GPSFR: GPS-Free Routing Protocol for Vehicular Networks with Directional Antennas

International Journal of Wireless & Mobile Networks (IJWMN), Vol 1, No 2, November 2009
64
GPSFR: GPS-FREE ROUTING PROTOCOL FOR
VEHICULAR NETWORKS WITH DIRECTIONAL
ANTENNAS
Qing Yang1
, Alvin Lim2
, and Prathima Agrawal3
1
Department of Computer Science and Software Engineering, Auburn University, AL,
USA
yangqin@auburn.edu
2
Department of Computer Science and Software Engineering, Auburn University, AL,
USA
limalvi@auburn.edu
3
Department of Computer and Electrical Engineering, Auburn University, AL, USA
pagrawal@eng.auburn.edu
ABSTRACT
Efficient and practical communications between large numbers of vehicles are critical in providing high
level of safety and convenience to drivers. Crucial real-time information on road hazard, traffic
conditions and driver services must be communicated to vehicles rapidly even in adverse environments,
such as “urban canyons” and tunnels. We propose a novel routing protocol in vehicular networks that
does not require position information (e.g. from GPS) but instead rely on relative position that can be
determined dynamically. This GPS-Free Geographic Routing (GPSFR) protocol uses the estimated
relative position of vehicles and greedily chooses the best next hop neighbor based on a Balance Advance
(BADV) metric which balances between proximity and link stability in order to improve routing
performance. In this paper, we focuses primarily on the complexity of routing in highways and solves
routing problems that arise when vehicles are near interchanges, curves, and merge or exit lanes of
highways. Our simulation results show that by taking relative velocity into account, GPSFR reduces link
breakage to only 27% that of GPSR in the dense network. Consequently, GPSFR outperforms GPSR in
terms of higher data delivery ratio, lower delay, less sensitivity of the network density and route paths’
length.
KEYWORDS
Geographic Routing Protocols, GPS Free, Wireless Vehicular Ad Hoc Network, Relative Position
Maintenance Algorithm, Directional Antennas
1. INTRODUCTION
In the future, large scale vehicular ad-hoc networks will be available to provide drivers with
higher level of safety and convenience. For instance, multi-hop wireless communication
between vehicles can enhance ACC (adaptive cruise control) systems by enabling rapid
adaptation of longitudinal control in response to traffic accidents that just occur a short distance
ahead of it (possibly a few wireless hops). It can also enable smart vehicles to react rapidly to
sudden braking when a "hidden" vehicle that is ahead of it by several vehicles is braking. Such
capabilities will be instrumental in improving highway traffic safety. As reported in [12] by the
National Highway Traffic Safety Administration (NHTSA), in U.S. alone, vehicle crashes on
the highway resulted in the loss of as many as 40,000 lives and an overall economic losses of
more than $230 billion. The motivation to reduce accidents on highways has sparked increasing
interest in research on improving vehicle safety through inter-vehicle communication in the

65
vehicular ad-hoc networks [1, 2, 3, 4, 26, 27], including an effort by IEEE on a standard for
inter-vehicle communication.
There are three possible network architectures of vehicular networks: infrastructure-based, ad-
hoc networks and hybrid. In this paper, we focus on vehicular ad-hoc network (VANET)
architectures because the cost of building such a network is very low and it can even operate in
the events of disasters. Deployment of such networks is flexible and self-organizing. The other
architectures require infrastructure support which has three drawbacks: high operating cost,
limited bandwidth and symmetric channel allocation for uplink and downlink. There have been
a number of research efforts on vehicular ad-hoc networks. For instance, the medium access
control (MAC) problem was addressed in [5, 6]. To improve safety and commercial services, a
multi-channel MAC protocol was proposed in [5]. Routing issues were addressed in [1, 2, 3, 7],
including vehicle-assisted trajectory-based routing protocol [1], mobility-centric data
dissemination [2] and position-base routing [7]. To further solve the network disconnection
problem, [3] used the historical traffic data from digital maps to compute the probabilities of
network connectivity of all road segments. Then, the path with the highest probability of
network connectivity will be selected to forward packets.
However, all the above routing protocols rely on the positions of nodes and require vehicles to
be equipped with GPS receivers. Though GPS will become standard equipment in vehicles in
the future, it may still fail when the power source is depleted or the signals from satellites are
blocked by tall buildings in “urban canyons”, tunnels or bad weather. In this paper, we present a
GPS-free geographic routing (GPSFR) protocol that uses only relative positions of vehicles
which can be determined dynamically. Based on the relative distance and velocity, a new
routing metric called balance advance (BADV) is designed to balance between proximity and
link stability. Unlike other route optimization metrics [2, 7], BADV improves performance in
routing without relying on nodes’ locations. Our simulation results of vehicular networks in
highway scenarios show that GPSFR outperforms GPSR [16], achieving fewer link breakage,
higher data delivery ratio and low network delay.
The remainder of this paper is organized as following. In Section II, we summarize several
related work. Then, we discuss our motivation and assumptions in Section III. The Relative
Position Maintenance (RPM) and routing algorithms are described in Section IV. In Section V,
we discuss our simulation environment and present the simulation results. Section VI presents
the conclusion and future work.
2. RELATED WORK
There are a number of existing techniques for finding the location of nodes in wireless ad hoc
network [19, 20, 21, 22, 23, 24, 25]. As stated in [18], these techniques require nodes to be able
to measure the distances between itself and the neighbors using signal strength or time
differences. Therefore, the effectiveness of this sort of approaches will rely heavily on the
accuracy of distance or time estimation which may be adversely affected by large spurious
variation in signal strength and time synchronization, the absence of line of sight, and
specialized signal processing hardware or software.
A fully distributed, infrastructure-free positioning algorithm for mobile ad hoc network has been
proposed in [8] that do not rely on anchor nodes. However, it is not suitable for vehicular
networks for two reasons. First, after determining the relative positions of neighbors, each node
must change its local coordinate system to the network coordinate system. Such update
overhead increases as the network size increases. In fact, it is proven in [9] that the volume of
message exchanges in [8] increases exponentially with the node density. Secondly, in highly
mobile vehicular networks, the overhead of updating location reference group composed of

66
nodes with lower moving speed is significant. Although cluster-based method in [9] can
generate less communication overheads compared to [8], the number of message exchanges is
still huge because it requires coordinate translation of master nodes throughout the network.
A scheme was proposed in [10] which can localize nodes through fewer message exchanges.
However, the scheme in [10] is applicable to ad-hoc networks with less mobility, such as sensor
networks, but unsuitable for vehicular networks. The high mobility in vehicular networks will
result in large network overhead because of periodic bootstrapping beacons. The number of
flooding nodes during the bootstrapping phase will increase as network size increases [10].
Relative position is also used to warn if a collision is happening by checking the relative
distance between vehicles [11], but has not been used for solving the routing problem.
Unlike all the existing methods, the relative position information of nodes in GPSFR can be
maintained through only localized broadcasting and hence significantly reduces position update
overhead compared to those that require flooding the entire network. In addition, to the best of
our knowledge, GPSFR is the first method to improve the performance of geographic routing in
VANET without reliance on nodes’ position.
3. ASSUMPTIONS
We focus primarily on vehicular networks in the rural highway scenario. A rural highway
provides a link between urban areas. To determine and maintain neighbors’ relative position,
each node requires a compass and two directional antennas [13] pointing in opposite directions.
One antenna is for sending/receiving data to neighbors in front of it, and the other for those
behind it. Other advantages of directional antennas are longer radio ranges, absence of exposed
stations problems and reduction of co-channel interference. We also assume that the coverage
area of each antenna is a semi-circle, thus the area covered by the two antennas will form a
circle.
4. ROUTING PROTOCOL
4.1. Relative Position Maintenance
While moving at the same direction on highways, vehicles will construct a linear network, as
shown in Figure 1. Problems of vehicles at interchanges and ramps will be discussed later. For
now, we will just focus on linear networks. In such networks, the delivery of packets can be
categorized as forwarding or backwarding. In forwarding (backwarding), the routing algorithm
needs only to choose the next hop from neighbors that are moving in the same (opposite)
direction as packets being delivered. In order to achieve this, each node has to compute and
maintain the relative positions of all its neighbors. As shown in Figure 1 (a), suppose all nodes
from 1 to 6 are neighbors, then from node 3’s perspective, node 5 is at a closer relative position
than node 6. Also from node 3’s perspective, the relative position of node 1 should be further
than node 2’s.
Suppose the forward and backward directional antennas are f_antenna and b_antenna,
respectively. If the message arrived at one’s f_antenna, then it must be sent by a node in front of
the receiver; otherwise, it comes from a rear node. Therefore, each node can divide its neighbors
into two groups (fgroup and bgroup) by checking from which antenna the messages are
received. For example in Figure 1, the fgroup of node 3 should be {4, 5, 6} and the bgroup is
{1, 2}. Even on a curve, as shown in Figure 1 (b), vehicles can also divide neighbors into two
groups (assume the curve is not be very sharp which is reasonable on the highway). In addition,
we can easily distinguish packets received from the vehicles moving in the opposite direction
because if such message was sent from f_antenna (or b_antenna) of nodes moving at the
opposite direction, then the receiver will obtain it also from the f_antenna (or b_antenna).

67
Figure 1. Illustration of nodes’ relative positions on straight and curve road
After dividing neighbors into proper groups, each node will send such group information
periodically. The format of this beacon message (called group_update) is: <bgroup, id, velocity,
direction, fgroup>. If no group_update was received after a certain time-out period T, the
neighbor will be considered out-of-range and deleted from the neighbor list. Actually, we could
make GPSFR’s beacon mechanism fully reactive, in which nodes will solicit beacons only when
they have data to transmit. However, we felt this is unnecessary since the one-hop beacon
overhead does not cause too much congestion.
Although computing the relative positions of nodes are straightforward, there is still a hidden
neighbor problem. Suppose vehicle A has two apparent neighbors B and C in front (or behind),
but B and C are not neighbors; then we can say B and C are the hidden neighbors of A. For
example, as shown in Figure 2, node 2 and 4 are the hidden neighbors of node 1.
Figure 2. Node 2 and 4 are hidden neighbors of node 1
Since the hidden neighbors are actually not neighbors, then it is not always possible to obtain
their relative positions through directional antennas. As shown in Figure 3, let the
communication range of nodes be R and the width of each lane be d. On each lane, there is only
a small area where hidden neighbors may exist. We denote the length of such area as li, which
can be calculated as follows:
2 2 2
/ 2 ( / 2)
i
l R R i d
= − − × (1)

68
where m is the number of lanes which is usually from 2 to 6, R is 250 meters and d is 3.6
meters, which is the typical width of lanes in highways [14]. Then the length of each piece will
be very short, so the hidden neighbor problem is an unlikely event in vehicular networks. Note
that the hidden neighbor problem also arises when vehicles are side by side, close to an
interchange or try to leave the highway. In our protocol, the hidden neighbors are not included
in the neighbor list, since they are only hidden for a short time.
Figure 3. There are only small pieces where hidden neighbors may exist
4.1.1. Relative Position Maintenance (RPM) Algorithm
Within a certain neighborhood, except for those hidden neighbors, each node can determine
whether or not a neighbor is in front or not through the directional antennas. The Relative
Position Maintenance (RPM) algorithm uses group_update message exchanges to compute and
maintain nodes’ relative positions. As shown in the Table 1, each node ni will maintain two
linked lists Fi and Bi. Fi is used to record the front neighbors and Bi is for rear neighbors.
Elements of each linked list are ordered by the nodes’ relative positions.
If a message is received from nj, then node ni will first check whether or not this message is in
its cache. If this message matches an entry, then there will be no change on nj in the list. It just
updates the lifetime of node nj in the list. If it is a new message, ni will first add/update the
message in its cache and then arrange nj’s new relative position in the list. Line 2-9 is used to
deal with the scenario of new vehicles merging into the networks, which will be examined later.
Line 14-28 arrange node nj in the corresponding list. We note when node nj overtook ni during
the last beacon period, the distance between nj and ni should be approximately equal to 0. From
ni’s perspective, nj will be an anchor node since its exact relative distance is known now. In the
later beacon periods, we can estimate nj’s new relative distance from the length of beacon
period and the relative velocity between nj and ni. Clearly, the relative distances of anchor nodes
are more accurate.
So far, we have established the relative positions of all neighbors, but we do not know the
relative distance between them. As described above, some neighbors may become anchor nodes
that have more accurate position. Thus, we can use those anchor nodes’ relative distance to
estimate other distances. If there is no anchor node in the list, then nodes’ distances will be
estimated to be evenly distributed between each neighbor.

69
Table 1. Relative position maintenance algorithm.
Algorithm: Relative Position Maintenance (RPM)
Input: Message mj <GFj, IDj, Vj, Dj, GBj> received from nj
Output: Ordered link list Fi and Bi for current node ni
C: Cache for all recently received messages
Fi: Ordered link list of neighbors located in front of ni
Bi: Ordered link list of neighbors located behind ni
e: Temp variable holding the element of the ordered link list
GFi: Neighbors located in front of ni
GBi: Neighbors located behind ni
f-antenna: whether message received from f antenna
Interval: Period of beacon message
1. if(mj is not in C) then
2. if(size of mj == 2&&Di==Dj)
3. if(f-antenna) then
4. add IDj into GFi;
5. else
6. add IDj into BFi;
7. endif
8. exit
9. endif
10. if (size of mj >=3 && Dj is on clock-wise direction of Di) then
11. drop this msg.
12. endif
13. add/update the entry of<GFj, IDj, Vj, Dj, GBj> in C
14. if(IDj is in Bi && f-antenna) then /* nj overtake ni */
15. e←remove element corresponding to IDj from Bi
16. set e as an anchor and reset e’s life time
17. e.position←0;
18. endif
19. if(IDj is in Fi && !f-antenna) then /* nj move backwards of ni */
20. e←remove element corresponding to IDj from Fi
21. set e as an anchor and reset e’s life time
22. e.position←0;
23. endif
24. if(f-antenna) then /* add nj into the corresponding list */
25. Insert(Fi, GFj, GBj, IDj, Vj)
26. else
27. Insert(Bi, GFj, GBj, IDj, Vj)
28. endif
29. update GFi and GBi by new Fi and Bi
30. if(one Interval passed) then
31. RPU(Vi, Fi, true) and RPU(Vi, Bi, false)
32. endif
33. else
34. reset the life time of element related to IDj in Fi or Bi
35. endif
4.1.4. Vehicles Leaving Highways

70
If packets need to be forwarded to the front and the forwarder (vehicle) is trying to leave the
highway, then this packet will go out of networks. To avoid this problem, a backup scheme is
adopted to send the packet to another rear node. The following are the details of this scheme.
While vehicles make a right turn, two cases may occur: this vehicle is on a right-turn curve of
the highway, or it is leaving the highway. In both cases, packets are both forwarded and sent to
a rear neighbor. If the rear neighbor is also leaving the highway, this backup process continues
for a certain number of times. Though this backup scheme can avoid routing packet out of the
network, it still has some overhead while the vehicles are moving on a curve. Suppose the
backup process is repeated k times, then the problem is how to choose a minimal k.
Figure 4. Vehicles moving nearby an exit of the highway
Suppose near the exit of a highway, as shown in Figure 4, there are m cars connected through
wireless links and one of them (e.g. q) is located at the junction of the highway and exit. Then
there will be m + 1 possible deployment of these nodes. For example, one is in the area A and
(m – 1) in the area B. If there are one or more nodes in the area A, then the packet will pass the
exit successfully. The problem arises only if there is no vehicle in the area A, so the probability
of this case will be
1
( 1)
m +
. Now suppose all m nodes are in area B, and then the probability of
k continuous nodes leaving off the highway will be
1
k
m
C
. Therefore, we can obtain that after k
times of backup, the probability of packet being routed out of the network is:
1 1
1
k
out k
m
P p
m C
= ⋅ ⋅
+
(2)
where p is the probability that the node leaves the highways and the largest value for it is 0.5.
Now assume we are trying to forward a packet through a routing path with n exit junctions.
Then the probability p of the last exit is 0.5, while the probability of the vehicle leaving at the
first exit is
1
( 1)
n +
. Therefore, we can calculate the probability of the packet successfully
reaching the destination as:
1
1 1 1
1
1 1
i
k
i n
suc k
i i m
P
m C i
=
=
 
 
= − ⋅ ⋅
 
 
 
+ +
 
 
∏ (3)

71
It should be noted that sometimes mi might be smaller than k. In this case, the packet will be
routed out of the networks. In fact, this failure is caused by a network partition, so it will not be
considered in the performance of routing protocol. In most cases, mi is larger than k, so the
minimal value of Psuc can be obtained when m is equal to k. Now we have
1
1 1
1
1 1
k
i n
suc
i
P
k i
=
=
 
 
> − ⋅
 
 
 
+ +
 
 
∏ (4)
Let k equals to 3, then the Psuc will be within [0.95, 0.96]. Although the value of Psuc increases
as k increases, we believe the value of Psuc while k is 3 is already good enough for our network
routing protocol. Therefore, we choose k = 3 in the implementation of our GPSFR protocol.
4.2. Routing Algorithm
4.2.1. Distance Advance
In GPSR [16], the current node ni greedily selects one neighbor that is closet to the destination
as the next hop. The implicit goal of such strategy is to maximize the distance advance and
eventually minimize the total hop numbers. Let us denote such distance advance (ADV) of a
neighbor nj as
( )
0 ( )
ij
ij i
i
j
ij i
d
if d r
r
ADV
if d r

≤

= 
 >

(5)
where dij is the relative distance between node ni and nj. Whether nj is behind or in front of ni,
the value dij is always larger than or equal to zero. Clearly, the conventional geographic routing
protocols try to maximize ADV of the next hop.
4.2.2. Balanced Advance (BADV)
Balanced Advance (BADV) aims to avoid choosing an unstable node as the next hop while
gaining as much distance advance as possible. The goal of BADV is to balance between large
distance advance and good link stability.
(1 ) ( 0, 1)
ij ij
v v
d d
j j ij
BADV ADV e if v e
α α
∆ ∆
− −
= ⋅ + − ⋅ ∆ < = (6)
where ij j i
v v v
∆ = − is the velocity difference between nj and ni, and d is the distance from nj’s
current position to the edge of ni’s communication range. Therefore, suppose tj is the time used
by nj to move out of ni’s range; then a longer tj implies a more stable link between ni and nj. If
ij
v
∆ is less than zero, it means node nj is moving closer towards ni. In this case, we consider the
link stability as one because such link will become stronger until nj move into the different
neighbor group of ni. Since the beacon period is only a few seconds, ij
v
∆ will not change much
within such a short time. Thus, we can trust this value for at least one beacon period.
Although the concept of BADV is simple, it has many benefits in wireless vehicular networks.
First, the data delivery ratio will be increased because of reliable transmission links. While only
the distance advance is used, as in GPSR, the link to the selected next hop may suffer from a
poor quality due to larger distance. Second, the hit rate of finding next hop’s MAC address from

72
the cache table will be increased, so the times of ARP request and reply will be decreased.
Consequently, network delay can be reduced because of fewer retransmissions due to stable
links. Third, fewer changes in the next hop will reduce channel switching overhead. For
example, if [15] was adopted as the MAC protocol, then there will be a huge time slot allocation
overhead due to the frequent next hop change. However, if the data is an emergency message,
GPSFR will then use the maximal ADV policy by setting α as 1, because there is more benefit
to choose the shortest path than a stable one.
In summary, the GPSFR protocol will select a next hop with the maximal BADV to forward
normal packets. Besides, it will utilize the original greedy forwarding policy to transmit
emergency messages. When packets are routed around a exit ramp, the packets will be backup
three time to make sure they are not delivering out of the highway networks.
5. SIMULATIONS AND RESULTS
We use ns2 (ns2.29) to simulate and measure the networking performance of GPSFR. To
compare the performance of GPFSR with the prior work for vehicle ad hoc networks, we
choose the well-known GPSR [16] protocol. Since modeling of complex vehicle movement is
important for accurately evaluating protocols, we generated the movement of nodes using
VanetMobiSim [17] whose mobility patterns have been validated against TSIS-CORSIM, a well
known and validated traffic generator. In simulation, we focus on the two-lane two-direction
highway scenario with different node densities and velocities. Details of the simulation’s
parameters are list in Table 2.
Table 2. Parameters of simulation setups.
Parameter Value
Number of nodes 100
Communication range 250m
Velocity 65-80 miles per hour
Packet size 1024 Bytes
Data sending rate 1-8 packets per second
Beacon interval 5.0 seconds
Alpha 0.7
The simulation time is 2000 seconds and each scenario is repeated 20 times to achieve a high
confidence level. At each run, arbitrary vehicle pairs were selected as the source and
destination. To evaluate the performance on different data transmission density, we vary the
data sending rate from 1 to 8 packets per second. The performances metrics are link stability,
data delivery ratio and data delivery delay.
5.1. Link Stability
The link stability between two nodes is measured by the number of times wireless link breakage
occurs. As shown in Figure 5, GPSFR always generate less link breakage than that of GPSR
[16]. Network density is defined as the average number of neighbors at each node. In dense
networks, the number of link breakages in GPSFR is only 27% of that in GPSR. This is because
the next hop selected by maximizing BADV will be more stable, resulting in fewer changes in
the next hop. However, in sparse networks, GPSFR outperforms GPSR to a lesser degree. This
is because in spare networks, the number of candidate nodes that can be chosen as the next hop
is limited. So GPSFR may have no choice but to choose the same nodes as GPSR. However, it
still suffers from fewer link breakages. In Figure 5, the percentage value denotes the probability

73
of velocity change at each vehicle, which is used to model how dynamic the network is. As we
can see, the frequent velocity change of vehicles does not affect the link stability of GPSFR too
much.
Figure 5. Percentage of next hop changes at various different network densities
5.2. Routing Path Length
Since BADV considers the trade-off between stability and distance advance, the length of
routing path in GPSFR may be increased because of the slight reduction in distance advance at
each hop. Figure 6 and 7 presents the histograms showing the extra routing hops of GPSFR
compared to that of GPSR [16]. No matter how dense the network is, most of the routes in
GPSFR have the same length as GPSR. In addition, the longer routes are mostly one or two
hops more than that of GPSR. Therefore, to maximize BADV, we indeed increase the number
of hops by just enough to ensure higher data delivery ratio and lower network delay. Figure 6
shows the scenario where all vehicles are in cruise control (no velocity change). In Figure 7,
vehicles change their velocity all the time during the simulation. Note that no matter how
dynamic the network, GPSFR always delivers large majority of packets along the path with the
fewest number of hops.
Figure 6. Path length beyond GPSR when there is no velocity change

74
Figure 7. Path length beyond GPSR when velocity change probability is 100%
5.3. Data Delivery Ratio
Data delivery ratio is the number of packets received at the destination divided by the total
number of packets sent into networks. GPSFR3 (GPSR3) denotes that the neighbor time-out
period is 3 times the beacon period, while GPSFR (GPSR) means that the time-out period is
equal to the beacon period. Geographic routing in VANET may suffer from the problem of out-
of-date neighbors due to the high mobility of vehicles. One possible solution is to shorten the
time-out period of neighbors. Since the neighbors’ information is more accurate, higher delivery
ratio can be achieved, as shown in Figure 8. At each hop, GPSR [16] always try to maximize the
distance advance. However, the chosen one may be out of range after a short time, and this may
cause packet loss. On the other hand, in GPSFR only those that still are in range after
transmission will be considered as candidates. Therefore, the data delivery ratio of GPSFR is
higher than that of GPSR.
Figure 8. Data delivery ratio at various different data sending rate
5.4. End-to-End Delay
The end-to-end delay is defined as the average time taken for a packet to be transmitted from
the source to the destination. Figure 9 shows that the delay of GPSFR is much lower than that of

75
GPSR [16]. This is because links selected by GPSR are not as stable as those in GPSFR. Thus
link breakage happens more often in GPSR, requiring data retransmissions that increase delay.
Another reason is that there is a smaller ARP delay in GPSFR. For example, if the chosen next
hop is not a new one then the ARP request/reply process will not be required because the MAC
information of receiver can be retrieved from the cache table. GPSFR prefers to use stable links,
which means fewer changes in next hops. This reduces both the ARP delay and data
transmission delay. We also note that shortening the time-out period does not help to reduce the
delay because only successfully delivered packets are used for determining delay. Though
shorter time-out period can increase the number of successfully delivered packets, it does not
reduce the queuing, ARP and transmission delay.
Figure 9. Network delays at various different data sending rate
5.5. Impact of Routing Distance
The data delivery ratio of GPSFR and GPSR [16] will decrease as route length increases, as
shown in Figure 10. However, the delivery ratio of GPSFR is always higher than that of GPSR.
For high and low density networks, the performance of both protocols is measured as the
distance between the source and destination increases. For high and low density networks, the
average distance between vehicles are 50m and 75m, respectively. In all cases, GPSFR has
higher delivery ratio than GPSR. Note that GPSFR is also not as sensitive as GPSR to variation
in network density.
Figure 10. Data delivery ratio at various different routing path length

76
6. CONCLUSIONS
The GPS-Free Geographic Routing (GPSFR) algorithm uses relative positions and velocity to
achieve higher packet delivery ratio, lower delay and smaller per-node routing state than GPSR
[16], on densely and highly dynamic vehicular networks. Furthermore, it does not require
nodes’ positions. The BADV metric in this geographic routing ensures that only stable links are
selected, resulting in a higher data delivery ratio and lower delay. Actually, the performance of
GPSFR can be further improved if some nodes with GPS locations were added into the network.
Our future work is to design an enhanced protocol based on GPSFR to meet the communication
requirement of vehicles in urban areas. While we have shown herein the benefits of GPSFR as a
routing protocol for VANET, combining the GPSFR algorithm with a location database system
will further reduce the overhead in using external geographic information for routing. An
efficient distributed location service would enable the network to be more useful and powerful.
REFERENCES
[1] J. Zhao and G. Cao, (2008) "VADD: Vehicle-Assisted Data Delivery in Vehicular Ad Hoc
Networks," Vehicular Technology, IEEE Transactions on , vol.57, no.3, pp.1910–1922.
[2] H. Wu, R. Fujimoto, R. Guensler and M. Hunter, (2004) “MDDV: A mobility centric data
dissemination algorithm for vehicular networks,” in VANET ’04: Proceedings of the 1st ACM
international workshop on Vehicular ad hoc networks. New York, NY, USA: ACM Press, pp.
47–56.
[3] Q. Yang, A. Lim, S. Li, J. Fang and P. Agrawal, (2008) "ACAR: Adaptive Connectivity Aware
Routing Protocol for Vehicular Ad Hoc Networks," Computer Communications and Networks,
2008. ICCCN '08. Proceedings of 17th International Conference on, pp.1–6.
[4] Y. Ding, C. Wang and L. Xiao, (2007) “A static-node assisted adaptive routing protocol in
vehicular networks,” In Proceedings of the Fourth ACM international Workshop on Vehicular
Ad Hoc Networks VANET 2007, ACM, New York, NY, 59–68.
[5] R. M. Yadumurthy, A. Chimalakonda, M. Sadashivaiah and R. Makanaboyina, (2005) “Reliable
MAC broadcast protocol in directional and Omni-directional transmissions for vehicular ad hoc
networks,” in VANET '05: Proceedings of the 1st ACM international workshop on Vehicular ad
hoc networks. Cologne, Germany, pp. 10–19.
[6] M. Sadashivaiah, R. Makanaboyina, B. George and R. Raghavendra, (2005) “Performance
evaluation of directional MAC protocol for inter-vehicle communication,” in Vehicular
Technology Conference, 2005. VTC 2005-Spring. 2005 IEEE 61st. vol. 4, pp. 2585–2589.
[7] H. Füβler, M. Mauve, H. Hartenstein, M. Käsemann and D. Vollmer, (2002) “Poster: Location-
Based Routing for Vehicular Ad-Hoc Networks,” in Proc. of ACM MobiCom '02: Proceedings
of the 8th annual international conference on Mobile computing and networking, Atlanta,
Georgia.
[8] S. Capkun, M. Hamdi and J.P. Hubaux, (2001) “GPS-free positioning in mobile Ad-Hoc
networks,” in Proc. of Hawaii International Conference on System Sciences, Maui, HW, pp.
3481–3490.
[9] R. Iyengar and B. Sikdar, (2003) “Scalable and Distributed GPS free Positioning for Sensor
Networks,” in ICC '03, IEEE International Conference on Communication. vol.1, pp. 338–342.
[10] A. Rao, C. Papadimitriou, S. Shenker and I. Stoica, (2003) “Geographic routing without location
information,” in Proc. of ACM MobiCom '03: Proceedings of the 9th annual international
conference on Mobile computing and networking. San Diego, CA, USA, pp. 96–108.
[11] V. Kukshya, H. Krishnan and C. Kellum, (2005) “Design of a system solution for relative
positioning of vehicles using vehicle-to-vehicle radio communications during GPS outages,” in
Vehicular Technology Conference, 2005. VTC-2005-Fall. 2005 IEEE 62nd. Vol. 2, 25-28, pp.
1313–1317.

77
[12] Report and Press Release – The Economic Impact of Motor Vehicle Crashes (2000), posted on
National Highway Traffic Safety Administration website.
[13] C. Liberti and T. S. Rappaport, (1999) “Smart Antennas for Wireless Communications: IS-95
and Third Generation CDMA Applications,” Prentice Hall.
[14] Geometric Design of Highways and Streets, (2004) American Association of State Highway and
Transportation Officials.
[15] F. Borgonovo, L. Campelli, M. Cesana and L. Coletti, (2003) “MAC for ad-hoc inter-vehicle
network: services and performance,” in Vehicular Technology Conference, 2003. VTC 2003-
Fall. 2003 IEEE 58th. Volume 5, pp. 2789–2793.
[16] B. Karp and H. T. Kung, (2000) “GPSR: Greedy perimeter stateless routing for wireless
networks,” in MobiCom ’00: Proceedings of the 6th annual international conference on Mobile
computing and networking. New York, NY, USA: ACM Press, pp. 243–254.
[17] J. Harri, F. Filali, C. Bonnet, and M. Fiore, (2006) “VanetMobiSim: Generating realistic
mobility patterns for vanets,” in VANET ’06: Proceedings of the 3rd international workshop on
Vehicular ad hoc networks. New York, NY, USA: ACM Press, pp. 96–97.
[18] D. Deb1, S. Roy2, and N. Chaki3, (2009) "LACBER: A new location aided routing protocol for
GPS scarce MANET," International Journal of Wireless & Mobile Networks, vol.1, no.1, pp.22–
35.
[19] H. Akcan, V. Kriakov, N. Bronnimann, (2006) “GPS-Free node localization in mobile wireless
sensor networks,” Proceedings of the 5th ACM international workshop on Data engineering for
wireless and mobile access 2006, pp.35–42.
[20] A. Caruso, S. Chessa, S. De and A. Urpi, (2005) "GPS free coordinate assignment and routing in
wireless sensor networks,” Proceedings of 24th Annual Joint Conference of the IEEE Computer
and Communications Societies INFOCOM 2005, pp.150–160.
[21] H. Chu and R. Jan, (2007) “A GPS-less, outdoor, self-positioning method for wireless sensor
networks,” Ad Hoc Networks, Elsevier Science, vol.5, no.5, pp.547–557.
[22] K. Liu and N. Abu-Ghazaleh, (2006) "Aligned Virtual Coordinates for Greedy Routing in
WSNs," Mobile Ad hoc and Sensor Systems (MASS), 2006 IEEE International Conference on,
pp.377–386.
[23] J. Sheu, Y. Chang and G. Song, (2007) "Logical coordinate assignment for geographic routing in
wireless sensor networks," International Journal of Pervasive Computing and Communications,
vol.3, no.3, pp.274–288.
[24] T. Watteynea, I. Augé-Bluma, M. Dohlerb, S. Ubédaa and D. Barthelc, (2009) "Centroid virtual
coordinates – A novel near-shortest path routing paradigm," Computer Networks, Elsevier
Science, vol.53, no.10, pp.1697–1711.
[25] M. Tsai, H. Yang, B. Liu and W. Huang, (2009) "Virtual-coordinate-based delivery-guaranteed
routing protocol in wireless sensor networks," IEEE/ACM Transaction on Networking, vol.17,
no.4, pp.1228–1241.
[26] D. Reichardt, M. Miglietta, L. Moretti, P. Morsink, and W. Schulz, (2002). “Safe and
comfortable driving based upon inter-vehicle-communication,” CarTALK 2000, vol.2, pp.545–
550.
[27] F. Li and Y. Wang, (2007) “Routing in vehicular ad hoc networks: A survey,” Vehicular
Technology Magazine, IEEE, vol.2, no.2, pp.12–22.

78
Authors
Qing Yang received his B.E. and M.E. degree in computer science and technology
from Nankai University (China) and Harbin Institute of Technology (China) in 2003
and 2005, respectively. He is currently a Ph.D candidate in the department of
computer science and software engineering, Auburn University. He is the student
member of IEEE and ACM, and has been the Vodafone Fellow from 2005 to 2008.
His research interests include target tracking in distributed sensor networks, routing in
vehicular networks (VANETs), location security and privacy in VANETs and
performance measurement and analysis.
Alvin Lim is currently an associate professor of computer science and software
engineering at Auburn University. He received his Ph.D. degree in computer science
from University of Wisconsin at Madison in 1993. His research interests include self-
organizing sensor networks, mobile and pervasive computing, network security,
wireless networks, reliable and dynamically reconfigurable distributed systems,
complex distributed systems, mobile and distributed databases, distributed operating
systems, and performance measurement and analysis. He has published widely in
journals and conferences in these networking and distributed systems areas. He is a
subject area editor of the International Journal of Distributed Sensor Networks. His work had been
supported by the National Science Foundation, the DARPA SensIT program, U.S. Air Force Research
Lab and the U.S. Army.
Prathima Agrawal received her B.E. and M.E. degrees in electrical communication
engineering from the Indian Institute of Science, Bangalore. She received her Ph.D.
degree in electrical engineering from the University of Southern California in 1977.
She is the Samuel Ginn Distinguished Professor of Electrical and Computer
Engineering at Auburn University, Alabama. She is also the director of the Wireless
Engineering Research and Education Center at Auburn. Her research interests are
computer networks, mobile and wireless computing, and communication systems.
Earlier, she was assistant vice president of the Network Systems Research Laboratory
and executive director of the Mobile Networking Research Department at Telcordia Technologies
(formerly Bellcore), Morristown, New Jersey, where she worked from 1998 to 2003. Prior to this, she was
head of the Networked Computing Research Department at AT&T Lucent Bell Laboratories, Murray
Hill, New Jersey, where she worked from 1978 to 1998 in various capacities. She has published over 200
papers and holds 50 patents with several more pending. She is a Fellow of the Institution of Electronics
and Telecommunications Engineers, India, and a member of the ACM.

GPSFR: GPS-Free Routing Protocol for Vehicular Networks with Directional Antennas

More Related Content

What's hot (20)

Similar to GPSFR: GPS-Free Routing Protocol for Vehicular Networks with Directional Antennas (20)

Recently uploaded (20)

GPSFR: GPS-Free Routing Protocol for Vehicular Networks with Directional Antennas