Content uploaded by Hsin-Ya Pan
Author content
All content in this area was uploaded by Hsin-Ya Pan on Apr 22, 2014
Content may be subject to copyright.
Mobile-Gateway Routing for Vehicular Networks
1
Hsin-Ya Pan, Rong-Hong Jan
2
, Andy An-Kai Jeng,
and Chien Chen
Department of Computer Science
National Chiao Tung University
Hsinchu, 30010, Taiwan
{hypan, rhjan, andyjeng, chienchen}@cs.nctu.edu.tw
Huei-Ru Tseng
Information and Communications Research
Laboratories
Industrial Technology Research Institute
Hsinchu, 31040, Taiwan
hueiru@itri.org.tw
Abstract—Development of vehicular ad hoc networks (VANETs)
has drawn intensive attention in recent years. Designing routing
protocols for vehicle-to-vehicle (V2V) communication in VANET
may suffer from frequent link change and disconnection. Vehicle-
to-infrastructure (V2I) communication can overcome the
challenge by relaying packets through the backbone network, but
is limited to those areas where a RSU exists. In this paper, we
present a position-based routing protocol, named mobile-gateway
routing protocol (MGRP), for VANETs. The MGRP combines
V2V and V2I communications, and utilizes certain vehicles as
mobile gateways. Each mobile gateway connects with a base
station through a 3G interface and communicates with other
vehicles without the 3G interface through an IEEE 802.11
interface. Upon receiving packets from a vehicle, the mobile
gateway forwards the packets to a gateway controller via the base
station. The gateway controller then searches the position of the
destination vehicle and determines a set of gateway vehicles close
by the destination to forward the packets. Simulation results
show that the MGRP can significantly improve the packet
delivery ratios and reduce the transmission hop count.
Keywords- vehicular ad hoc network; vehicle-to-vehicle;
vehicle-to-infrastructure; position-based routing protocol;
I. INTRODUCTION
Vehicular Ad Hoc Networks (VANETs) have received
increasing attention from the research and industrial
communities recently. Many valuable applications, such as
entertainments, trip planning, and accident avoidance, have
been envisioned in VANETs.
Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure
(V2I) are two major types of communications in VANET.
Each vehicle equipped with an On-Board Unit (OBU) can
either transmit hop-by-hop to the destination using V2V
communication or transmit to a Roadside Unit (RSU) using
V2I communication.
Different from traditional wireless networks, designing a
routing protocol for V2V communication is more challenging.
The network topology may change rapidly due to the high
speed characteristic of vehicles. Thus, a proactive routing,
such as DSDV [1] that pre-establishes shortest paths between
nodes, is not appropriate. In other words, a packet could be
transmitted in a longer hop in V2V communication. Moreover,
the disconnection problem may happen at the areas of low
traffic density, further degrading the packet delivery ratio.
The challenges in V2V can be overcome by the support of
V2I communication. A vehicle can firstly transmit packets to a
RSU. The RSU connects to the backbone network and thus
can forward packets for vehicles using a more efficient and
reliable way. However, due to the limited transmission ranges
of OBUs, the support of V2I communication is only restricted
to those areas where a RSU is reachable. In other places, the
above challenges still exist.
In this paper, we propose a position-based routing protocol,
named
mobile-gateway routing protocol (MGRP) for VANETs.
The MGRP combines V2V and V2I communication, and
utilizes certain vehicles as mobile gateways to extend the
coverage of fixed RSUs. The OBU on each mobile gateway is
equipped with an IEEE 802.11 interface and a 3G interface.
The other vehicles without the 3G interface can forward
packets through 802.11 links to the nearest mobile gateway.
Upon receiving a packet, the mobile gateway forwards the
packet to a gateway controller through the 3G interface. The
gateway controller then searches the position of the
destination vehicle and determines a set of gateway vehicles
close by the destination to forward the packet.
Simulation results show that the MGRP can significantly
improve the packet delivery ratios and reduce the transmission
hop counts. In other words, the proposed protocol can provide
vehicles with more instantaneous services. We also
investigated the percentage of gateway vehicles to find the
appropriate ratio to guarantee the successful delivery ratio.
The rest of the paper is organized as follows. In Section II,
we review the previous studies and related works. Then, the
system model, assumptions and detailed description of MGRP
are presented in Section III. In Section IV, we evaluate the
proposed approach by simulation and compare with the GPSR
routing protocol. Finally, a conclusion is given in Section V.
II. R
ELATED WORK
Research on V2I communication can be divided into two
categories: One is that the RSU just plays the role of packet
storage but does not provide the function of transmission, such
as SADV [2]. The other focuses on utilizing the pre-
established RSU for transmission. When a vehicle enters the
transmission range of a RSU, it will start to send or receive the
data packet to the RSU.
1
This paper was supported in part by the National Science Council of
the ROC, under Grant NSC-97-2221-E-009-049-MY3
2
Corresponding Author; Fax: 886-3-5721490
IEEE VTS APWCS 2011
A number of routing protocols that use fixed infrastructures
to improve the packet delivery have been proposed. In
MPARP [3], each vehicle is equipped with an IEEE 802.11
and an IEEE 802.16 interfaces. When routes exist, vehicles
can communicate directly with each other using the IEEE
802.11 mode; otherwise, their communication will be taken
over by a base station using the IEEE 802.16 mode. In RAR [4]
and DDR [5], each section of the road is embraced by two
RSUs. When a vehicle has some packets for another vehicle
on a different section, it transmits to one of the RSU on which
it is located. Then, the packets will be forwarded to the
destination through the backbone network. Similarly, the
routing protocol in [6] tries to make a proper decision on
whether to broadcast or to use end-to-end transmission based
on the information provided by RSUs.
Although vehicular communication can be assisted by the
support of V2I model, the advantage is restricted to those
areas where fixed infrastructures exist. To overcome it, the
MIBR protocol in [7] introduces the concept of mobile
gateways. It employs each bus as a mobile gateway to forward
packets for vehicles. Because buses have fixed travel routes
and can be equipped with radios of larger transmission ranges
(over 300m), it is beneficial to improve the delivery radio and
throughput. However, the connectivity between buses is still
limited to the period of bus schedules and the covered region
of bus routes.
To conquer the above limitations, this paper uses certain
vehicles, e.g. taxi, in which 3G infrastructures are added to
their OBUs, as mobile gateways. Other vehicles without the
3G infrastructures can deliver packets to destinations through
those gateway vehicles. Because the coverage of 3G
infrastructure is large enough to cover the whole area, the
proposed protocol can significantly reduce the hop count and
improve the packet delivery ratio.
III. M
OBILE GATEWAY ROUTING PROTOCOL
In this section, we first introduce the architecture and
assumptions of the mobile gateway routing. Then, the MGRP
routing protocol is presented in details.
A. The Architecture of Mobile Gateway Routing
As shown in Fig. 1, we utilize certain vehicles as mobile
gateways to substitute traditional RSUs. The OBU on each
gateway vehicle is equipped with an IEEE 802.11 interface and
a 3G interface. The 3G interface is used to communicate with a
base station in a cellular network and the 802.11 interface is
used to communicate with other vehicles without the 3G
interface. When the base station received data packets from a
gateway vehicle, it will deliver the packets to a gateway
controller. The gateway controller then searches the position of
the destination, determines a set of gateway vehicles close by
the destination, and sends packets to each of the chosen
gateway vehicles via the based station. Finally, those gateway
vehicles will transmit the data packets to the destination with
IEEE 802.11 links.
Figure 1: Architecture of mobile gateway routing.
B. Assumptions
We assume that each vehicle can obtain its position,
velocity, and direction through a global positioning system
(GPS) equipped on the vehicle. This information will be
periodically broadcasted to nearby vehicles within the
transmission range using a hello message. We also assume that
a digital map with traffic load condition of roads is installed in
each vehicle. Besides, if a vehicle founds that the route has
been broken, it will buffer any received data packet and send a
RRER packet to the source vehicle for selecting an alternative
route. Different from ordinary routing protocols, such as
AODV [8], the MGRP limits the time-to life (TTL) value to
three hops.
C. MGRP Routing Protocol
In MGRP, each vehicle can deliver data packets via a
mobile gateway to decrease the transmission hop count and to
achieve more reliable communication quality. Fig. 2 shows
how a source vehicle (left hand side) sends the packet to a
destination vehicle (right hand side). When the source vehicle
has some packets for the destination vehicle, it first searches a
mobile gateway closest to itself, i.e. Gateway1, and sends
packets to the gateway vehicle. Then, Gateway1 forwards the
data packets to a base station using the 3G interface. Upon
receiving the packets, the base station delivers the packets to
the back-end gateway controller in order to search the position
of the destination vehicle and transmits the packets to a set of
gateway vehicles close by the destination vehicle, i.e.
Gateway4. Finally, Gateway 4 will forward the packets to the
destination vehicle via the 802.11 interface. Without the assist
from Gateway1, Gateway4, and the gateway controller, the
source vehicle has to carry the packets for a while until it meets
vehicle1, vehicle 3, or vehicle9. The same problem happens on
the next vehicle carrying those packets. The above relaying
process may cause a longer delay time before reaching the
destination vehicle. Even worse, if the packet is forwarded to
vehicle9 and there is no further vehicle connecting vehicle9 to
the destination, the packet will lose, causing an unreliable
transmission.
Figure 2: Scenario of the source vehicle sending packets to a
destination vehicle
Now we descript a detail processes. Similar to the AODV
protocol, when a vehicle needs to send packets, it first
broadcasts a RREQ packet to the neighboring vehicles. Once a
neighbor received the RREQ packet, if it has no routing path
to the destination, it will rebroadcast the route request to other
neighbors. Different from the AODV, the TTL of RREQ is
limited to three hops in the MGRP. Once a vehicle receives
the RREQ, it first checks whether the hop count is still less
than three. If so, the vehicle will become the next forwarder to
rebroadcast the RREQ packet; otherwise, the vehicle will drop
the RREQ packets. If the information of the destination
vehicle was in its routing table or the gateway vehicle receives
the RREQ packet, it will send back a RREP packet to the
source vehicle. Furthermore, if the vehicle waits for a while
and does not receive the RREP packet, it will rebroadcast the
RREQ packet and repeat above steps.
Figure 3: Packet forwarding in different scenarios
After a source vehicle broadcasts the RREQ packet to
require a routing path, there are three situations may happened.
The first is that there is no other vehicle beside the source
vehicle or cannot find a neighboring vehicle instantly. In this
case, the vehicle carries the packets until a vehicle appears
within its transmission range. Then, it will forward the packets
to the vehicle.
The second situation is that there are more than one
neighboring vehicles, but none of them having a path to a
mobile gateway or destination vehicle in three hops. In this
case, the source vehicle determines a forwarding direction for
the packets according to the road density information. When
the vehicle located at an intersection, it will select the
direction which has the highest road density. As shown in Fig.
3, there is no routing path that can forward data packet from
source 2 to a gateway vehicle or destination vehicle, and the
density of road A (3 vehicles) is higher than that of road B (2
vehicles). So, source 2 forwards the packets to vehicle 8 which
is on road A. This method can improve the packet delivery
ratio, because a higher road density usually implies a higher
probability of finding a mobile gateway if the percentage of
gateway vehicles to ordination vehicles on each road has no
significant difference.
The third situation is that there are more than one routing
paths that can forward to destinations or gateway vehicles. In
this case, the source vehicle needs to select a suitable route
path. The MGRP will select the most reliable path to forward
the data packets. The reliability of routing path is evaluated by
the lifetime of the route. We utilize the following link lifetime
formula proposed in [3] to predict the inter-vehicle lifetime,
ji
ij
VV
DR
lifetimeLink
−
−
=_
,
where R is the transmission range of each vehicle, D
ij
is the
distance between vehicle i and vehicle j, V
i
: the velocity of
vehicle i, and V
j
is the velocity of vehicle j. The lifetime of a
routing path is the smallest link lifetime on this path.
As shown in Fig. 3, there are two routes can forward data
packets from Source1 to a gateway vehicle. The first path is
through Source1→Vehicle2→Vehicle3→Gateway1, and the
second path is Source1→Vehicle2→Vehicle4→Gateway2.
The lifetimes of links on the first route path are 9s, 4s and 7s,
and lifetimes of links on the second route are 9s, 3s and 5s. So,
the route lifetimes of the first and second paths are 4s and 3s
respectively. As a result, Source 1 will select the first route to
forward data packets, which has a longer route lifetime.
Note that if the routing table has recorded the routing path
to the destination vehicle, it has the higher priority to select
this routing path for transmission.
After the data packets are forwarded to a gateway vehicle,
the gateway vehicle will forward these packets to the base
station via 3G network and the gateway controller. The
gateway controller will choose gateway vehicles nearby the
destination vehicle as forwarders. The gateway controller
periodically updates the gateway vehicles’ position. The
forwarding decision of the gateway controller server depends
on whether the distance between gateway vehicles and the
destination vehicle is less than 500 meters. If there are many
gateway vehicles’ distance less than 500 meters, all of those
gateway vehicles will be selected as the forwarders. And the
packets will be delivered to the destination vehicle via V2V. It
could enhance the probability of successfully forwarding data
packets to the destination vehicle. However, if there is no
gateway vehicle forwarding data packets to the destination
vehicle, the gateway controller will drop the data packets. As
shown in Fig. 4, three gateway vehicles will receive the data
packets from the base station.
Figure 4: Gateway controller forwards packets to a set of
gateway vehicles which are less than 500m to the destination
IV. S
IMULATION AND PERFORMANCE EVALUATION
In this section, we evaluate the performance of MGRP
using ns2 simulator [9] (version 2.34). We compare MGRP
with the traditional position-based routing protocol GPSR [10]
routing protocol, and analyze the relationship between the
percentage of gateway vehicles and the success delivery ratio.
We perform the simulation on a real street map, captured
from TIGER database (Topologically Integrated Geographic
Encoding and Reference System) [11]. We simulated the
MGRP within two scenarios, highway and urban. The urban
street layout is within a 1100m*1100m area as shown in Fig. 5.
There are totally sixty-one roads and 150 vehicles. We offer
10 CBR flows and the packet size is 512-byte. The simulation
parameters are summarized in Table 1.
Figure 5: Simulation street layout
TABLE 1. Simulation parameters
Parameter Value
Simulation scenario Highway/ Urban
Speed of vehicles 40–90 km/h
Simulation time 300 sec
Interval time of data delivery 0.5 sec
Data packet size 500 bytes
Transmission range 250 m
Figure 6: The percentage of gateway vehicles vs. packet
delivery ratio
We first evaluate the relationship between the percentage
of gateway vehicles and the packet delivery ratio for a scenario
of 4km highway with four lanes and double directions. The
results are shown in Fig. 6. We can see that when more
vehicles play the role as mobile gateways, the packet delivery
ratio increases significantly. Even in low density scenario (20
vehicles/km), the MGRP can achieve 80% packet delivery ratio
if the percentage of the gateway vehicles is over 60%. When
the percentage of gateway vehicles is 10%, the packet delivery
ratio is down to 38%. Besides, the MGRP needs at least 70%
gateway vehicles to reach the 100% delivery ratio in the low
density scenario. On the other hand, the result shows that it can
perform better in middle (30 vehicles/km) and high density
scenarios (40 vehicles/km). In these cases, the MGRP needs
just 30% of gateway vehicles to achieve nearly 100% delivery
ratio. That is, if there are ten vehicles in the highway scenario,
using our protocol just needs three vehicles to equip the OBU
devices. Therefore, it does not need too much cost in this
network architecture.
Now, we compare the performance of MGRP and GPSR.
Fig. 7 shows the packet delivery ratio versa the maximum
speed. We can see that although the packet delivery ratios of
both protocols decrease as long as the velocity of vehicles
rises, our protocol still perform better, because the MGRP
utilizes the 3G network and gateway controller to assist packet
forwarding so that link disconnect due to high mobility can be
greatly avoided. Notice that the packet delivery ratio of
MGRP is lower than GPSR when vehicles’ velocity is over 85
km/h. The reason is that MGRP has to frequently maintain the
routing table in high velocity scenarios. As a result, it may
raise the opportunity of packets lost.
Fig. 8 shows the average hop count to the maximum speed.
The results reveal that the average hop count increases when
the velocity of vehicles rises regardless of MGRP or GPSR.
The MGRP can keep count within 6 while the maximum hop
count in GPSR is 10. It is because of the fact that we use 3G
network to reduce the transmission hops. In addition, we limit
the hop counts when the source vehicle intends to find a route
to the gateway vehicle or destination vehicle.
Fig. 9 shows the routing overhead versa the maximum
speed. The results show that the packet overhead increases
when the velocity of vehicles rises regardless of MGRP or
GPSR. The packet overhead in MGRP is more than GPSR
because we need to maintain the routing table. However, by
establishing the routing table we can avoid the local maximum
problem in GPSR. Furthermore, our method can decrease the
total hop count to the destination node.
Fig. 10 shows the routing overhead (without hello
message) and the packet delivery ratio versa the hop count
between source vehicle to gateway vehicle. We test average
velocities of 45km/h, 65km/h and 85km/h. The results show
that the packet delivery ratio and overhead increase when hop
count rises, because our protocol has more success rate to
forward packets to gateway vehicles the hop count is raised.
And it also raises the packet overhead.
Figure 7: Packet delivery ratio vs. maximum node speed
Figure 8: Average hop count vs. maximum node speed
Figure 9: Routing overhead vs. maximum node speed
Figure 10: Packet delivery ratio and overhead vs. hop count
V. C
ONCLUSIONS
In this paper, the position-based routing protocol, called
mobile gateways routing protocol, (MGRP) has been proposed
for vehicular ad hoc networks. We utilize certain vehicles as
mobile gateway vehicles equipped with the OBU which can
forward the data packets through interfaces of 3G or IEEE
802.11. Other vehicles without 3G interface can forward the
packets through wireless network to mobile gateway vehicles,
then using 3G interface to forward packets to the gateway
controller. Finally, the gateway controller will forward the
packets via mobile gateway vehicles nearby the destination
vehicle. We design the routing protocol suitable for this hybrid
network architecture and it decreases the total hop counts and
the probability of links disconnection obviously. The
simulation results show that MGRP performs better than the
traditional position-based routing protocol GPSR.
R
EFERENCES
[1] Charles E.Perkins and Pravin Bhagwat. “Highly Dynamic Destination-
Sequenced Distance-Vector Routing (DSDV) for Mobile Computers,”
SIGCOMM Comput. Commun., 1994, Rev., 24(4):234–244.
[2] Y. Ding and L. Xiao, “SADV: Static-node-assisted adaptive data
dissemination in vehicular networks,” Vehicular Technology, vol. 59, no.
5, Jun. 2010, pp. 2445-2455.
[3] C. C. Hung, H. Chan, and E. H. K Wu, “Mobility pattern aware routing
for heterogeneous vehicular networks,” Wireless Communications and
Networking Conference (WCNC), Apr. 2008, pp. 2200-2205.
[4] Y. Peng, Z. Abichar, and J. M. Chang, “Roadside-aided routing (RAR)
in vehicular networks,” IEEE International Conference on
Communications (ICC), Jun. 2006, pp. 3602-3607.
[5] R. He, H. Rutagemwa, and X. Shen, “Differentiated reliable routing in
hybrid vehicular ad-hoc networks,” IEEE International Conference on
Communications (ICC), May 2008, pp. 2353-2358.
[6] N. Brahmi, L. Boukhatem, N. Boukhatem, M. Boussedjra, N. D. Nuy, H.
Labiod, and J. Mouzna, “End-to-end routing through a hybrid ad hoc
architecture for V2V and V2I communications,” Ad Hoc Networking
Workshop (Med-Hoc-Net), Jun. 2010, pp. 1-8.
[7] J. Luo, X. Gu, T. Zhao, and W. Yan, “A mobile infrastructure based
VANET routing protocol in the urban environment,” IEEE International
Conference on Communications and Mobile Computing (CMC), Apr.
2010, pp. 432-437.
[8] C. Perkin, E. Belding-Royer, and S. Das, “Ad hoc on-demand distance
vector (AODV) routing,” IETF Experimental RFC, MANET working
group, RFC 3561, Jul. 2003.
[9] Network simulator. ns-2. http://www.isi.edu/nsnam/ns, 2011.
[10] B. Karp and H. T. Kung, “GPSR: Greedy perimeter stateless routing for
wireless networks,” ACM/IEEE International Conference on Mobile
Computing, Aug. 2000, pp. 243-254.
[11] TIGER. http://www.census.gov/geo/www/tiger/, 2011.
