] “Dtn reference implementation,” Nov 2013. [Online]. Available: http://www.dtnrg.org/wiki/Code
[98] J. Eriksson, H. Balakrishnan, and S. Madden, “Cabernet: vehicular content delivery using WiFi,” in Proceedings of the 14th ACM
international conference on Mobile computing and networking, ser. MobiCom ’08, New York, NY, USA, 2008, pp. 199–210.
[99] M. C. G. Paula, J. N. Isento, J. a. Dias, and J. J. P. C. Rodrigues, “A real-world VDTN testbed for advanced vehicular services and
applications,” 2011 IEEE 16th International Workshop on Computer Aided Modeling and Design of Communication Links and Networks
(CAMAD), pp. 16–20, Jun. 2011.
[100] A. Amoroso, G. Marfia, M. Roccetti, and G. Pau, “Creative testbeds for VANET research: A new methodology,” in 2012 IEEE Consumer
Communications and Networking Conference (CCNC), Las Vegas, NV, Jan. 2012, pp. 477–481.
38
ACRONYMS
ADPBSW Average Delivery Probability Binary Spray and Wait
BS&W Binary Spray & Wait
CAN Controller Area Network
CDF Cumulative Distribution Function
DDT Distance Defer Transmission
DRTAR Distributed Real-time Data Traffic Statistics Assisted Routing
DSCF Directional Store-Carry-Forward
DTFR Delay Tolerant Firework Routing
DTN Delay Tolerant Network
DTNRG Delay Tolerant Network Research Group
DV-CAST Distributed Vehicular BroadCast
FEC Forward Error Correction
FFRDV Fastest Ferry Routing in DTN-enabled Vehicular Ad-Hoc
GORF Generic Opportunistic Routing Framework
GPS Global Positioning System
IPN InterPlanetary Network
ITS Intelligent Transport System
IVC Inter-Vehicle Communication
MANET Mobile Ad-hoc NETwork
METD Minimum Estimated Time of Delivery
MSDP Map-based Sensor-data Delivery Protocol
NS Navigation System
P2P Peer-to-Peer
PER Packet Error Rate
RENA REgioN-bAsed
ROD Road Oriented Dissemination
ROI Region of Interest
RSU Road Side Unit
SADV Static-Node-Assisted Adaptive Data Dissemination in Vehicular Networks
TTL Time To Live
UV-CAST Urban Vehicular BroadCast
V2V Vehicle to Vehicle
V2I Vehicle to Infrastructure
39
VADD Vehicle-Assisted Data Delivery
VANET Vehicular Ad-Hoc Network
VDTN Vehicular Delay Tolerant Network
VN Vehicular Network
WAVE Wireless Access for Vehicular Environment
WSMP WAVE Short Message Protocol
Sergio M. Tornell earned his grade in Telecommunications Engineering by the Universidad Polit´
ecnica de Cartagena,
Spain in July 2010. In July 2011, after a short experience in the industry, he joined the GRC research group in the
Universidad Polit´
ecnica de Valencia, Spain. He received his M.Sc. degree in Dec 2011. Currently his is going through
a Ph.D in Computer Science. His research interest include Wireless communications, network and mobility modelling,
and resource management.
Carlos Calafate is an associate professor at the Technical University of Valencia. He graduated with honors in
Electrical and Computer Engineering at the University of Oporto (Portugal) in 2001, and he received his Ph.D. degree
in Computer Engineering from the Technical University of Valencia in 2006. He is a member of the Computer Networks
Group (GRC), and author of more than 250 publications. His research interests include vehicular networks, mobile and
pervasive computing, security and QoS on wireless networks, as well as video coding and streaming.
Juan Carlos Cano is a full professor in the Department of Computer Engineering at the Polytechnic University of
Valencia (UPV) in Spain. He earned an MSc and a Ph.D. in Computer Science from the UPV in 1994 and 2002
respectively. From 1995-1997 he worked as a programming analyst at IBM’s manufacturing division in Valencia. His
current research interests include Vehicular Networks, Mobile Ad Hoc Networks, and Pervasive Computing.
40
Pietro Manzoni received the MS degree in computer science from the “Universit`
a degli Studi” of Milan, Italy, in
1989, and the PhD degree in computer science from the “Politecnico di Milano”, Italy, in 1995. He is currently a
full professor of computer science at the “Universitat Polit`
ecnica de Val`
encia”, Spain. His research activity is related
to mobile wireless data systems design, modelling, and implementation, particularly oriented to Intelligent Transport
Systems. He is a member of the IEEE.
41
