HAL Id: hal-01174850
https://hal.archives-ouvertes.fr/hal-01174850
Submitted on 10 Jul 2015
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires
Definition and specification of ”CANlike” protocols in the context of wireless networks
Guy Juanole, Gérard Mouney
To cite this version:
Guy Juanole, Gérard Mouney. Definition and specification of ”CANlike” protocols in the context of
wireless networks . [Research Report] LAAS/CNRS. 2015. �hal-01174850�
ontext of wireless networks
GuyJuanole
1,2
, GérardMouney
1,2 1
CNRS, LAAS, 7 avenuedu ColonelRohe, F-31400, Toulouse,Frane
2
Univ de Toulouse,UPS, INSA, INP, ISAE, LAAS,F-31400 Toulouse,Frane
juanolelaas.fr, mouneylaas.fr
Abstrat
Theimplementationofdistributedreal-timeappliationsonwirelessnetworksonstitute
todaya newimportanthallengeand,inthisontext,theMACprotools, whihimplement
theframeexhangesheduling,haveanessentialrole. Thispaperispreisely onernedby
thespeiationofsuhMACprotools. WespeifyMACprotoolsalled CANlikeproto-
ols beause theyareinspired by theMACprotool of thenetwork CAN whih is a wired
network. Thepresentationmadeinthispaper,afterareminderofbasiknowledges(wireless
networkphysiallayer,dierenttopologies,CANwirednetwork,MACprotoolharateris-
tis)showshowtointegratethesebasiknowledgesinordertospeifytheCANlikeprotools
forseveraltopologies(mono-hoptopology andthreedierentmulti-hoptopologies(hains)).
In the onlusion too, weprove (by onsidering a mono-hop topology) the interest of the
CANlikeprotoolsforimplementingappliationsin networkedontrolsystems (byompar-
isonwiththeWiFi-DCF protool).
1 Introdution
Wireless networks andmore partiularlyWireless LoalAreaNetworks(WLANs) aremoreand
more used today inthe industrial area where we have real-time distributed appliations whih
requireQualityofServie(QoS) guaranteesfortheir ommuniations. Inthis ontext,theMAC
protools, whih implement theframe sheduling, have anessential role. WLANsan be either
mono-hannelor multi-hannel. Hereweonsider themono-hannelase.
In the ontext of the wireless networks, the protools of the CSMA (Carrier Sense Multiple
Aess) type and, partiulary, with the attribute CA (Collision Avoidane) [1℄ are very often
onsidered and used. The attribute CA is based on a Bakoproedure whih allows, in om-
parison to the strit CSMA type, to redue the ollision ourrene but not to eliminate this
ourreneand thenwe annotgiveQoS guarantees fortheframe transfer.
Mastering the ollisions and giving QoS guarantees is possible by assoiating priorities to the
frames of the ows (the role of the priorities is to allow to implement a Collision Resolution
(CR)mehanismi.e. to transformwhatwouldbeaollision situation withaCSMAtypepro-
toolinto awinner-looser(s)situation whihresultsfrom atournament basedonthepriorities
omparison; the winner is the frame whih has the highest priority). The rst approah is to
use the BlakBurst tehnique [2℄. The idea is to let the ontending nodes send rst jamming
signals (alled BlakBurst (BB) messages)of length aordingto thepriority. The node whih
hasthelongestjammingsignal(i.e. thehighestpriority)winstheompetitionandthensendsits
frame. Thedrawbakofthis tehnique isthat, if we havea greatprioritynumber,thejamming
signals will be very long and give important delays [3℄. The seond approah is to adapt the
MACprotoolofthewiredCANbus(thepriorityoftheframeisexpressedbytheIDeldwhih
this paper(oneptof CANlike protool).
This paperinludes threeparts:
•
therstpart presentsbasi knowledges,•
theseond part onernsmainly the speiation of themain parametersof the CANlikeprotools for dierentwireless networktopologies; itpresentsalso solutionsfor aproblem
whih ours in a hain topology and whih is alled the intraow problem (onurrent
frame transferinaframe owgoingfrom asoure nodeto a destinationnode).
•
thethird partisa onlusion.2 Preliminaries: Basi knowledges
Three types of basi knowledges are neessary. The type 1 onerns the harateristis of the
wireless networks physial layer and some important onsequenes for the MAC layer with a
protool of the CSMA type (pure CSMA or CSMA-CA). The type 2 onerns dierent node
interonnetion strutures (i.e dierent topologies) in a wireless ontext. The type 3 onerns
thepriniplesthat underlietheCAN-like protools for thedierenttopologies whih havebeen
onsidered.
2.1 Type 1 of the basi knowledges
2.1.1 The wireless transeiver
Inawireless ontext(ontrarilyto thewiredontext),atranseiverannotsimultaneously send
and reeive on a hannel and has three states: transmitter, reeiver, sleeper. Here we do not
onsiderthestate sleeper whih isusedfor onsiderationsof energyeonomy.
Two time attributes haraterize the transeiver behavior: the hannel Sensing Time
τ ST
andtheTurnaround Time
τ T T
.τ ST
allows thetranseiver(inthereeiverstate)to test thehannelstate (busy oridle)dependingon whetherthedetetedEnergy on
τ ST
ishigher or lowerthanaprexedthreshold (noted
Ethr
).τ T T
is thetime to go from thereeiver(transmitter) state to thetransmitter(reeiver)state.Ifthehannelisdetetedidle,thetranseiverango(aftera
τ T T
)inthetransmitterstatewhih allowstheMACentitytosendaframe. Aftera frametransmission, thereeiveran(afterτ T T
)ome baktothereeiverstate.
Relativelyto aframetransmission(byonsideringframeswhereallthebitsuses thesame ode,
and thenhave,fromthepowerpoint ofview,identialtransmissiononstraints),thehannelis
denedbymeans of two parameters(bandwidth, signal reduing)and thenthe transmissionof
anodeisharaterized, interm ofthesignalreduing(withrespetto thepowerofthesignalof
theemitted frame)bytwo ranges: CarrierSenseRange(
R CS
) and TransmissionRange(R T
).2.1.2 Carrier Sense Range (
R CS
)The
R CS
,whihisassoiatedtoanodei
(notedR CS (i)
),isrepresentedbyairleofenteri
andofradius noted
r CS (i)
. Theradiusr CS (i)
is themaximal rangeinwhihthesending of aframebythe node
i
indues for all nodej
being inthe irle, thedetetion of a signal, the Powerofwhih ishigher thanor equaltoa threshold noted
P (R CS (i))thr
(theprodutofP (R CS (i))thr
by
τ ST
gives the thresholdEthr
, i.e the limit of the detetion of a busy hannel state after aframetransmissionbythenode
i
). Notethat,thefatthatanodej
,intheirleR CS (i)
,detetsasignalresultingfromthesending ofa framebythenode
i
,doesnotmean neessarilythatthenode
j
isableto deodethis frame(thatdependsonthedistaned ij
). Thisremark justiestheneessityto introduetheonept ofTransmissionRange(
R T
).Thedenition ofthe
R CS (i)
requiresstillto preise thefollowingpoints:1. thenode
i
isalledexposednode to allthenodesj
whihareintheR CS (i)
(beause thetransmission of a frame by the node
i
induesthe busy hannel state whih prevents allthenodes
j
to usethehannelduring thistransmission duration),2. thenodes
j
arethenodeswhih,intheframeworkof theR CS (i)
,areinompetition withthenode
i
for thesending of aframe. Morepreisely:•
if onenodej
startsa transmissionjustbeforean attemptofthenodei
,thisindues,forthenode
i
,thesituationbusyhannel whihdelaysitspossibilityoftransmission,•
if onenodej
andthenodei
transmitssimultaneously,thisinduesasituation emis- sionollision.We have tonote thatthese twosituations arenormalsituationsbydenitionof thestrit
ontext CSMA. We desribe,relativelyto thenode
i
,these two situationsasendogenousinterferenes beause they resultfromations ofthenodes
j
whih areintheR CS (i)
.3. nodesan be outsidethe
R CS (i)
. Amongthesesnodes, some ofthem an have theirR CS
whihhave an intersetionwith
R CS (i)
. Callk
suha node andR CS (k)
itsCarrierSenseRange, and rename
jj ′
the nodesj
whih are at the intersetion ofR CS (i)
andR CS (k)
.The nodes
jj ′
an hear the attempts of thetransmission of the nodesi
andk
whih anthenreate,inthesenodes
jj ′
,interferenesituationsthatweallexogenousinterferenes (beauseresultingofationsof thenodesi
andk
,whiharenotinthesame CarrierSenseRange). This harateristi willhelp us topresent thehiddennode problem.
2.1.3 Transmission Range (
R T
)Consideragainanode
i
and itsassoiatedR CS (i)
. TheR T
,whihisalsoassoiated tothenodei
(notedR T (i)
), isrepresentedbya irleof entrei
and ofradiusr T
,notedr T (i)
. The radiusr T (i)
is the maximal range in whih the sending of a frame by the nodei
sets, for all nodesj
being in the irle,the detetion of a signal, thepowerof whih is higher thanor equalto the
power neessaryto deode theframe sent bythenode
i
i.e. a power higher thanor equal to athresholdnoted
P(R T (i)thr)
. ObviouslyP (R T (i)thr) > P (R CS (i)thr)
.Remark: In pratie,generally, we have
r CS (i) > r T (i)
whih an still be expressedR CS (i) >
R T (i)
. However, we an also onsider a partiularase whih an be expressedP (R T (i)thr) = P (R CS (i)thr)
and thenR CS (i) = R T (i)
(i.e. any node in theCarrierSense Rangeof thenodei
an deode thesignaloftheframe sent bythenodei
). Inshort,wean sayR CS (i) ≥ R T (i)
.2.1.4 Transmission hop
A transmission hop is the basi element of a ommuniation path between omputers i.e. it
represents, in the framework of an implementation, the distane (noted
d
) between a node,transmitterof a frame, andthe next node, inthe path,whihreeivesdiretly and deode this
frame. Wehave :
d ≤ r T < 2d
i.e. we an havea pathof one hopinR T
and obviouslymore inR CS
whenr CS ≥ 2d
.2.1.5 Hidden node
a)Consideragainthepresentationinthesubsetion2.1.2andonsidertheaseofatransmission
of a frame in one hop from thenode
i
to a nodejj ′
(then this frame will be well reeived andwell deoded inthenode
jj ′
if there isno kind ofinterferene). A node hidden to thenodei
isa node
k
[4℄, [5℄, beause the nodek
is a node exposed to the nodejj ′
(i.e the nodejj ′
is inthe
R CS (k)
), whih an lead, relatively to an attemptof a frametransfer bythe nodei
,to twoexogenous interferenesituations inthereeptionativityof thenode
jj ′
:•
situation 1: asituation alledbusyhannel resultingfromthesending ofa framebythenode
k
before thesendingattempt bythenodei
(butthis one annotseethestate of thehannel asthenode
k
is notin itsR CS
); the result will be the non onsideration, by the nodejj ′
,oftheframe omingfromthenodei
(thenits loss),•
situation 2: a situation alled ollision in reeption whih results from a simultaneoussending ofa framebythenodes
i
andk
,whih anlead, on theframesentbythenodei
,to thedeodingimpossibilitybythenode
jj ′
.The ourrene of the situation 2 depends, at the node
jj ′
, on the ratio Signal Power of theframe oming from the node
i
(allP i
this power) on Signal Power of the frame oming fromthenode
k
(allP k
thispower). Byallingd
thelength ofthehopi, jj ′
andl
thedistanek, jj ′
we have:
P
iP
k= ( d l ) 4
. The ondition for a orret deoding of the frame sent by the nodei
isP
iP
k≥ 10
[6℄,whih denesthelimit valueofl
alledInterferene Range etnotedR I
. WehaveR I = 1.78d
.b) We an now give the quantitative onditions [6℄ whih express the behaviour of a node
k
hiddento thenode
i
. Asitis anode outsidetheR CS (i)
,wehaved + l > r CS ( i )
.•
Ifl ≤ 1.78d
,we an have thesituations1 and 2,•
Ifl > 1.78d
,wean onlyhavethesituation 1. Thesituation 1an alwayshappen beausethenode
jj ′
is intheR CS (k)
.2.2 Type 2 of the basi knowledges
Weonsidertopologieswherealwaystheframes,exhangedbetweenallthenodes,useforalltheir
bits the same ode (i.e. all the bits have, from thepowerpoint of view,identialtransmission
onstraints)andthenthetransmissionofall thenodesareharaterizedbythevalues
R CS
,R T
and
d
. This situation is the ase of the protools of the CSMA type (pure CSMA or CSMA-CA).We an have eithertopologies,alledmono-hop topologies,or topologies alledmulti-hop
topologies(important examplesarethehainsthat we onlyonsider here).
2.2.1 Mono-hop topologies
Mono-hop topologies aretopologies where eah node an ommuniate diretly(one hop) with
all the other nodes. In suh a topology, all the nodesare in theintersetionof their range
R T
(andobviouslytoo oftheir range
R CS
asR CS ≥ R T
;herewetakeR CS = R T
). So wehave notthehidden node problem. Thisdenes full-meshed topologies. On thegure 1we represent an
exampleof suha topologywhihis madeupof
4
nodes1
,2
,3
,4
whereeahnodeis theenterof airle ofradius equalto
r T
.Inthis topology,we an only have endogenousinterferenes.
2.2.2 Chains (Multi-hop topologies)
Note that, for drawing size reasons, we only represent the
R CS
ranges and, furthermore, their irles arerepresentedbyellipses.We onsider nodeswhere theradius of the range
R CS
an inlude at themosth
hops(h ≥ 1
).We dene three types of hains. The rst one with
h > 1
(noted hain-1) is a hain where3 1
4 2
R
CS(3) = R
T(3) R
CS(1) = R
T(1)
R
CS(2) = R
T(2) R
CS(4) = R
T(4)
Figure1: Mono-hop topology(fullmeshedtopology)
all the nodes are in the intersetion of their range
R CS
and then we have not still the hiddennodeproblem (beausenone node isoutsidethe ranges
R CS
of theother nodes). Weonly haveendogenous interferenes. Onthe gure2,we represent an example of suh atopology(hain
of3hops)whihismadeupof4nodes
1
,2
,3
,4
wheretheradiusoftherangesR CS
ofthenodesinludeat themaximum3hops(
h = 3
). ObviouslyR T (i) < R CS (i)
.1 2 3 4
d d d R
CS(3) R
CS(4)
R
CS(2) R
CS(1)
Figure2: hain-1 (4 nodes, 3hops)
Thetwoothertypes(notedhain-2andhain-3)havethehiddennodeproblembeausenodes
areoutsidetherange
R CS
ofothernodes. Wedistinguishtwoases aordingtothevalueofh
:h = 1
haraterizes a hain notedhain-2 where we onsiderR T = R CS
;h > 1
haraterizes ahainnotedhain-3 whereobviously
R T < R CS
.We represent,on thegures 3 and4, respetivelyan example of thehain-2 and anexample of
thehain-3 with
h = 2
(the two hains have7
nodes numbered from1
to7
). We didnotdrawthe
R CS
ofall thenodesfor reasonsofgure larity.d d d d d d
1 2
3
4
5
6 7
R
CS(7) = R
T(7) R
CS(2) = R
T(2)
R
CS(1) = R
T(1)
Figure3: hain-2
Wean easily see
•
on the gure3, the nodes (i + 2
) are thehidden nodesof the nodesi
(i ∈ [1, 5]
) and thenodes(
i − 2
) arealsothehidden nodesof thenodesi
(i ∈ [3, 7]
),•
on the gure4, the nodes (i + 3
) are thehidden nodesof the nodesi
(i ∈ [1, 4]
) and thenodes(
i − 3
) arealsothehidden nodesof thenodesi
(i ∈ [4, 7]
).d d d d d d
1 2 3 4
5
6
7
R
CS(1) > R
T(1) R
CS(4) > R
T(4) R
CS(7) > R
T(7)
Figure4: hain-3
It is important to note thedierene in the role of the hidden node depending on whether we
havea hain-2 ora hain-3 (see2.1.5)
•
hain-2: asthehiddennodei + 2
(ori − 2
) ofa nodei
isat thedistaned
(i.e.< 1.78d
)of the node
i + 1
(ori − 1
), we an have the two situations 1 and 2 of the exogenousinterferenes,
•
hain-3: asthehiddennodei + 3
(ori − 3
)ofa nodei
isat thedistane2d
( i.e.> 1.78d
ofthenode
i + 1
(ori − 1
),weonly have thesituation1 of theexogenousinterferenes.We an nowextrapolate from this observation the general ase where we onsider a radius
of
R CS
inludingh
hops: thenodes(i + (h + 1)
)and(i − (h + 1)
) are thehidden nodes for thenodes
i
(anhidden node to anodei
istherstnode outsidetheR CS
assoiated to thenodei
).2.2.3 Conept of topology lasses
By looking at the onsequenes of the transmission of a frame by a node
i
, we an distinguish two topologieslasses:•
the lass 1 (mono-hop and hain-1), whih represents one broadast domain, i.e. thetransmissionofaframegeneratesasignalwhihisheard byalltheothernodes(beause
allthe nodesareintheintersetionoftheir
R CS
ranges),•
the lass 2 (hain-2 and hain-3), whih represents multiple broadast domains, i.e. thetransmissionof aframe generatesa signalwhih isonly heard bytheothernodeswhih
areintherange
R CS (i)
;all thenodes, whih areoutsideR CS (i)
,donothearanysignal.Aboutthe word heard, we an distinguish two semantis (whih will allowto underline simi-
laritiesbetweentopologies ofthetwolasses): thestrong semantiwhih isthe signalwhihis
reeivedbya node,an be deodedbythis node;theweaksemantiwhihis thesignal,whih
isreeived bya node,induesonlya busyhannel state.
The topologies mono-hop (lass 1) and hain-2 (lass 2) are only haraterized by the strong
semanti(as
R CS = R T
). Thetopologieshain-1(lass1)andhain-3(lass2)areharaterized bythetwo semantis (asR CS > R T
): thestrong semantifor thenodes (i − 1
) and (i + 1
) i.e.thenodes whih arethe neighbours of thenode
i
(they areone hop distant of thenodei
); theweak semantifor thenodesdistant ofthenode
i
from2
hops tillh
hops.2.3 Type 3 of the basi knowledges
ThesebasiknowledgesonernthemainpriniplesoftheCANlikeprotools. Astheseprotools
are inspired bythe CAN network MAC protool, we rst make a reminder of thepriniples of
thisMACprotool. Then,weshowhowweanadaptthesepriniplestothedierenttopologies.