Mobile Ad hoc Network (MANET) (Giordano and Lu, 2001; Scaglione et
al., 2006; Kim and Toh, 2006) comes from the Defense Advanced Research
Projects Agency (DARPA), which is a network that each node processes packet
routing without infrastructure. The fast equipping and mobile self-organization
characters push MANET the hot point in wireless communication, especially
the route protocol, e.g., now Internet Engineering Task Force (IETF)`s
MANET working Group is searching for the comments (Mnaouer et al.,
2007; Perkins et al., 2003).
Today routing protocol is approximately divided into table-driven routing
protocol and On-Demand routing protocol (Lee et al., 1999; Kim
et al., 2006) according to its trigger mechanism. Table-driven
routing protocol is based upon the established routing tables, thus owns
small delay. Its virtue is to not wait for establishing route while information
are transmitted and delay is small. Meanwhile, on-demand routing protocol
looks for routing when demanded, where the node usually needn`t maintain
routing technically. Hence it is suitable for MANET and is widely researched
On-demand routing is usually made up of two processes named routing set
up and routing maintain. When nodes need transmit information, they broadcast
route set up packets and search the optimal path without storage of the
routing information, resulting in effective bandwidth and memory save.
But it is apparently at the cost of delay due to the set up procedure.
Accordingly, this style represents in Ad hoc On-Demand Vector (AODV) (Perkins
et al., 2003).
This study presents a novel AODV routing protocol based on mobility prediction
(MAODV). It can control node route by keeping and switching route through
estimating the neighbor node`s distance and predicting the neighbor node`s
mobility to fit fast changed network topology. This mechanism reduces
the end-to-end delay effectively and enhances the real-time character,
which is very important to the voice communication and the video communication.
MOBILITY PREDICTION ALGORITHM
Figure 1 shows a sketch map on mobility prediction.
Node D moves to node D` in Velocity V. The radial distance varies from
d to d` relative to node S and the radial velocity act as Vr
and V`r. Thus the movement difference is the Δd = dd`.
|| Sketch map on mobility prediction
After calculating the received power Pr, d is determined by
Here, we suppose that all nodes have the same transmission power pt.
The neighbor node`s velocity is calculated by formula 2:
where, Δt is spacing time for twice measurement. By the positive
or negative of
value, we may determine movement direction parameter of neighbor nodes
that are outward node S or inward node S for instance (3).
By estimating space distance d, neighbor`s movement velocity v and neighbor`s
movement direction, we may predict effective communication area whether
node leaves off one another. If node`s movement direction is outward and
d is already close to effective communication radius and v is relatively
large, we may in advance start up routing detection and establish a new
route before current path breaks off. In the suitable time path switching
will be finished. So we may decrease delivery delay duo to suddenly route
breaking off by prediction technique effectively.
According to analysis above mentioned, we provide one formula
4 for estimating effective communication area that neighbor node will
there d denotes neighbor node`s distance at some time, Dcomm denotes
effective communication distance, Vr is the radial velocity
and Tmin denotes a new routing detective time estimated. When
neighbor node`s distance d satisfied with formula 5,
a new routing detection is stated up. Before primary routing path breaks
off, a new routing will set up.
MAODV PROTOCOL SIMULATIONS
This study uses Network Simulator 2 (NS2) (Cavin et al., 2002).
In NS2, we choose propagation model based on 802.11 and apply lucent`s
WaveLAN. Under these simulation conditions, the wireless node is configured
||End-to-end packet delay vs node`s maximum velocity
|Address resolution protocol
|AODV and MAODV
|Two-ray ground model
In simulations, we use Constant Bit Rate (CBR) and choose effective communication
area of 250 m. Moreover, the mobility scene adopts CMU`s generator. Then
in the 1500x300 m scenes, 50 nodes and 20 CBR, of which the length of
date packet chooses 64 and 1024, respectively. The node`s maximum mobility
is 1, 5, 10, 15 and 20 m sec-1.
From Fig. 2 , we may obtain end-to-end packet delay
in the MAODV protocol superior to in the AODV protocol. The decrease of
delay is mainly introduced by route updating predicted. Because route
updating is predicted, the delay by route breaking off will Greatly reduce.
So average delay character in the MAODV will be improvement.
Figure 3 shows the relationship of the packet delivery
ratio vs node`s mobility velocity, where packet delivery ratio is the
ratio of received packets vs transmitted packets. As the maximum mobility
velocity increases, packet delivery ratio in AODV and MAODV all decrease.
But decreased value is not dramatically. Simultaneously the MAODV follows
on-demand routing character and its packet delivery ratio will slightly
inferior to AODV. After setting up prediction path, the MAODV don`t process
further for network topology. The prediction algorithm that is finds out
is the temporal excellent path. By memory, the path may be do not the
most stable path. Meanwhile the error between the prediction algorithm
and the actual status will bring some packet loosen ratio. Therefore under
the complex network
|| Packet delivery ratio vs node`s maximum velocity
|| Control overhead vs node`s maximum velocity
circumstances, the MAODV will decrease packet delivery ratio because
the action of MAODV protocol reply network topology.
From Fig. 4, control overhead will increase with the
increases of node`s mobile velocity. This is because route switching varies
business with the increases of node`s mobile velocity. Thus the routing
overhead rises and control overhead increase. The control overhead of
the MAODV will slightly outgo to AODV, which is bring because of additional
consume of the path prediction.
Form Fig. 5, routing average numbers of hops do not
dramatically vary with the increases of node`s mobile velocity. Average
numbers of hops of the MAODV routing are slightly outgo to AODV. This
is because the MAODV adopts the predictive algorithm. The selected predict
paths do not always the best path of the route switching. Nevertheless
the AODV adopts the routing strategy that set up route searching process
after routing is break off. So the AODV routing is not corresponding
||Average number of hops vs node`s maximum velocity
delay than the MAODV. Sometimes the selected AODV path is more get well
than the MAODV. This is because the average numbers of hops of the AODV
is small than the MAODV. What is a cost that the MAODV can decrease end-to-end
packet delay adopted predictive algorithm.
This study presents a novel MAODV routing protocol based on mobility
prediction combined with the AODV routing protocol. The algorithm controls
route discovery, route keeping and route switching by estimating the neighbor
node`s distance and predicting the neighbor node`s mobility. Then we introduce
NS2 to verify the algorithm in terms of the average end-to-end packet
delay, the packet delivery ratio, the control overhead and the average
number of hops.
As a result, the MAODV reduces the end-to-end delay effectively and enhance
the data transmission rate, which is very important to real-time communication
such as the voice communication and multimedia image communication. On
the other hand, the MAODV will reduce packet delivery ratio, increase
control overhead and increase average numbers of hops. However, comparing
with the obvious real-time performance improvement, our MAODV affects
on the other third performances slightly.
The research is supported by the ZJNSF (Grant No. Y106162), China.