Content uploaded by Balamuralidhar Purushothaman
Author content
All content in this area was uploaded by Balamuralidhar Purushothaman on Jun 16, 2014
Content may be subject to copyright.
QoS-enabled Group Communication in Integrated
VANET-LTE Heterogeneous Wireless Networks
Rajarajan Sivaraj∗, Aravind Kota Gopalakrishna†, M. Girish Chandra‡, P. Balamuralidhar§
Embedded Systems Innovation Lab,
Tata Consultancy Services,
Bangalore, India.
rsivaraj@ucdavis.edu∗, get.aravindkg@gmail.com†,{m.gchandra‡,balamurali.p§}@tcs.com
Abstract—Ubiquitous integration of high-speed WLANs with
wide-range 3GPP systems results in the service extension of the
backbone cellular network. This paper envisions such heteroge-
neous wireless network architecture by integrating IEEE 802.11p
VANETs with 3GPP LTE to achieve seamless data connectivity
for uninterrupted multimedia sessions amongst spatially-apart
vehicular clusters. Issues on cluster head-based multicasting
and QoS are explored in this paper. An adaptive multi-metric
Cluster Head (CH) election mechanism is proposed to manage
the VANET sub-clusters. In addition to this, construction of a
2-hop virtual overlay mesh-based shared multicast tree for lower-
level multicasting within VANETs is discussed. Following this, the
process of VANET-LTE upper-level communication is detailed,
addressing the issues of CH and gateway handover, and resource
allocation of the LTE eNB. The envisioned architecture enables
the LTE to effectively schedule multimedia sessions based on
the service requirements of the VANET gateways, thus satisfying
QoS. Requisite simulation results are presented to evaluate the
integrated network.
Keywords-VANET; LTE; Group Communication; Gateway
Candidates; Cluster Head; Mesh topology; Shared tree; QoS;
I. INTRODUCTION
Recent research in wireless networking has been focusing
upon the heterogeneous integration of IEEE 802.11-based
wireless ad hoc networks with 3GPP cellular networks, as in
[1], [2]. One such example is integrating Universal Mobile
Telecommunication Systems (UMTS) with Vehicular Ad hoc
Networks (VANET), as discussed in [1]. It provides anytime,
anywhere seamless data access to vehicles due to the service
extension of the 3G network. It, thereby, aims to cater to
the next-generation group communication service applications
Manuscript created June 10, 2011, revised August 4, 2011. This work was
fully supported by TCS Innovation Labs, research division of Tata Consultancy
Services Ltd., and was carried out in TCS Embedded Systems Innovation Lab,
Bangalore, India.
Mr. Rajarajan Sivaraj∗was a researcher at TCS Innovation lab, Bangalore,
India. He currently is a Ph.D student in Computer Science at the University
of California, Davis, USA.
Mr. Aravind Kota Gopalakrishna†was a researcher at TCS Innovation Lab,
Bangalore, India. He is currently a Ph.D Candidate at Eindhoven University
of Technology, the Netherlands.
Dr. M. Girish Chandra‡is a senior scientist at TCS Innovation Lab,
Bangalore, India.
Mr. P. Balamuralidhar§is the head of TCS Innovation Lab, Bangalore,
India
such as video conferencing, mobile TV etc. over vehicles.
Vehicular communication has been gaining a great momen-
tum in the academic and industrial fronts. Toyota has come
forward with the design of an LTE-connected car and Intel,
with a WiMax-connected car, for deploying cloud-based and
monitoring applications, respectively.
IEEE 802.11p, an enhanced version of the IEEE 802.11
family for Wireless Access in Vehicular Environments
(WAVE), providing 5.9 GHz bandwidth with data rates be-
tween 6 Mbps and 27 Mbps, is used in VANETs. The
Federal Communications Commission has allocated spectrum
for Inter-Vehicle Communications (IVC). Though the rates
of the VANET-UMTS integrated network are higher, they
still pose significant challenges for deploying delay-sensitive
multimedia applications, requiring high bandwidth. On the
other hand, 3GPP Long Term Evolution (LTE) provides uplink
and downlink data rates of atleast 50 Mbps and 100 Mbps
respectively, and a round-trip time interval of around 10 ms.
A. Existing issues in heterogeneous wireless networks
The prominent issue in such integrated networks revolves
around the selection and management of vehicular gateway
nodes, which act as liaisons between the IEEE 802.11p-based
VANETs and 3GPP LTE. But, the vehicles do not exhibit
identical characteristics(Eg. Vehicles could travel in different
directions), and so, it is challenging for the gateways to reach
out to them. One of the approaches used to address this issue is
to form clusters of vehicles, based on similar vehicular charac-
teristics and choose Cluster Heads (CH) for each cluster, which
could eventually become gateways [1]. Though clustering is
a good mechanism to address this problem, the cluster head
election process was based only on the geographic position of
the vehicles for a lower TTL value, in [1]. This could result
in the cluster heads not possessing the essential credentials to
manage the VANET clusters and to enable them communicate
with the LTE backhaul, especially when it comes to group
communication.
The common modes of multicasting include a mesh-based
and a tree-based multicasting. While mesh-based multicasting
is more scalable, it always yields to significant overhead
affecting the performance of the VANET. Whereas a tree could
reduce the amount of overhead, it is not robust as a mesh. In a
dynamic scenario, both robustness as well as limited overhead
are essential requisites because of the need to sustain the inter-
connectivity with the LTE backhaul, in spite of the network
dynamics. Identifying the traffic profiles of different VANET
groups for service and scheduling prioritization by the LTE
for end-to-end group communication is another relevant issue.
B. Contributions of the paper
Our paper addresses the above-identified issues by the
following key contributions:
1) Devising a QoS-enabled VANET-LTE integrated ar-
chitecture : For end-to-end communication between
spatially-apart VANET groups through the LTE, we de-
vise a VANET-LTE heterogeneous network architecture,
incorporating the relevant QoS components.
2) Proposing a multi-metric CH election mechanism :
Here, we identify the essential metrics from the perspec-
tives of both the VANET and LTE for a vehicle within a
cluster to act as the CH.
3) Incorporating a dynamic mesh-based multicast tree
: To incorporate a dynamic mesh-based multicast tree,
combining the robustness of a mesh and the scalability
of a tree, for lower-level communication amongst VANET
groups.
The remainder of the paper is structured as follows. The
envisioned VANET-LTE integrated network architecture is
described in Section II. Section III delineates the methodology
of dynamic clustering for grouping vehicles and proposes
a Cluster Head (CH) election mechanism. Section IV dis-
cusses multicasting in the integrated network and addresses
the related QoS issues. The performance of the proposed
mechanisms is evaluated in Section V. The paper concludes
in Section VI with a brief on the future research directions.
II. PRO PO SE D VANET-LTE HE TE ROG EN EO US WIRELESS
NET WORK ARCHITECTURE
The envisioned QoS-based VANET-LTE integrated network
architecture is shown in Fig 1. The topology depicts two
spatially-apart roads with a VANET existing in each of them.
The vehicles in the VANETs are equipped with IEEE 802.11p
radio interfaces. Further, each road has two different tracks,
corresponding to the direction of movement of vehicles.
An LTE Evolved Node B (eNB) base station transceiver
is deployed alongside each road, and the two VANETs are
assumed to be under the coverage region of different eNBs.
The Evolved Universal Terrestrial Radio Access Network (E-
UTRAN) interface enables the vehicles communicate to the
eNB so as to access the core components of the LTE [3].
The main purpose of our paper is to perform an effective
group communication between the spatially-apart VANETs
through the backhaul LTE network, making the best use of E-
UTRAN and eNB resources. Optimal end-to-end group com-
munication requires effective lower-level multicasting within
VANETs and upper-level communication with the LTE eNB.
The architecture focuses on dynamic clustering and cluster
head election mechanisms to achieve this. Referring to the
architecture shown in Fig 1, the region within the LTE eNB’s
coverage, where the LTE received signal strength is intense, is
termed as the 4G active region. Vehicles in the VANET, which
are lying in or moving into the 4G active region and equipped
with the E-UTRAN interface , are termed as the Gateway Can-
didates (GWCs). The E-UTRAN interface is enabled on the
GWCs. Rest of the vehicles that are either not instantaneously
present in the 4G active region or unequipped with LTE E-
UTRAN interfaces are termed as Ordinary Vehicles (OVs).
The E-UTRAN interface is either absent or disabled on them.
Dynamic clustering is performed on the GWCs, resulting
in individual GWC sub-clusters with a Cluster Head (CH),
present in each of them. In group communication scenarios,
multi-casting within VANETs is controlled and co-ordinated
by the CHs. A minimum number of gateways (GWs) are
adequately elected out of them. Only the GWs are activated
with their E-UTRAN interfaces to communicate with the LTE
eNB. The QoS requirements of the multicast sessions are
handled by the four modules: SACM, PPM, GWMM and
SCMM, detailed in Section IV. Based on these requirements,
the LTE eNB multicasts data to more than one GW, which
shall further multicast data to the intended destination vehicles.
It is to be noted that the 4G active region is only a portion of
the entire coverage boundary of the LTE eNB.
III. PROP OS ED CL US TE R HEA D ELECTION MECHANISM
For group communication to be carried out, it is essential
for the vehicles to be dynamically clustered, according to
different relevant metrics, as discussed in [1]. The metrics
used here include the vehicular direction of movement, LTE
Received Signal Strength and IEEE 802.11p wireless transmis-
sion range. Dynamic clustering is performed using the above
metrics, in order. Clustering is followed by the process of
election of a Cluster Head (CH) for each sub-cluster. But,
the CH election, proposed here, involves the IEEE 802.11p
transmission rate, the LTE Uplink/Downlink Channel Quality
Indication (CQI) and the relative distance metrics of the
GWCs, unlike in [1], where the GWC closest to the center
of the sub-cluster deems itself as the CH.
The transmission rate, which is the net bit rate of the
GWC at the link with the minimum channel capacity, is
a critical factor for the CH in team multicasting. A good
transmission rate indicates a higher link capacity across the
channel. Similarly, the CQI value in the LTE E-UTRAN
channel is a reflection of the Signal-to-Interference-plus-Noise
Ratio (SINR) across the communication channel. By this,
both the topological as well as physical layer paradigms are
involved in the CH election process. As stated in [1], in each
sub-cluster, the leading edge GWC is identified by the absence
of GWC neighbors behind it and the trailing edge GWC is
identified by the absence of GWC neighbors before it. The
total number of hops between these border edge GWCs‘ is
termed as the hop length of the sub-cluster. A HELLO packet
broadcast is initiated by the leading-edge GWC within the
sub-cluster and after the broadcast, the GWC possessing the
Fig. 1. Envisioned QoS-enabled VANET-LTE integrated network architecture
maximum weight, computed as detailed in [4], is notified and
elected as the CH. Following are the fields in the HELLO
packet:
•TS : Current Time stamp of the broadcast packet
•ID : Relative identity of the GWC within the sub-cluster.
The leading and trailing edge GWCs are identified as
GWC1and GWCn, where ndenotes the size of the
sub-cluster.
•W : Net weight of the GWC
•DEG : One-hop neighborhood degree of the GWC
•GWCmax : Structure of the GWC with the maximum
weight within the sub-cluster (till the current time stamp)
–GWCid :Relative identity of the GWCmax
–Dr:Relative hop Distance from the leading-edge
GWC
–TxRate :IEEE 802.11p transmission rate of
GWCmax
–CQIeNB :Uplink / Downlink Channel Quality Indi-
cation value of the LTE eNB in the GWCmax
–Wid : Net weight of the GWCmax
–Hop dist: Hop distance from GWCmax to the GWC
•Link State: Structure of the one-hop neighbors to the
GWC, with total number matching value in DEG field
–NBid - Relative ID of the neighbor GWC
–Wid : Net weight of the neighbor GWC
The procedure for the CH election is given in Algorithm
1. The ACK packet piggybacked from each GWCicomprises
the following fields in addition to the TS, ID and GWCmax
fields as in HELLO packet.
•GWCdisc - List of discarded GWCs within the sub-cluster
–GWCid - Relative identity of the GWC discarded
–Wid - Weight of the GWC discarded
–Hop dist - Hop distance from the current GWCi
•Position - Location information of the current GWCi
–GPS co-ordinates (x, y, z)
–Angle of Inclination with respect to Cartesian Space
(Θ)
–Velocity of the current GWCi(ν)
–Link Expiration Time (LET) with the sender
The elected CHs periodically broadcast CH Advertisement
(CHADV) within the sub-cluster. The list of discarded GWCs
in the ACK packet is to prevent re-transmission of control
packets by these GWCs, reduce computational complexity in
comparing the weights and thereby forbidding these GWCs
from contesting the CH election.
The packet structures for the HELLO is in Fig. 2 (a).
Here, basically, each GWC, within the sub-cluster, starting
from the leading-edge GWC (GWC1) broadcasts its param-
eters, and computes its weight based on the parameters. It
also keeps track of the GWC with the maximum weight
(GWCmax)and indicates its metrics in the HELLO packet.
Initially, GWCmax=GWC1. So as to enable every GWC
maintain its two-hop information, the HELLO packet also
includes information of the one-hop neighbors of the source,
to which it sends an ACK packet, and forwards it to the next-
hop. This maintenance of two-hop information is essential for
the construction of the virtual mesh as indicated in the next
section. When the HELLO packet reaches the trailing-edge
GWC and after its metric comparisons are carried out, the CH
is elected as the GWC with the maximum weight. A notifi-
cation NOTIFYCH is sent out, starting from the trailing-edge
GWC. Once the GWC with the maximum weight receives it,
it deems itself to be the CH of that sub-cluster. Following this,
it computes the TTL and broadcasts the CHADV within the
sub-cluster.
Algorithm 1 CH ELECTION: Cluster Head Election Proce-
dure
1: Initiate HELLO packet broadcast by GWC1.
2: Assign the hop dist field of the GWCmax structure to 0
and the remaining fields to the corresponding values of
GWC1.
3: Assign NBid and Wid to NULL.
4: for all one-hop neighbor GWCireceiving HELLO do
5: Compare Dr, TxRate and CQIeNB metrics of
GWCmax with those of GWCi(Xij ) and determine
maximum of each metric as max(Dr), max(TxRate)
and max(CQIeNB ), where 1≤j≤3
6: Compute the scaled value (Yij) of each
metric as (Dr/max(Dr), TxRate/max(TxRate),
CQIeNB /max(CQIeN B )).
7: Determine weight (Wi) of each GWCias
Wij =
3
X
j=1
(Yij ∗P RI ORIT Y F ACT ORj)(1)
Assign Wito field W of the current GWCi
8: if Wi>Wid of GWCmax then
9: Assign the GWCid, Dr, TxRate, CQIeN B and Wid
fields of GWCmax with the respective values of
GWCi.
10: Assign hop dist field value of GWCmax to 0.
11: else
12: Increment hop dist field value of GWCmax by 1
13: end if
14: Assign NBid and Wid of Link State field to the sender
of the current GWCi
15: Piggyback ACK to the sender GWC and notify the list
of discarded GWCs.
16: Update the Link State field of the previous relaying
neighbor with the corresponding ID and W values of
the current GWCi
17: Forward HELLO packet with updated metric informa-
tion to the one-hop neighbor(s) of GWCi
18: if ACK is received then
19: Forward neighbor ACK packet to sender GWC.
20: end if
21: if GWCiis the trailing edge GWC then
22: Send NOTIFYCH packet to GWCmax using the
Time-To-Live (TTL) value, equal to the value in hop
dist field.
23: exit for
24: end if
25: end for
26: if GWCmax receives NOTIFYCH then
27: Compute the TTL value of the CH as in [1]
28: Broadcast Cluster Head Advertisement (CHADV)
within the sub-cluster
29: end if
As every GWC relays the HELLO packet, it also acknowl-
edges receipt of the HELLO packet back to its sender. The
ACK comprises the set of GWCs, discarded from the GW
election, which could include the senders as well, if their
weight is not the maximum at that point of time. This is
to forbid them from participating in the CH election, which
reduces the overhead from these GWCs. The ACK packet
also contains information of the current GWCmax. Its packet
structure is shown in Fig. 2 (b).
Fig. 2. HELLO and ACK packet structures
IV. MULTICASTING AND QOS IN TH E INT EG RATE D
NET WORK
Referring to the architecture in Fig 1, end-to-end multicast-
ing in the integrated network takes place at two levels: upper-
level communication between LTE eNB and VANET CHs, and
lower-level multicast within VANET sub-clusters.
A. Low-level Multicasting
For intra-cluster group communication, this paper discusses
a mechanism to construct a virtual overlay 2-hop mesh within
each sub-cluster and build a shared multicast tree using it.
Construction of a multicast tree is centralized and results in
good efficiency with low overhead. On the other hand, a
distributed mesh is more robust and scalable with alternate
routes for managing link failures. The advantages of both
these procedures are clubbed in the proposed mechanism.
The virtual mesh is constructed during the process of CH
election, discussed earlier. When a HELLO packet is relayed
by a GWC, it encapsulates the link state information of its
sender GWCs within the packet. So, the next receiving GWC
deciphers information about the current GWC’s senders. On
receiving the ACK packet, the current GWC pads information
about the receiver to the ACK and transmits it back to the
sender. Thus, a two-hop link state information is available for
each GWC, from which, it maintains a partial mesh view of
the sub-cluster, using its unicast tunnels.
The virtual mesh topology is shown in Fig 3. It precisely
shows the ID, weight and hop distance values of the link state
neighbors of each GWC. Thus, an adequate TTL value of 2
is configured for every GWC within the sub-cluster, so as to
be communicable up to its 2-hop mesh members. For every
pre-defined time stamp interval, the topological changes, if
any, are updated in the mesh table. Once the underlying mesh
is constructed, by then, the HELLO packet would have been
broadcast within the entire sub-cluster and the GWC with the
maximum weight would have been determined. The trailing
edge GW candidate transmits notification to the GWC with
the maximum weight, electing it as the CH. A core-based
multicast shared tree [5], [6] is constructed, rooted at the CH,
such that the communication between the GWCs and CH is
bidirectional. The CH is designated as the rendezvous point,
and controls and co-ordinates group communication within the
sub-cluster. Using the unicast tunnels of the underlying mesh
with high LET, the shared tree connects all the members of
the multicast group with the CH. The CH is configured with
a TTL value, equal to the maximum number of hops between
the leading and trailing edge GWCs with the CH, as in [1], so
that the CH is communicable with all the GWCs within the
sub-cluster.
Fig. 3. 2-hop link state information for virtual mesh topology
Let V1(i) and T1(i) denote the 1-hop neighborhood set
for GWCiin the underlying virtual mesh and the shared
tree. Let V2(i,j) denote the 2-hop mesh neighborhood set for
GWCivia GWCj. Let χin(x) denote the incoming one-hop
tree neighborhood set of a GWC, say x. Once the CH is
selected, the CH initiates the construction of tree using the
virtual mesh. For the CH, initially, T1=V1. The procedure
for the establishment of links in the shared multicast tree from
the underlying mesh is described in Algorithm 2. The idea
behind the algorithm is that a GWC should determine with
which GWCs, its one-hop tree neighbors should establish an
edge, from its 2-hop mesh view. That is, if jand m, neighbors
of i, are at one-hop with x, then idetermines which GWC,
between jand m, should have an edge with x, by comparing
Link Expiration Time between jand x;LETjx with LETmx.
It follows a Breadth First Search (BFS) strategy by means of
which all edges are established for each GWC in the sub-
cluster. This results in the construction of the shared tree. The
time complexity taken by a GWC to directly determine its tree
neighbors and establish an edge with them is O(n2).
Algorithm 2 CONSTRUCT EDGE: Establishment of edges
for construction of shared multicast tree
1: for all j∈T1(i)do
2: for all x∈V2(i, j), where χin (x) = ∅do
3: if x∈V2(i, m), where m∈T1(i)then
4: if LETj x ≥LETmx then
5: Construct an edge (j,x) in the multicast tree.
6: T1(j) = T1(j)S{x}
7: end if
8: else
9: Construct an edge (j,x) in the multicast tree.
10: T1(j) = T1(j)S{x}
11: end if
12: end for
13: Forward T1(j)to j.
14: end for
B. QoS-enabled Upper-level VANET-LTE communication
After the CHs of different sub-clusters are elected, each
vehicular source elects its respective Gateway (GW) from the
available CHs using the GW selection mechanism, discussed
in [1]. However, the metrics used here include IEEE 802.11p
transmission rate and CQIeNB of the CHs, along with their
mobility speed and Route Expiration Time (RET) with the
sources. The Hybrid Gateway Discovery mechanism, detailed
in [1], notifies the sources about the elected GWs. The E-
UTRAN interfaces of the GWs are activated to enable com-
munication with the LTE eNB. This section addresses critical
issues of CH/GW handover and LTE resource allocation to
schedule GWs.
As a first step, a domain-level hierarchical multicasting
approach, as discussed in [7], is followed in which, the
GWs serve as the sub-roots for communication between the
source and destination VANET sub-clusters. When the existing
gateway loses its optimality, a gateway handover mechanism
as proposed in [1] is initiated. But here, the optimality check
is performed for the IEEE 802.11p transmission rate, CQIeNB
and RET metrics. The E-UTRAN interface of the serving
GW is de-activated and the newly-elected GW registers with
the LTE eNB. A similar handover is subjected to CH by
checking its optimality for CQIeNB , IEEE 802.11p Txrate
and Drmetrics. If the vehicles in the destination group are
GWCs, they are identified by their respective CHs, designated
as their serving GWs. But, if the destination vehicles are OVs,
there are a few cases to be observed, based on service level
agreements.
Case 1: When the destination OVs are already aware of
their participation in the multicast sessions
In this case, the OVs identify the instantaneous CHs using
the hybrid discovery mechanism by broadcasting CH solicita-
tion messages.
Case 2: When the destination OVs are unaware of their
participation in the multicast sessions
Here, the CH discovery is purely pro-active as the CHs
periodically broadcast their advertisement messages with ap-
propriate metrics within the destination VANET.
In both these cases, when the destination OVs receive
metrics simultaneously from more than one CHs, they select
their respective gateways out of these CHs, as discussed above.
After selection, they communicate to the respective GWs,
which add them to their multicast group. A virtual mesh
view of the multicast group, as discussed above, is maintained
by the CH to facilitate effective group communication with
the destination vehicles. However, the OVs are not a part
of the sub-cluster. Then, the elected GWs will have their E-
UTRAN interface activated to communicate with the backhaul
LTE network. The Multicast Broadcast Multimedia Services
(MBMS) is a feature of the 3GPP LTE [8], the subscription of
which could be activated by the GWs. The MBMS enables the
same multimedia content to be transmitted to different GWs
on a point-to-multipoint (p-t-m) basis, if the vehicles of the
same destination group are handled by more than one GW
by sprawling over more than one sub-cluster. LTE scheduling
for the purposes of resource allocation to GWs is another
concerned issue because the number of GWs keeps varying
across different time instances. So, the bandwidth for each GW
should be dynamically, yet effectively, allocated. In view of the
same, this paper proposes a DiffServ-based QoS framework,
as discussed in [9]. The framework, as in Fig 1, comprises 4
modules:
1. Policy Provisioning Module (PPM) - Handles priority re-
quirement of the multicast sessions on the GW side
2. Session Admission Control Module (SACM) - Decides
session admission/drop based on the requirements of the
GW.
3. Sub-Cluster Management Module (SCMM)-Manages ve-
hicular mobility, multicast mesh maintenance and resources
within the VANET sub-cluster.
4. GW Management Module (GWMM) - Reserves and man-
ages resource for GWs on the LTE eNB side.
The LTE-SAE traffic classes include Conversation Voice,
Conversational, Streaming, Interactive and Background, as
detailed in [8]. The GW decides the QoS requirements for the
multicast session. Based on the requirements, the PPM sets the
subscription profile of the sub-cluster. The QoS metrics include
delay, jitter, packet error rate and loss ratio, and throughput
and are decided by the LTE eNB by differentiating its services
based on the priority of the following parameters:
•Number of destination vehicles to be served (nd)
•Net Bandwidth required for each GWC (β)
•Number of multimedia sessions to be served (ns)
The SACM module on the side of the LTE eNB is interfaced
with the PPM to schedule sessions. If ndis to be given
more priority for the LTE, then the SACM module prioritizes
session admission for the GW with the maximum cluster
size or multicast group size. This requirement could serve
for Interactive and Background classes. If βis given more
priority, then the SACM module admits session for that GW,
which serves less number of vehicles, so that the available net
bandwidth for an individual vehicle is more. The traffic classes
served as a result include Conversation and Streaming, as they
are delay-sensitive, requiring higher individual bandwidth. On
the other hand, if nsshould be prioritized, then the appropriate
traffic class is Streaming and the GW having the highest
downlink CQIeNB and IEEE 802.11p transmission rate will
be prioritized, since these parameters reflect good end-to-end
group communication.
V. RE SU LTS AND DISCUSSIONS
The simulations of the proposed mechanisms in an inte-
grated VANET-LTE network environment have been carried
out in Network Simulator NS2.34. A source and destination
VANET, each comprising of 50 vehicles, along with the LTE
E-UTRAN and core network are considered in topographical
area of 8000x1000m2, for a simulation time of 200s. The
standards of IEEE 802.11p-based Wireless Access for Vehic-
ular Environments (WAVE), discussed in [10], are utilized for
vehicular communication amongst vehicles (V2V) or between
vehicles and internet gateways (V2I). The Manhattan mobility
model is considered to model the vehicular traffic. Protocol
for Unified Multicasting through Announcements (PUMA),
detailed in [11], is used for multicasting over VANETs.
Packet size of 1 KB, suited for multimedia data transfer, is
used. Simulation of LTE Access and Core Networks, and its
integration with IEEE 802.11p network interface is carried
out using Multi-Interface Cross Layer Extension for NS2
(NS-MIRACLE), as discussed in [12]. Configuration of the
LTE E-UTRAN and core network is referred from [3], [8],
[13]. The performance of the integrated network for Group
Communication purposes is evaluated in terms of Data Packet
Delivery Ratio (DPDR), Packet Error Rate (PER), Delay and
Throughput metrics. The DPDR and throughput metrics gen-
erally decrease, whereas the PER and delay metrics increase
with the increase in the number of multicast sessions and
participating vehicles. The dynamic clustering, reflecting the
CH and GW management mechanisms, is iteratively carried
out every second. The mechanism proposed in this paper
is dubbed as Clustered Virtual Mesh-based Tree (CMVT).
PUMA routing protocol is used to evaluate the performance
of CVMT and it is compared with the standard PUMA multi-
casting (without CVMT) and simultaneous AODV unicasting
over CMGM [1] in VANETs. Equal priorities are given for all
the concerned metrics for the CH election as well as CH/GW
management functionalities.
In Fig 4, the DPDR for the three protocols is evaluated
against the number of on-going multicast sessions in the
integrated network. Each multicast session is considered to
compose of a maximum of two vehicular sources. DPDR
for PUMA over our proposed CVMT shows an average of
6.49% increase over DPDR for standard PUMA and 16.91%
increase over DPDR for simultaneous AODV unicasts using
CMGM. The good performance of PUMA over CVMT is
attributed to the effective management of sub-clusters by the
respective CHs. Further, the CVMT is scalable as well as
efficient in handling dynamic topologies and the CH/GW
handovers sustain inter-connectivity with the backbone LTE
network. Fig 5 shows the PERs of the three routing protocols
as against the IEEE 802.11p wireless transmission range of
vehicles. PER is taken as the fraction of the net data rates
(max. 27 Mbps) of the vehicles, due to packet errors, measured
in Mbps. On an average, PUMA over CVMT shows a 4%
decrease in PER compared to the standard PUMA and around
20.32% decrease, compared to simultaneous AODV unicasts
using CMGM. Shorter transmission ranges result in more
number of VANET sub-clusters and the subsequent generation
of larger number of control packets results in the unwanted
consumption of the available bandwidth. Packet drops and
hence, the error rates increase. Lower PER values in PUMA
over CVMT is also attributed to the robustness of the virtual
mesh and the lower overhead of the shared multicast tree.
Fig. 4. Performance of the three protocols in terms of DPDR for varying
number of multicast sessions.
Fig. 5. Packet Error Rates for different average IEEE 802.11p wireless
transmission ranges.
In Fig 6, the time elapsed since the broadcast of GW-
SOL/GWADV messages till the point of establishment of a
path from the vehicular sources to the gateways is plotted
for varying number of VANET sub-clusters. PUMA over
CVMT shows 10.45% more delay than AODV over CMGM
and around 8.95% less delay compared to standard PUMA
multicasting. This is because of the higher amount of delay
accounted for multicasting due to communication with more
than one vehicle within the group. However, by dynamic and
stable connectivity, CVMT reduces the latency for PUMA,
compared to the standard PUMA for VANET multicasting.
Fig. 6. Delay involved in establishment of communication path from the
vehicular sources to the elected gateways after broadcast of GWSOL/GWADV.
Fig. 7. Performance of the average downlink LTE throughput for varying
number of vehicular gateways in destination VANET.
In Fig 7, the average LTE downlink throughput is measured
against the number of vehicular gateways in the destination
VANET, by varying the session priorities of the gateway.
Profiles P1, P2 and P3 indicate the priority requirements of the
gateway and these profiles are mapped on to the QoS classes
of the LTE for scheduling purposes. Referring to section IV,
the following are the policy requirements of the profiles.
Profile P1 : (nd= 0.5, β= 0.3, ns= 0.2),
Profile P2 : (nd= 0.2, β= 0.5, ns= 0.3) and
Profile P3 : (nd= 0.3, β= 0.2, ns= 0.5)
Mapping the gateway’s priorities to the QoS profiles of the
LTE eNB, P2 indicates an improvement of 3.65% over P1 and
3.71% over P3, indicating the emphasis given to bandwidth
(Conversational and Streaming classes) requirement of the
vehicular gateways. From the graphs, different priority policies
of the gateways are effectively handled, thereby giving a good
QoS guarantee. Further, the total delay in the LTE uplink
and downlink communication with VANET gateways is also
determined and is practically found to be around 15 ms, which
is an acceptable round trip latency value for delay-sensitive
multimedia applications.
VI. CONCLUSION AND FUTURE RESEARCH DIRECTIONS
This paper envisioned a novel VANET-LTE integrated ar-
chitecture to provide multimedia communication services over
spatially-apart vehicular groups. An effective cluster head
election mechanism is proposed to effectively manage VANET
sub-clusters. Following this, a virtual 2-hop overlay mesh-
based shared tree for lower-level VANET multicasting is
proposed. Upper-level multicasting dealt with concepts of
CH/GW handover and discussed a QoS framework for the LTE
eNB to schedule and serve the VANET gateways. The simu-
lation results demonstrated that the integrated system shows
acceptable values in terms of LTE throughput and end-to-end
delay, thereby indicating an improved performance. As future
work, further research is in progress towards exploring the
capabilities of the MBMS feature of the LTE and incorporation
of appropriate Erasure Correction Codes, as the latter are
very useful candidates for robust multicasting to arrive at an
enhanced scheme to support QoS.
REFERENCES
[1] A. Benslimane, T. Taleb, and R. Sivaraj, “Dynamic clustering-based
adaptive mobile gateway management in integrated VANET - 3G
heterogeneous wireless networks.” IEEE Journal on Selected Areas in
Communications, vol. 29, no. 3, pp. 559–570, March 2011.
[2] R. Manoharan, R. Sivaraj, S. Sashtinathan, and K. Sriram, “A novel
multi-hop B3G architecture for adaptive gateway management in hetero-
geneous wireless networks,” in Proc. 5th IEEE WiMob 2009, Marrakech,
Morocco, October 2009.
[3] D.M-Sacristan, J.F.Monserrat, J.C-Penuelas, D.Calabuig, S.Garrigas, and
N.Cardona, “On the way towards fourth-generation mobile: 3GPP LTE
and LTE-Advanced,” EURASIP Journal on Wireless Communications
and Networking, vol. 2009, p. 10, 2009.
[4] F. Setiawan, S. Bouk, and I. Sasase, “An optimum multiple metrics
gateway selection mechanism in MANET and infrastructured networks
integration,” in Proc. IEEE WCNC 2009, Las Vegas, NV, March 2009.
[5] C. Gui and P. Mohapatra, “Overlay multicast for MANETs using
dynamic virtual mesh,” Wireless Networks, vol. 13, no. 1, pp. 77–91,
May 2006.
[6] A.Benslimane and O.Moussaoui, Hierarchical Multicast Protocols with
Quality of Service. WILEY, January 2010, ch. 2.
[7] C. Gui and P. Mohapatra, “Scalable multicasting in mobile ad hoc
networks,” in Proc. IEEE INFOCOM 2004, Hong Kong, March 2004.
[8] E.Dahlman, S.Parkvall, J.Skold, and P.Beming, 3G Evolution - HSPA
and LTE for Mobile Broadband. Academic Press, 2008.
[9] X.G.Wang, G.Min, J.E.Mellor, K.Al-Begain, and L.Guan, “An adaptive
QoS framework for integrated cellular and WLAN networks,” Computer
Networks, vol. 47, no. 2, pp. 167–183, August 2004.
[10] T. Murray, M. Cojocari, and H.Fu, “Measuring the performance of IEEE
802.11p using ns-2 simulator for vehicular networks,” in Proc. IEEE Int’l
Conf. on Electro/Information Technology (EIT), Ames, IA, May 2008.
[11] R.Vaishampayan and J.J.Garcia-Luna-Aceves, “Efficient and robust mul-
ticast routing in mobile ad hoc networks,” in Proc. IEEE Int’l Conf. on
Mobile Ad-hoc and Sensor Networks, Florida, USA, October 2004, pp.
304–313.
[12] N. Baldo, F. Maguolo, M. Miozzo, M. Rossi, and M. Zorzi, “ns2-
MIRACLE: a modular framework for multi-technology and cross-layer
support in network simulator 2,” in Proc. 2nd International Conference
on Performance Evaluation Methodologies and tools, Nantes, France,
October 2007, pp. 1–8.
[13] Q. Qiu, J.Chen, L. Ping, Q. Zhang, and X. Pan, “LTE/SAE model and
its implementation in NS2,” in Proc. 5th IEEE Int’l Conf. on Mobile
Ad-hoc and Sensor Networks 2009, Fujiyan, China, December 2009,
pp. 299–303.