方建印,謝欣欣,2
( 1.河南工程學(xué)院 理學(xué)院,河南 鄭州 451191; 2.鄭州大學(xué) 數(shù)學(xué)與統(tǒng)計學(xué)院,河南 鄭州 450001)
?
具有飽和輸入和通信約束的多智能體系統(tǒng)的包含控制
方建印1,謝欣欣1,2
( 1.河南工程學(xué)院 理學(xué)院,河南 鄭州 451191; 2.鄭州大學(xué) 數(shù)學(xué)與統(tǒng)計學(xué)院,河南 鄭州 450001)
研究了有向拓撲下具有飽和輸入和通信約束的線性多智能體系統(tǒng)的包含控制問題,利用代數(shù)圖論和Lyapunov控制方法證明了相應(yīng)結(jié)論.假設(shè)系統(tǒng)是有界輸入漸近零可控的且任意追隨者至少存在一個領(lǐng)導(dǎo)者到它有一條有向道路,若持續(xù)通信時長大于某一閾值,利用代數(shù)Riccati方程設(shè)計的低增益間歇反饋控制協(xié)議可以使系統(tǒng)達到半全局包含控制.最后,給出仿真實例證明了理論分析的正確性.
線性多智能體系統(tǒng); 有向拓撲; 輸入飽和; 間歇通信; 半全局包含控制
由于多智能體系統(tǒng)協(xié)同控制在許多熱門領(lǐng)域都有著廣泛的應(yīng)用,如衛(wèi)星群的協(xié)調(diào)和控制、無人駕駛飛行器編隊控制、群集控制等,所以受到了越來越多的關(guān)注.多智能體系統(tǒng)協(xié)同控制中一致性的研究較為廣泛,一般來說,一致性的研究可以分為無領(lǐng)導(dǎo)者的一致性[1-4]和領(lǐng)導(dǎo)者追隨者跟蹤一致性[5-6].包含控制類似于跟蹤控制,是指有多個領(lǐng)導(dǎo)者的系統(tǒng)對追隨者設(shè)計控制協(xié)議,使它們最終趨于多個領(lǐng)導(dǎo)者形成的凸包之中[7-12].由于多智能體系統(tǒng)工作環(huán)境并非實時可控,系統(tǒng)可能會面臨許多問題,比如存在通信約束、輸入飽和等.文獻[12]研究了間歇通信和輸入飽和同時存在的情況下線性多智能體系統(tǒng)的包含控制,考慮追隨者之間的信息交互是無向的,通信間歇是周期性的.本研究考慮智能體間信息交互是有向的(即有向拓撲),間歇通信為非周期情況下具有輸入飽和的線性多智能體系統(tǒng)的包含控制,更具有一般性.
1.1 圖論知識
本研究考慮由N個追隨者和M個領(lǐng)導(dǎo)者組成的多智能體系統(tǒng),用g=(ν,ε,A)表示個體間的通信拓撲.其中,頂點集ν=νF∪νL,νF={1,…,N},νL={N+1,…,N+M},邊集ε?νF×νF∪νF×νL,記有序頂點對(i,j)為圖g中的有向邊,表示第j個個體可以收到第i個個體的信息.i1到ik存在有向道路是指存在有序頂點序列(i1,i2),(i2,i3),…,(ik-1,ik),其中ij∈ν,(ij,ij+1)∈ε,j=1,…,k-1.A=[aij]∈(N+M)×(N+M)表示鄰接矩陣,其中aij≥0.元素aij>0當且僅當(j,i)∈ε,否則aij=0.假設(shè)圖中沒有自環(huán),即aii…=0.圖g的度矩陣為一對角矩陣,記為D=diag{d1,d2,…,dN+M},對角線上的元素定義為節(jié)點的入度,即第i行對角線上元素為圖g的Laplacian矩陣L=D-A=[lij]∈N×N,其中滿足
一般來說,有向圖的Laplacian矩陣是不對稱的.考慮到記為N+1,…,N+M的M個領(lǐng)導(dǎo)者沒有鄰居,可以將L∈(N+M)×(N+M)劃分為如下形式:
式中:L1∈N×N,L2∈N×M.
1.2 問題描述
本研究考慮具有飽和輸入和通信約束的線性多智能體系統(tǒng)的包含控制,通信約束由于通信信道的不可靠、傳感器的限制等因素,智能體只能在一些不連續(xù)的時間段內(nèi)與其鄰居節(jié)點通信.輸入飽和是指由于物理設(shè)備的限制等因素,控制輸入存在界限.
多智能體系統(tǒng)的動力學(xué)描述如下:
(1)
(2)
式中:xi∈n,ui∈m分別為第i個智能體的狀態(tài)變量和控制輸入.A和B是適當維數(shù)的系統(tǒng)矩陣,σ(·)是飽和向量函數(shù),定義如下:
半全局包含控制問題是指對每個追隨者構(gòu)造一個控制協(xié)議,使得對于任意預(yù)先給定的有界集χ∈n,只要任意的追隨者和領(lǐng)導(dǎo)者的初值都在此有界集中,即xi(0)∈χ,i=1,2,…,N+M,追隨者最終趨于由多個領(lǐng)導(dǎo)者形成的凸包之中.
下面引入本研究涉及的一些假設(shè)和引理:
假設(shè)1在有向圖g中,任意一個追隨者至少存在一個領(lǐng)導(dǎo)者到它有一條有向道路.
假設(shè)2[13]線性系統(tǒng)(A,B)是有界控制輸入漸近零可控的,即
(1)(A,B)可鎮(zhèn)定;
(2)A的特征值都有非正實部.
引理3[14]假設(shè)M∈N×N是一個正定矩陣,N∈N×N是一個對稱矩陣,那么對?x∈N有如下矩陣不等式成立:
λmin(M-1N)xTMx≤xTNx≤λmax(M-1N)xTMx.
引理4[15]若假設(shè)2成立,則對?ε∈(0,1]存在唯一的矩陣P()>0,使如下代數(shù)Riccati方程成立:
引理5對任意給定的實矩陣A=[aij]∈n×n,都存在一個正定實矩陣Q=QT∈n×n使得
ATQ+QA-βQ<0,
對具有通信約束和輸入飽和的多智能體系統(tǒng)利用如下算法構(gòu)造狀態(tài)反饋間歇包含控制協(xié)議,保證系統(tǒng)的低反饋增益.
算法1第1步解如下代數(shù)Riccati方程
ATP(ε)+P(ε)A-γ1P(ε)BBTP(ε)+εIn=0,ε∈(0,1],
(3)
第2步對第i個追隨者構(gòu)造如下形式的狀態(tài)反饋控制協(xié)議:
(4)
當t∈[tk,tk+δk)時,智能體可以接收到鄰居的信息,即通信正常;當t∈[tk+δk,tk+1)時,智能體不能接收到鄰居信息,即通信中斷.其中,k∈.該算法建立在引理4的基礎(chǔ)上保證了低反饋增益.
證明在狀態(tài)反饋控制協(xié)議(4)下,系統(tǒng)(1)動力學(xué)改寫為如下形式:
(5)
式中: i=1,2,…,N;k=0,1,2,….
(6)
令c>0為一常數(shù)且滿足:
(7)
(8)
由方程(7)和(8)及ψ的定義將其寫為矩陣形式:
ψ=(L1?In)XF+(L2?In)XL,
(9)
則有
(10)
若通信中斷,則有
(11)
(12)
下面分別在通信正常和中斷的情況下對Lyapunov函數(shù)進行求導(dǎo).
(1)當t∈[tk,tk+δk),?k∈時,對Lyapunov函數(shù)求導(dǎo):
ψT(IN?AT-LT?P(ε)TBBT)(W?P(ε))ψ+ψT(W?P(ε))(IN?A-L1?BBTP(ε)ψ=
(13)
由引理3及式(3)得
ξmaxψT[IN?(ATP(ε)+P(ε)A-γ1P(ε)BBTP(ε))]ψ=-εξmaxψTψ≤
-εξmaxλmin(W-1?P-1(ε))ψT(W?P(ε))ψ=
-ελmin(P-1(ε))V(ψ(t))-αV(ψ(t)),
(14)
式中:α=ελmin(P-1(ε))>0.
(2)當t∈[tk+δk,tk+1),?k∈時,對Lyapunov函數(shù)求導(dǎo):
βψT(IN?Q)ψ=βV(ψ(t)).
(15)
由引理5易得式(15)中的不等式,其中β>0. 綜上可知:
(16)
V(ψ(t1+δ1))=ψT(t1+δ1)(IN?Q)ψ(t1+δ1)≤λmax(Q)ψT(t1+δ1)ψ(t1+δ1)≤
μλmin(P(ε))ξminψT(t1+δ1)ψ(t1+δ1)≤μψT(t1+δ1)(W?P(ε))ψ(t1+δ1).
(17)
又由于
limt→(t1+δ1)-V(ψ(t))=ψT(t1+δ1)(W?P(ε))ψ(t1+δ1)≤V(ψ(t1))e-αδ1,
(18)
故
V(ψ(t1+δ1))≤μV(ψ(t1))e-αδ1.
(19)
V(ψ(t2))=ψT(t2)(W?P(ε))ψ(t2)≤λmaxP(ε)ξmaxψT(t2)ψ(t2)≤
μλminQψT(t2)ψ(t2)≤μψT(t2)(IN?Q)ψ(t2).
(20)
又因
(21)
那么
V(ψ(t2))≤μV(ψ(t1+δ1))eβ(t2-t1-δ1)≤μ2V(ψ(t1))eβ(t2-t1-δ1)-αδ1=V(ψ(t1)e-(α+β)δ1+β(t2-t1)+2ln μ.
(22)
令Δ1=(α+β)δ1-β(t2-t1)-2ln μ,由條件知Δ1>0.
(23)
V(ψ(t))≤K0e-kt,t∈(0,∞].
(24)
從而,‖ψ(t)‖→0即
也就是說,利用算法中的反饋控制協(xié)議,追隨者最終會趨于領(lǐng)導(dǎo)者形成的凸包.
考慮如下的系統(tǒng)矩陣:
易知(A,B)可鎮(zhèn)定且A的特征值具有非正實部.
圖1 多智能體系統(tǒng)的通信拓撲圖Fig.1 Communication topology of multi-agent system
圖2 當ε=0.5時,誤差變量及控制輸入各分量的軌線圖Fig.2 ε=0.5,trajectories of error variable and control input
圖3 當ε=0.8時,誤差變量及控制輸入各分量的軌線圖Fig.3 ε=0.8,trajectories of error variable and control input
由圖2與圖3可知,若定理條件成立,控制輸入的值隨著ε值的減少而減少,誤差變量逐漸趨于0,即所有的追隨者最終趨于領(lǐng)導(dǎo)者形成的凸包之中.
在固定有向拓撲下研究具有間歇通信和飽和輸入同時存在的線性多智能體系統(tǒng)的包含控制問題,在系統(tǒng)是有界輸入漸近零可控的,并且拓撲圖中任意追隨者至少存在一個領(lǐng)導(dǎo)者到它有一條有向道路的假設(shè)下,若持續(xù)通信時長大于某一閾值,利用代數(shù)Riccatic方程設(shè)計的低增益間歇反饋控制協(xié)議可以使系統(tǒng)達到半全局包含控制.
[1] WEN G,DUAN Z,YU W,et al.Consensus in multi-agent systems with communication constraints[J].International Journal of Robust and Nonlinear Control,2012,2(22):170-182.
[2] WEN G,DUAN Z,REN W,et al.Distributed consensus of multi-agent systems with general linear node dynamics and intermittent communications[J].International Journal of Robust and Nonlinear Control,2014,24(16):2438-2457.
[3] WEN G,DUAN Z,CHEN G.Distributed consensus of multi-agent systems with general linear node dynamics through intermittent communications[C]∥24th Chinese Control and Decision Conference,Taiyuan:IEEE,2012:1-5.
[4] HUANG N,DUAN Z,ZHAO Y.Leader-following consensus of second-order non-linear multi-agent systems with directed intermittent communication[J].IET Control Theory and Applications,2014,8(10):782-795.
[5] LI Z,WEN G,DUAN Z,et al.Designing fully distributed consensus protocols for linear multi-agent systems with directed graphs[J].IEEE Transaction on Automatic Control,2015,60(4):1152-1157.
[6] SU H,CHEN M,LAM J,et al.Semi-global leader-following consensus of linear multi-agent systems with input saturation via low gain feedback[J].IEEE Transaction on Circuits and Systems,2013,60(7):1881-1889.
[7] JI M,FERRARI-TRECATE G,EGERSTEDT M,et al.Containment control in mobile networks[J].IEEE Transaction on Automatic Control,2008,53(8):1972-1975.
[8] MENG Z,REN W,YOU Z.Distributed finite-time attitude containment control for multiple rigid bodies[J].Automatica,2010,46(12):2092-2099.
[9] CAO Y,REN W,EGERSTEDT M.Distributed containment control with multiple stationary or dynamic leaders in fixed and switching directed networks[J].Automatica,2012,48(8):1586-1597.
[10]LI Z,REN W,LIU X,et al.Distributed containment control of multi-agent systems with general linear dynamics in the presence of multiple leaders[J].International Journal of Robust and Nonlinear Control,2013,23(5):534-547.
[11]SU H,JIA G,CHEN M.Semi-global containment control of multi-agent systems with input saturation[J].IET Control Theory and Applications,2014,8(18):2229-2237.
[12]SU H,JIA G,CHEN M.Semi-global containment control of multi-agent systems with intermittent input saturation[J].Journal of the Franklin Institute,2015,352(9):3504-3525.
[13]SABERI A,LIN Z,TEEL A.Control of linear systems with saturating actuator[J].IEEE Transaction on Automatic Control,1996,41(3):368-378.
[14]HUANG L.Linear Algebra in System and Control Theory[M].Beijing:Science Press,1984.
[15]LIN Z.Low Gain Feedback[M].London:Springer,1999.
Containment control of multi-agent systems with input saturation and intermittent communications
FANG Jianyin1,XIE Xinxin1,2
(1.CollegeofScience,HenanUniversityofEngineering,Zhengzhou451191,China;2.SchoolofMathematicsandStatistics,ZhengzhouUniversity,Zhengzhou450001,China)
This paper investigates containment control of multi-agent systems with input saturation and aperiodic intermittent communications,the corresponding conclusion is proved based on algebraic graph theory and Lyapunov control method. Under the assumptions that each agent is asymptotically null controllable with bounded controls and there exists at least one leader that has a directed path to each follower,low-gain state feedback aperiodic intermittent control protocol is designed according to the algebraic Riccati equation,simi-global containment control of multi-agent systems is proved theoretically if the communication duration is larger than a certain threshold. Finally,numerical simulation is given to illustrate the theoretical result.
linear multi agent systems; fixed directed topology; input saturation; intermittent communication; simi-global containment control
2016-08-20
方建印(1964-),男,河南南陽人,教授,博士,主要從事運籌與控制研究.
TP13;TP18
A
1674-330X(2016)04-0062-07