Presentation on Ad Hoc Networks - PDF

Document Sample
Presentation on Ad Hoc Networks - PDF Powered By Docstoc
					Master’s Thesis: HUT, Networking Laboratory
Seminar presentation 3.9.2002




                      Choosing routing protocol for military ad hoc networks
                            based on network structure and dynamics

                                              Petteri Kuosmanen


                                  Supervisor: Ma. Prof. Samuli Aalto
                      The contests of the presentation

                      •       Ad hoc networks
                      •       Methodology
                      •       Modelled network structure and conclusions based on the model
                      •       Evaluation of network dynamics
                      •       Ad hoc routing protocols
                      •       Suitability of routing protocols
                      •       Proposal for routing protocol framework
                      •       Future work




Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002
              The subject of the thesis and its main goals


   Choosing routing protocol for military ad hoc networks
   based on network structure and dynamics

   Main question:

   What kind of routing protocol is suitable for military radio networks, if the structure of
   the network and its dynamics is known?

   The purpose is to find, not only one protocol, but protocol framework, which will fulfil the
   needs of the military network and its applications.




Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002
              Ad hoc networks and their application areas


   Ad hoc network is a mobile network, in where the routing is based on
   multihop routing from a source node to a destination node or nodes.

   In the ad hoc network there is no fixed infrastructure and the configuration of the
   network is autonomous, so it is suitable for many applications such as:

   -rescue, emergency and civil defense operations
   -conference and team working applications
   -military radio systems

                                                                S


                                                                                         D



Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002
                             Military ad hoc networks


   The basis of the ad hoc networks is in the military. Early research have been done
   by the DARPA and its PRNet and SURAN projects from the early 1970s.

   Now the emphasis of the research is in the commercial side; maybe ad hoc
   networks have some usable applications in mobile communications within
   a few years.

   Military applications will also benefit from the research of commercial
   applications, but there exist quite many features which are only typical for
   military networks.

   -> These features are studied in this research




Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002
                                       Methodology



Network structure and dynamics

                                                         Suitability of routing
                                       qualitative and   protocols and the needed
      Routing protocols                quantitative      features of protocols
                                       evaluations


           Evaluation criteria




  Petteri Kuosmanen
  Seminar Presentation. HUT 3.9.2002
                                 Modelled military organisation

                                                           Brigade


                     Battalion           Battalion         Battalion         Artillery          Artillery
                                                                             Battalion          Battalion


                                         Separate          Logistics        AA-artillery
                                          units              units           Battalion




                                     Brigade Battalion Company Platoon      Squad     Soldier Operational
                                                                                                 area
                        Brigade            1                                                         1600
                        Battalion          9           1                                               25
                        Company           36           4      1                                         4
                        Platoon          144          16      4         1                            0,25
                        Squad            432          48     12         3           1              0,0025
                        Soldier         3888         432    108        27           9        1




                        Simple model by which it is possible to evaluate node numbers and the
                        density of nodes in deployment area



Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002
                             The deployment of the organisation


                         Logistics                         By the deployment model it is possible to describe
                           units
Separate
 units                                   Brigade
                                                           the geographical distribution of nodes, which is the
                                           HQ              basis of the topology of the network
             Artillery         AA-




                                                   40 km
                             artillery
                                                           Also, it is possible to describe the distribution of traffic
                 Artillery                                 and network dynamics by the model

                                     Battalion
 Battalion       Battalion




                    40 km




 Petteri Kuosmanen
 Seminar Presentation. HUT 3.9.2002
                        Different generations in network topology

       There exists three generations in military networks:

       -1st gen. is based on radio networks constructed according to command chain
       -2 nd gen. is based on base stations and links between them
       -3 rd gen. is based on ad hoc principle




Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002
                                     The distribution of the nodes

  Depending on node penetration level, the nodes will be distributed on the deployment
  area differently. The problem is that the distribution will not be even.




   80 nodes.                                  112 nodes.              144 nodes.

   Distributed to:                            Distributed to:         Distributed to:
   -3 x 16 in battalions                      -3 x 16 in battalions   -3 x 16 in battalions
   -2 x 16 in artillery                       -2 x 16 in artillery    -2 x 16 in artillery
                                              -32 other nodes         -64 other nodes


Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002
                                  Trade-off between transmission power
                                             and throughput

Because of uneven distribution of nodes, there is not always the possibility to reach connections between
all nodes.

The increase of transmission range and thus transmission power will not help, because of channel
congestion.




    Petteri Kuosmanen
    Seminar Presentation. HUT 3.9.2002
                                     Distribution of the nodes
                                          - terrain effects
 Also, the terrain has its effects to connectivity. The terrain effects and uneven distribution of nodes
 have been simulated by a radio planning tool.




                                             The tool uses digital terrain elevation model, but
                                             does not take into account the effects of foliage.

                                             The stations have been placed on the same way as
                                             in the previous cases.

                                             The parameters are so that in free terrain the effective
                                             range of communications is over 10 km with the
                                             transmission power of 1 W.




Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002
                                      Needed transmission power


With the transmission power of 2 W it is able to reach good connectivity at the battalions,
but the total connectivity in the brigade area is very poor.




 Petteri Kuosmanen
 Seminar Presentation. HUT 3.9.2002
                                           Needed transmission power




By increasing the transmission power up to 10 W, then
the connectivity is fairly good also at the brigade area,
but at battalions there surely is disturbances between
nodes.

Conclusion is that there should be different power levels
in use at different areas.




      Petteri Kuosmanen
      Seminar Presentation. HUT 3.9.2002
                                        Field strength distribution


Also, the field strength distribution reveals the fact that by even transmission power and low
antenna height, it is not possible to guarantee needed communications services for all areas




               Transmission power 2 W                           Transmission power 10 W
                 Antenna height 2 m                                Antenna height 2 m



                  Solution is to use two levels of hierarchy in nodes:



   Petteri Kuosmanen
   Seminar Presentation. HUT 3.9.2002
                                     Two levels of hierarchy

 There could be two different kinds of nodes:

 -Lower level nodes are carried by soldiers and have a transmission power about 1 W. The antenna
  height is about 2 m and transmitters can have quite high frequency in use

 -Upper level nodes are carried by vehicles and have a transmission power about 10 W,
  even 50 – 100 W is possible and so the resistance against jamming is gained. The antenna height
  can be 2 m, but in quite stationary stations, antennas can be up to 10 – 15 m in height.




Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002
                                  Evaluation of network dynamics


The dynamics of the network has been measured by the link change frequency:
-every time an existing link breaks, broken link goes up or a new link emerge, there is an even
-the more events there is per time unit the more dynamic the network is

The network dynamics has been evaluated by calculating link changes caused by relative node
movements and by link failures caused by terrain

The movement inside the military organisation is controlled, so the arbitrary movement models are
not suitable for evaluating the network dynamics

According to the calculations, movement of nodes will cause about 0,12 link changes per second
per node if a node has about 6 neighbours in average and if it is moving about 30 km/h




 Petteri Kuosmanen
 Seminar Presentation. HUT 3.9.2002
                                           Evaluation of dynamics
                                              - terrain effects


The terrain effects have been simulated by a radio planning tool

According to the simulations the terrain has effects at the around
of 0,05 link changes per second, if nodes are moving 30 km/h

The terrain effect is around 40% of link changes caused by node
mobility

The total network dynamics is about 0,2 link changes per node




                   If the total dynamics is compared to simulation
                   results about routing protocols, then a normalisation
                   is needed


      Petteri Kuosmanen
      Seminar Presentation. HUT 3.9.2002
                              Normalisation of simulation results


Normally in simulations done by ns2-simulator,
mobility is defined by max speed and pause time

Normalisation is needed in order to compare link
change rates to this mobility measures

Normalisation was done by the ns2’s tool ”setdest”

E.g. pause time 400 s is equal to link change rate
of 0,1 Hz, if the node density, radio range and max speed
are same in both simulations



              0,2 link changes per second is moderate,
              but not very high dynamics from the point
              of view of protocols


   Petteri Kuosmanen
   Seminar Presentation. HUT 3.9.2002
                                                    Route lengths


   Average route lengths have been evaluated

   The routes seem to be quite short in average, this
   is one issue to be considered when choosing
   suitable routing protocol




                                        Number of     Percentage of
Route                        Hops        routes          routes       Total number of routes                83

Battalion to company           2             4             4,8        Relative share of route lengths (%)
Company to platoon             1             16           19,3        1 hop                                 19,3
Platoon to squad               2             48           57,8        2 hops                                62,7

Company to company             3             3             3,6        3 hops                                18,1

Platoon to platoon             3             12           14,5
                                                                      Weighted average number of hops       1,99



        Petteri Kuosmanen
        Seminar Presentation. HUT 3.9.2002
                                          Traffic flows



By rough evaluation of minimum hop routes, it seems
that the traffic will be concentrate on a few nodes

This causes congestion, if the routing do not take
into consideration average queue lengths



One reason why some kind of cost-function is
needed as a route selection criterion




     Petteri Kuosmanen
     Seminar Presentation. HUT 3.9.2002
                                          Routing protocols


There exists dozens of unicast routing protocols

Many protocols have different means to gain the
goals of the routing

The studied protocols have been classified according
to a taxonomy, which is able to reveal the implementation
and design choices of the protocols




  It is possible to evaluate the suitability of
  different protocol types




     Petteri Kuosmanen
     Seminar Presentation. HUT 3.9.2002
                                         Some protocols


Proactive protocols maintain all the time all possible routes
-high overhead and also routes, which are never needed are maintained

Reactive protocols find routes only when needed
-lower overhead, but longer delay when route is needed
-uses route request / route reply phases for route discovery
-route error messages are used to inform source node about link failures

Topology based protocols maintain large scale network information in all nodes
Destination based protocols only maintain information to reach the nearest neighbours
-> scalability is better with destination based protocols

In very large networks some kind of differentiation between nodes is needed
-hierarchical protocols maintain e.g. clusters and other control structures
-in neighbour selection protocols nodes divide network to remote and near nodes, and
 there is some differences in the handling of these nodes




    Petteri Kuosmanen
    Seminar Presentation. HUT 3.9.2002
                                           Some protocols
                                           - scalability issues
Neighbour selection protocols seem to have interesting
properties in scalability

Zone Routing Protocol (ZRP) is a hybrid protocol, which combines
proactive and reactive routing schemes
-proactive routing is used to interior nodes
-reactive routing is used to exterior routes

Fisheye Routing Protocol (FSR) is an extension to be used with link-state
protocols (e.g. GSR)
-link state update interval is different for near and remote nodes

Also the use of location as a aid of routing is a promising feature to
improve scalability
-Location Aided Routing (LAR) uses position information to reduce
 control overhead




      Petteri Kuosmanen
      Seminar Presentation. HUT 3.9.2002
                                              Some protocols
                                          -framework structure


Framework structure used in the ZRP-protocol enables the
use of different routing schemes and improves the flexibility
of the protocol

Routing in the military environment should support e.g.
the following features:
-QoS-routing
-Multiple routes between end nodes in order to improve bandwidth
 and reliability
-Location information to reduce control overhead

Framework structure to be used?




     Petteri Kuosmanen
     Seminar Presentation. HUT 3.9.2002
                                       Suitability of routing protocols


Protocols were evaluated by comparing different
published simulation results

According to the simulations and qualitative
evaluations the suitability of protocols can be
classified as seen by the chart

The evaluated military network is situated somewhere
in the right middle part of the suitability chart




                                                           Military networks




      Petteri Kuosmanen
      Seminar Presentation. HUT 3.9.2002
                            Proposal for routing protocol framework

The main idea is to use two levels of hierarchy in nodes, but flat routing scheme based on
neighbour selection structure:
-two levels is feasible because of economical and functional reasons
-flat routing is simple and robust

The use of hybrid, neighbour selection protocol such as ZRP
is based on short average routes, low delay of often needed routes
and good scalability features of reactive routing

The scalability can be improved by location information in use

                                           Feature                       Lower level        Upper level
                                           Transmitter power             1W                 50-100 W
                                           Transmission range            1 km               10-15 km
                                           Antenna                       Omni-directional   Directional, maybe adaptive
                                           Communications model          Single-channel     Multi-channel
                                           QoS-support                   No                 Yes
                                           Position information in use   No                 Yes
                                           Multiple routes in use        No                 Yes
                                           Route selection criteria      Shortest path      Multiple criteria


      Petteri Kuosmanen
      Seminar Presentation. HUT 3.9.2002
                         Proposal for routing protocol framework

Upper level nodes have probably better S/N and thus bandwidth, so if cost-based route selection
method is in use the upper level acts as a network core

With the core it is possible to use QoS-management system such
as e.g. in Core Extraction Distributed Ad hoc Routing protocol (CEDAR)

Methods for configuration information exchange and for
power control is also needed




   Petteri Kuosmanen
   Seminar Presentation. HUT 3.9.2002
                           Configuration information exchange
When two upper level nodes come to contact with each other, then the following information
(at least) is exchanged between nodes:
-Node IDs (e.g. IPv6 address in use)
-Node positions (x, y, z)
-Traffic code between nodes (WCDMA in use)
-Power distribution towards this peer node (percentage of total power)

The negotiation is conducted by using general code and after that the nodes are able to construct a link
and use communications code which is orthogonal to neighbouring codes

By controlling the power distribution it is possible to control available bandwidth of a link

Node positions can be used to evaluate terrain effects to a link



                   Code ”H”
                                       B
        A                                                   Multi-channel technique complicates the system, but
                                                            it is probably the only way to gain high available
                                           Code ”G”         capacity in the network
               Code ”F”
                                  C


  Petteri Kuosmanen
  Seminar Presentation. HUT 3.9.2002
                                            Power control

The meaning of the power control is to allocate transmission power,
so that S/N-ratio is the best possible for all nodes
-mutual disturbances are alleviated
-good signal levels are acquired on receivers

Control mechanism could be based on reaching suitable number of
neighbours, but also some signalling is probably needed

The use of adaptive antennas at the upper level nodes makes the
issue quite complicated
-it is perhaps possible to use node locations and approximations
about terrain effects to help the control mechanism




       Petteri Kuosmanen
       Seminar Presentation. HUT 3.9.2002
                                     Future work

         Still very many open issues on the field of military ad hoc networks
         - routing is only one problem, but very important
         - the routing is not alone, also lower level layers have effects to routing

         Developed routing schemes are suitable for use in military networks, but the military
         environment is different from commercial applications, so the features of military
         networks must be known when adapting routing protocols to this field

         The proposed routing scheme should be simulated at different levels and it should be
         developed together with physical-, MAC- and transportation layer design

         Paths to ad hoc networks will go through base station networks i.e. 2G will co-exists
         with the 3G. Also, semi-ad hoc networks are one obvious way of development.




Petteri Kuosmanen
Seminar Presentation. HUT 3.9.2002

				
DOCUMENT INFO
Description: Presentation on Ad Hoc Networks document sample