Considerthesecond-ordersystems:
˙i=ζiξ
˙i=uiζ
whicharealsothenon-zeroeigenvaluesofmatrixΘ.From
¯lie[7]weknowthatwithproperγ,alltheeigenvaluesofΘ
inthelefthalfplaneandthesystemisstablewhichmeansthatthesystem(3)canreachsecond-orderconsensus.Fur-thermore,theconvergencerateofsystem(3)isdeterminedbytheeigenvalueofΘwhichhasthelargestnegativerealpart.
Nowweassumethatalltheeigenvaluesofmatrix Larereal.Notethatanundirectednetworkisaspecialcase.Thefollowingresultfollows.
Lemma3.1:IfalltheeigenvaluesofmatrixLarereal,andtheassociatedtopologyhasaspanningtree,thentheprotocol(2)withanypositivevalueγsolvesthesecond-orderconsensusproblem.
Proof:SincealltheeigenvaluesofLarereal,wehaveµn≤...≤µ1=0,whereµiareeigenvaluesof L.For√22
γµ±γµ+4µi
eachµiotherthanzero,λi±=i.Ifγ2µ2+4µi<0,thenRe(λi±)=γµi/2<0.If0<γ2µ2+
224µi<√γµ,thenγµ+4µi< γµiandλi <λi+=
γµi+
γ2µ2+4µi(1)
whereξiandζiarethepositionandvelocityofagenti,i=1,2,···,n.Inthefollowing,werefertoξandζaspositionandvelocityforsimplicity.
Inthecaseoffullstateinformationexchangebetweenagents,RenandAtkins[7]proposesthefollowingprotocol:
ui=
n j=1
aij(ξi ξj)
n j=1
γaij(ζi ζj)
(2)
Bysubstitutingtheprotocol(2)into(1),theclosed-loop
multi-agentsystemcanbegivenby
˙ξ0n×nInξ
(3)˙= L γLζζwhereξ=[ξ1,ξ2,...,ξn]T,ζ=[ζ1,ζ2,...,ζn]T.
Inthispaperweassumethatagentsmoveinaone-dimensionalspace,forsimplicity.Itshouldbenoted,how-ever,thatitiseasytoextendthestudytoageneralmulti-dimensionalcasebyusingKroneckerproduct.
III.CONVERGENCERATEFORFULLSTATE
INFORMATIONEXCHANGE
Considerthedifferenceofsystemstates,anddenote i=ξi ξ1andηi=ζi ζ1,i=2,3,...,n.Furthermore,let¯=[ 2,..., n]Tandζ¯=[η2,...,ηn]T.Then,itfollowsfromξ
(3)that
˙¯¯ξ0n 1×n 1In 1ξ
(4)¯¯¯˙=¯ L γLζζwhere
l22 l12
.¯= .L.ln2 l12
···...···
l2n l1n
.. .
lnn l1n
¯haseigenvaluesµi,i=1,2,...,n Supposethatmatrix L
¯1,theneachµicorrespondstotwoeigenvaluesλi±ofΘ
satisfyingλ2i=(λiγ+1)µi.Indeed,bysolvingtheequationwehaveγµii
,λi±=
2
withlijthe(i,j)-entryoftheLaplacianmatrixL.
Ifthesystem(4)isasymptoticallystable,thedifferencebetweenstatesofanytwoagentswillconvergetozero,i.e.theconsensusisreachedbythemulti-agentsystem.
¯hasexactlythesameItiseasytoseethatthematrixL
n 1non-zeroeigenvaluesofL.Denotethestatematrixin
¯.Thecharacteristic(3)asΘanditscounterpartin(4)asΘ
¯isgivenby[7]polynomialofΘ
λI I¯ detλI Θ= ¯λI+γL¯ L
¯)=det(λ2I+(λγ+1)L
<0.Sowecanseethatforanyγ>0,
thematrixΘhasexactlytwozeroeigenvaluesandallothereigenvalueslieinthelefthalfplane.AccordingtoLemma4.1of[7],thesystemwillreachsecond-orderconsensus.Withthelemma,wehavethefollowingtheoremforthefastestconvergencerateforthesecond-orderconsensus.Theorem3.1:IfalltheeigenvaluesofLaplacianmatrix Larerealandµn≤...≤µ2<µ1=0,withprotocol(2)theoptimalconsensusconvergencerateofsystem(3)is
2n
2nandthecorrespondingoptimalgainofγisgiven
4µn
byγ=.
22n
Proof:Todeterminetheoptimalgainγforfastestconvergence,weneedtoanalyzetheset ofeigenvaluesofmatrixΘ.Itisobviousthatλi =(γµi ii)/2hasreal partlessthanorequaltoitscounterpartλi+=(γµi+ii/2,soweonlyneedtoinvestigateλi+.
Withdifferentγandµi,thepartunderthesquarerootmightbepositiveornegative,soitisrequiredtodividethevalueofγintothreeparts.
Case1:0<γ≤ 4
nInthiscase,(γµi)2+4µi<0,i=2,3,...,nforallµiandtherealpartsofthenonzeroeigenvalueswillbeγµi/2.Soγµ2/2isthelargestrealpartandthegreatertheγ,thesmallertheγµ2/2(sinceµ2< 0)andhencethefasterthesystemconverges.Whenγ1=ntheconvergence
√
rateisthefastest,whichisr1=|γ1µ2/2|= µ2/n
44
Case2: n≤...≤ l≤γ≤ l4≤...≤ 1
4 2Inthiscase,(γµi)2+4µi<0,i=2,3,...,l 1anditisthesameascase1andtheconvergenceratewillbeγµ2/2.Forotheri,(γµi)2+4µi>0,i=l,...,nandλi=
Consensus of Second-order Multi-agent Systems with Time-Delay_工学_高等教育...Lin and Y. Jia, Distributed Consensus Control for Second-Order Agents with...
Second-order Consensus of Multi-agent System with...0, E t 0 determines the decay rate of T T ...Atkins, “Distributed multi-vehicle coordinated ...
Convergence Rate of Di... 暂无评价 6页 2下载券 Second-order Consensus... 暂无评价 4页 2下载券 2008 Finite-time conse... 暂无评价 6页 1下载券 sec...
Agent Networks with Second-Order Agent Dynamics ?...distributed consensus protocol ui involving ...nite-time convergence analysis. A function V (x...
Consensus tracking pro... 8人阅读 6页 2.00 Extended consensus ana... 5人阅读 6页 2.00 Consensus of Multi-Age... 2人阅读 6页 2.00...
under directed communication topology via distributed adaptive nonlinear protocol?...[3] investigated the consensus problem of second-order multi-agent systems ...
MARCH 2010 Consensus of a Class of Second-Order Multi-Agent Systems With ...Moreau, “Stability of continuous-time distributed consensus algorithm,” in ...
2011Consensus of second-order multi-agent systems under Time-varying Communication...I. I NTRODUCTION Distributed coordination of multi-agent systems has been ...
proposed several dynamic consensus algorithms for second-order multi-agent systems..., LI L. : ‘Distributed robust H 1 consensus control in directed network...
NOTE ON MULTIPLE POSITIVE SOLUTIONS TO NON-HOMOGENOUS MULTI-POINT BVPS FOR SECOND ORDER p-LAPLAC_专业资料。In this paper, some new sufficient conditions ...
热门文档