CSEM Scientific and Technical Report 2008

Efficient Information Dissemination in Mobile Wireless Ad-hoc Networks
D. Piguet, J. Rousselot, C. Kassapoglou-Faist
Simulations of information dissemination techniques based on probabilistic broadcast have been performed. Their results will drive the
implementation of collaborative information exchange between mobile nodes in the context of the IPAC project.
Heterogeneous wireless ad-hoc networks, possibly made of
handheld devices, vehicles with communication capabilities
and wireless sensor nodes (Figure 1) are the center of interest
of the European project IPAC. This project aims at creating a
generic platform for mobile nodes that exchange information
in a collaborative way. IPAC benefits from the expertise of
CSEM in low-power wireless communication protocol design
for the development of its information dissemination solution
based on adaptive probabilistic broadcast. The range of
possible applications of such heterogeneous networks is wide:
crisis management, road safety, supply-chain optimization are
the examples that will be demonstrated at the end of the IPAC
project.
Figure 1: A network made of heterogeneous devices, using different
communication technologies source: IPAC.
For such applications which are highly dynamic, traditional,
address-based routing protocols may not be optimal: for
instance, in a disaster recovery situation, a rescuer who finds
an injured person does not necessarily know the network
address of an ambulance. Instead, a content based routing
protocol may allow the rescuer to disseminate its message
throughout the network to the interested destinations, an
ambulance in this case. Note that the task is complicated by
node mobility as links for node to node change over time.
The challenge of information dissemination is to ensure that
messages are very likely to reach whoever should receive
them while keeping the number of transmissions as low as
possible in order to minimize energy consumption and reduce
the risk of network congestion. Efficiency and high delivery
ration have to be provided under node mobility.
The solution proposed by CSEM and its partners assumes
that all messages are disseminated to all nodes. Messages
are delivered to the application running on a node based on
local subscriptions reflecting the application interest. Flooding
is a possible solution to this dissemination.
Traditional flooding is a very redundant technique: every node
which receives a message re-transmits it (using local
broadcast) once only. A possible solution to reduce
redundancy is to modify the node behavior: a node that
receives a message retransmits it with a certain probability
using local broadcast. This probability depends on the
perceived network characteristics (node density, traffic,
86
mobility, …); if nodes choose their probability of broadcast
wisely, there is good hope that a message that is sent will
spread over the entire network at a reduced cost compared to
traditional flooding. This routing method is called Adaptive
Probabilistic Broadcast. Its main advantages are:
• No overhead due to control messages: nodes decide on
their probability according to solely local rules
• Scalability
• Lightweight implementation
• The probability factor is primarily influenced by the
network status. It can also depend on indications from the
application (importance, priority of a message, …)
• To improve flexibility, another parameter has been added:
for a given message a node may attempt to forward it (flip
the coin) several times, for example to improve the
reliability of an important message.
The goal of the study that was conducted was to derive the
transmission probability calculation rules. In order to
understand the influence of different network parameters,
simulations of non-adaptive probabilistic broadcast were
performed: for each simulation run, the nodes were assigned
a probability and a number of broadcast attempts. The
simulated protocol was implemented in the Omnet++
simulator. They took advantage of the already existing realistic
physical and MAC layer models previously developed at
CSEM. Many different network configurations were compared,
i.e. number of nodes, density, mobility and traffic.
The results show that the coverage of the network mainly
depends on network density, that is on how many peers a
particular node will find in its direct range. For a given network
density, there is a minimal probability that will ensure full
network coverage. Further increasing the probability is
useless and only adds more redundancy. Varying mobility did
not allow the derivation of a clear rule regarding probability: in
certain cases, mobility favors network connectivity, in other
cases, it harms it. However, reliability can be greatly improved
by increasing the number of broadcast attempts with mobility.
Simulations of the adaptive algorithm which were derived from
the preliminary study show that, if the maximal number of
broadcast attempts is set to one and compared to traditional
flooding, 20% (for a minimal density but connected network) to
84% (for a dense network, all nodes in sight of each other) of
packets transmissions can be saved while 98% to 100% of the
messages reach their destinations.
The project partners are Siemens A.E. Electrotechnical
Projects and Products (GR), National and Kapodistrian
University of Athens (GR), CENTRO RICERCHE FIAT
S.C.p.A. (IT), University of Cyprus (CY). CSEM thanks the EU
Seventh Framework Programme for founding this project.