2. 新南威尔士大学 电机工程与通信学院,澳大利亚 悉尼 2052
2. School of Electrical Engineering and Telecommunications, The University of New South Wales, Sydney 2052, Australia
受到自然界中动物集体运动(或者集群、编队运动)行为的启发,人们发现由多个具有自主“计算”能力的个体组成的分布式多自主体系统通常能够完成相对复杂的任务,且具有单个个体所不具有的能力。一致性作为分布式多自主体协同控制的基础[1-4],现在已经成为人工智能控制方向的一个研究热点问题。
近年来,对于分布式多自主体的研究热度不断增长,并涌现出了大量的研究成果。文献[5-6]研究了在固定和变换拓扑下二阶多自主体系统协同控制信息的一致性算法。文献[7]针对有向拓扑结构研究了时变的二阶多自主体网络的一致性问题。在多自主体系统中如果减少自主体之间的通信次数,就可以减少网络中的节点之间数据的传递次数和频率,提高系统的收敛速度。文献[8]研究了具有交换拓扑和存在噪声的多自主体系统模型,提出了一种基于自主体自身以及邻居状态的一致性采样控制协议。文献[9]研究了在采样数据集中的情况下具有领导者的多自主体收敛问题。文献[10]研究了在时变拓扑下通过采样数据实现多自主体系统一致性的问题。尽管采样控制容易实现,且在一定程度上能够减少自主体之间信息交互次数,但在采样数据差值微小或当系统趋近于稳定的时候系统仍然采样还会存在资源浪费的现象。所以Dimarogonas等[11-13]提出了仅需要自主体自己的状态与其邻居自主体状态的事件触发机制,并且给出了可以使多自主体网络实现一致性的集中式、自触发式和分布式的一阶系统的事件触发函数。文献[14]研究并给出了非线性多自主体系统的基于Lyapunov方法和驻留时间方法两种事件触发机制的设计方法。文献[15-16]通过引入合理的状态误差研究了基于事件触发机制的多自主体系统有限时间快速收敛问题,得到了系统实现一致性的事件触发条件。文献[17]针对强连通图提出一种自适应事件触发控制方案,该方案根据采样数据动态改变系统参数可以实现触发间隔的动态调节。文献[18]研究了线性结构下异构多自主体系统的输出收敛问题,设计了一个事件触发下的输出控制机制。文献[19-21]综述了多自主体系统分布式事件触发控制的研究现状,分析了当前几种主要的多自主体系统事件触发一致性算法。在以上的文献中主要研究了一阶系统的分布式事件触发算法,或者二阶系统的集中式触发算法。
本文给出了一种分布式事件触发机制下的二阶多自主体系统的控制协议和触发函数,设计的事件触发函数的触发时刻仅依赖自主体自身状态和其邻居状态的误差值。应用矩阵论和现代控制理论等工具对分布式事件触发多自主体领导跟随一致性进行了分析,得到了二阶系统协同运动的收敛条件。
1 预备知识一个自主体可以利用图中的一个节点来表示,利用边来表示自主体之间的信息传递。带有领导者的网络中的通信拓扑可以由含有
邻接矩阵
引理1 L为连通图 G 的
1)
2) 令
引理2 对于对称矩阵
1)
2)
引理3 对于任意维度的矩阵
1)
2)
3)
4)
引理4 若
1)
2)
针对一个二阶多自主体系统中有
$\left\{ {\begin{array}{*{20}{c}} {{{\dot x}_i}(t) = {v_i}(t)} \\ {{{\dot v}_i}(t) = {u_i}(t)} \end{array}} \right.$ | (1) |
式中:
$\left\{ {\begin{array}{*{20}{c}} {{{\dot x}_0}(t) = {v_0}(t)} \\ {{{\dot v}_0}(t) = {u_0}(t)} \end{array}} \right.$ | (2) |
式中:
定义1 称跟随者自主体
$ \begin{aligned} & \mathop {\lim }\limits_{t \to \infty } \left| {{x_i}(t) - {x_0}(t)} \right| = 0\\ & \mathop {\lim }\limits_{t \to \infty } \left| {{v_i}(t) - {v_0}(t)} \right| = 0 \end{aligned} $ |
式中
假设连续时间下的控制协议为
$\begin{split} {u_i}(t) = & - \alpha \sum\limits_{j \in {N_i}} {({x_i}(t) - {x_j}(t))} - \beta \sum\limits_{j \in {N_i}} {({v_i}(t) - {v_j}(t))} - \\ & \alpha {a_{i0}}({x_i}(t) - {x_0}(t)) - \beta {a_{i0}}({v_i}(t) - {v_0}(t)) \\ \end{split} $ |
设基于事件触发机制的控制协议为
$ \begin{split} {u_i}(t) = & - \alpha \sum\limits_{j \in {N_i}} {({x_i}(t_k^{i}) - {x_j}(t_k^{i}))} - \beta \sum\limits_{j \in {N_i}} {({v_i}(t_k^{i}) - {v_j}(t_k^{i}))} - \\ & \alpha {a_{i0}}({x_i}(t_k^{i}) - {x_0}(t)) - \beta {a_{i0}}({v_i}(t_k^{i}) - {v_0}(t)) \end{split} $ | (3) |
式中:
定义误差函数为
$ \begin{split} {u_i}(t) = & - \alpha \sum\limits_{j \in {N_i}} {({x_i}(t) - {x_j}(t){\rm{ + }}e_x^{i}(t) - e_x^{j}(t))}- \\ & \beta \sum\limits_{j \in {N_i}} {({v_i}(t) - {v_j}(t){\rm{ + }}e_v^{i}(t) - e_v^{j}(t))} -\\ & \alpha {a_{i0}}({x_i}(t){\rm{ + }}e_x^{i}(t) - {x_0}(t))- \beta {a_{i0}}({v_i}(t){\rm{ + }}e_v^{i}(t) - {v_0}(t)) \end{split}$ | (4) |
因此式(1)可以变为
$ \begin{array}{l} \left\{ \begin{array}{l} {{\dot x}_i}(t) = {v_i}(t)\\ {{\dot v}_i}(t) = - \alpha \sum\limits_{j \in {N_i}} {({x_i}(t) - {x_j}(t){\rm{ + }}e_x^i(t) - e_x^j(t))} - \end{array} \right.\\ \quad \quad \quad \;\beta \sum\limits_{j \in {N_i}} {({v_i}(t) - {v_j}(t){\rm{ + }}e_v^i(t) - e_v^j(t))} - \\ \quad \quad \quad \;\alpha {a_{i0}}({x_i}(t){\rm{ + }}e_x^i(t) - {x_0}(t)) - \beta {a_{i0}}({v_i}(t){\rm{ + }}e_v^i(t) - {v_0}(t)) \end{array} $ | (5) |
为了简化书写,令
$ \begin{split} & {{x}}(t)={\left[ {{x_1}(t)\;\;{x_2}(t)\;\; \cdot \cdot \cdot \;\;{x_N}(t)} \right]^{\rm{T}}} \in {{\bf{R}}^N} \\ & {{v}}(t)={\left[ {{v_1}(t)\;\;{v_2}(t)\;\; \cdot \cdot \cdot \;\;{v_N}(t)} \right]^{\rm{T}}} \in {{\bf{R}}^N} \end{split} $ |
$ \begin{split} & {{{e}}_x}(t)={\left[ {e_x^1(t)\;\;e_x^2(t)\;\; \cdot \cdot \cdot \;\;e_x^N(t)} \right]^{\rm{T}}} \in {{\bf{R}}^N}\\ & {{{e}}_v}(t)={\left[ {e_v^1(t)\;\;e_v^2(t)\;\; \cdot \cdot \cdot \;\;e_v^N(t)} \right]^{\rm{T}}} \in {{\bf{R}}^N} \end{split} $ |
则式(5)可以改写为
$\begin{gathered} \left[ {\begin{array}{*{20}{c}} {{{\dot x}}(t)} \\ {{{\dot v}}(t)} \end{array}} \right] = - {{A}} \otimes {{{I}}_{m \times m}}\left[ {\begin{array}{*{20}{c}} {{{x}}(t)} \\ {{{v}}(t)} \end{array}} \right] - {{B}} \otimes {{{I}}_{m \times m}}\left[ {\begin{array}{*{20}{c}} {{{{e}}_x}(t)} \\ {{{{e}}_v}(t)} \end{array}} \right] + \\ \;\;{\left[ {\begin{array}{*{20}{c}} \alpha \\ \beta \end{array}} \right]^{\rm{T}}} \otimes ({{H}} \otimes {{{I}}_{m \times m}})\left[ {\begin{array}{*{20}{c}} {{{1}}{x_0}(t)} \\ {{{1}}{v_0}(t)} \end{array}} \right] \\ \end{gathered} $ | (6) |
式中:
$\left[ {\begin{array}{*{20}{c}} {\mathop {{{\hat x}}}\limits^ \cdot (t)}\\ {\mathop {{{\hat v}}}\limits^ \cdot (t)} \end{array}} \right] = - {{A}} \otimes {{{I}}_{m \times m}}\left[ {\begin{array}{*{20}{c}} {{{\hat x}}(t)}\\ {{{\hat v}}(t)} \end{array}} \right] - {{B}} \otimes {{{I}}_{m \times m}}\left[ {\begin{array}{*{20}{c}} {{{{e}}_x}(t)}\\ {{{{e}}_v}(t)} \end{array}} \right]$ | (7) |
由式(7)可以看出,通过模型变换将研究系统式(1)、(2)的领导跟随一致性问题,变为研究式(7)的稳定性问题。下面针对式(7)设计分布式事件触发函数。
定理1 考虑带有一个领导者的多自主体系统式(1)、(2),如果连通图
$\begin{array}{c} {f_i}({{\hat x}_i}(t),{{\hat v}_i}(t),{e_{xi}}(t),{e_{xj}}(t),{e_{vi}}(t),{e_{vj}}(t))= \\ \sqrt m \cdot \displaystyle\sum\limits_{j \in {N_i}} {{a_{ij}}[\alpha ({{\left\| {{e_{xi}}(t)} \right\|}^2} + {{\left\| {{e_{xj}}(t)} \right\|}^2})} + \\ \beta ({\left\| {{e_{vi}}(t)} \right\|^2} + {\left\| {{e_{vj}}(t)} \right\|^2}) + {a_{i0}}(\alpha {\left\| {{e_{xi}}(t)} \right\|^2} + \beta {\left\| {{e_{vi}}(t)} \right\|^2})] - \\ {\theta _i}(\delta - 1)({\left\| {{{\hat x}_i}(t)} \right\|^2} + {\left\| {{{\hat v}_i}(t)} \right\|^2}) \end{array} $ | (8) |
式中:
证明 令
对
$\begin{array}{c} \dot V(t) = {{\dot \xi }}{(t)^{\rm{T}}}{{P}} \otimes {{{I}}_{m \times m}}{{\xi }}(t) + {{\xi }}{(t)^{\rm{T}}}{{P}} \otimes {{{I}}_{m \times m}}{{\dot \xi }}(t) = \\ ( - {{\xi }}{(t)^{\rm{T}}}({{{A}}^{\rm{T}}}{{P}} + {{PA}}) \otimes {{{I}}_{m \times m}}{{\xi }}(t)) + \\ ( - {{e}}{(t)^{\rm{T}}}{{{B}}^{\rm{T}}}{{P}} \otimes {{{I}}_{m \times m}}{{\xi }}(t) - {{\xi }}{(t)^{\rm{T}}}{{P}} \otimes {{{I}}_{m \times m}}{{Be}}(t)) \end{array} $ |
其中
$\begin{array}{c} \dot V(t)= - {{\hat x}}{(t)^{\rm{T}}}(2\alpha {\bar{ L}}) \otimes {{{I}}_{m \times m}}{{\hat x}}(t) + {{\hat x}}{(t)^{\rm{T}}}{{{I}}_{n \times n}} \otimes {{{I}}_{m \times m}}{{\hat v}}(t) + \\ {{\hat v}}{(t)^{\rm{T}}}{{{I}}_{N \times N}} \otimes {{{I}}_{m \times m}}{{\hat x}}(t) - {{\hat v}}{(t)^{\rm{T}}}(2\beta {\bar{ L}} - 2{{{I}}_{N \times N}}) \otimes {{{I}}_{m \times m}}{{\hat v}}(t) + \\ ( - {{e}}{(t)^{\rm{T}}}{{{B}}^{\rm{T}}}{{P}} \otimes {{{I}}_{m \times m}}{{\xi }}(t) - {{\xi }}{(t)^{\rm{T}}}{{P}} \otimes {{{I}}_{m \times m}}{{Be}}(t)) \end{array} $ |
根据引理3和引理4可以得到:
$\begin{array}{c} {{\hat v}}{(t)^{\rm{T}}}\left[ {2\beta {\bar{ L}} - 2{{{I}}_{N \times N}}} \right] \otimes {{{I}}_{m \times m}}{{\hat v}}(t) =\\ {{\hat v}}{(t)^{\rm{T}}}\left[ {2\beta {\bar{ L}}} \right] \otimes {{{I}}_{m \times m}}{{\hat v}}(t) - 2{{\hat v}}{(t)^{\rm{T}}}{{{I}}_{N \times N}} \otimes {{{I}}_{m \times m}}{{\hat v}}(t) \geqslant \\ ({{\textit{λ}}_{\min }}(2\beta {\bar{ L}}) - 2){\left\| {{{\hat v}}(t)} \right\|^2} \end{array} $ |
所以
$\begin{array}{c} \dot V(t) \leqslant - {{\textit{λ}}_{\min }}(2\alpha {\bar{ L}}){\left\| {{{\hat x}}(t)} \right\|^2} + ({\left\| {{{\hat x}}(t)} \right\|^2} + {\left\| {{{\hat v}}(t)} \right\|^2}) -\\ ({{\textit{λ}}_{\min }}(2\beta {\bar{ L}}) - 2){\left\| {{{\hat v}}(t)} \right\|^2}- [{{\hat x}}{(t)^{\rm{T}}}\alpha ({{L}} + {{H}}) \otimes {{{I}}_{m \times m}}({{{e}}_x}(t)+ \\ {{{e}}_v}(t))+ {{\hat v}}{(t)^{\rm{T}}}\alpha ({{L}} + {{H}}) \otimes {{{I}}_{m \times m}}({{{e}}_x}(t) + {{{e}}_v}(t))] \end{array} $ |
那么
$\begin{array}{c} \dot V(t) \leqslant - {{\textit{λ}} _{\min }}(2\alpha {\bar{ L}}){\left\| {{{\hat x}}(t)} \right\|^2} + ({\left\| {{{\hat x}}(t)} \right\|^2} + {\left\| {{{\hat v}}(t)} \right\|^2}) - \\ ({{\textit{λ}}_{\min }}(2\beta {\bar{ L}}) - 2){\left\| {{{\hat v}}(t)} \right\|^2} - 2g({{\hat x}}(t),{{\hat v}}(t),{{{e}}_x}(t),{{{e}}_v}(t)) \\ \end{array} $ |
其中
$\begin{array}{c} g({{\hat x}}(t),{{\hat v}}(t),{{{e}}_x}(t),{{{e}}_v}(t)) = \\ \displaystyle\sum\limits_{i = 1}^n {{{\hat x}_i}(t)\left[\alpha \left(\displaystyle\sum\limits_{j \in {N_i}} {{a_{ij}} \otimes {{{I}}_{m \times m}}({e_{xi}}(t) - {e_{xj}}(t))} + {a_{i0}} \otimes {{{I}}_{m \times m}}{e_{xi}}(t)\right)\right]} + \\ \displaystyle\sum\limits_{i = 1}^n {{{\hat x}_i}(t)\left[\beta \left(\displaystyle\sum\limits_{j \in {N_i}} {{a_{ij}} \otimes {{{I}}_{m \times m}}({e_{vi}}(t) - {e_{vj}}(t))} + {a_{i0}} \otimes {{{I}}_{m \times m}}{e_{vi}}(t)\right)\right]} + \\ \displaystyle\sum\limits_{i = 1}^n {{{\hat v}_i}(t)\left[\alpha \left(\displaystyle\sum\limits_{j \in {N_i}} {{a_{ij}} \otimes {{{I}}_{m \times m}}({e_{xi}}(t) - {e_{xj}}(t))} + {a_{i0}} \otimes {{{I}}_{m \times m}}{e_{xi}}(t)\right)\right]} + \\ \displaystyle\sum\limits_{i = 1}^n {{{\hat v}_i}(t)\left[\beta \left(\displaystyle\sum\limits_{j \in {N_i}} {{a_{ij}} \otimes {{{I}}_{m \times m}}({e_{vi}}(t) - {e_{vj}}(t))} + {a_{i0}} \otimes {{{I}}_{m \times m}}{e_{vi}}(t)\right)\right]} \end{array} $ |
令
$\delta {\rm{ = max\{ inf}}\{ {{\textit{λ}}_{\min }}(2\alpha {\bar{ L}}) - 1,({{\textit{λ}}_{\min }}(2\beta {\bar{ L}}) - 2) - 1\} ,1{\rm{\} }}$ |
又因为
$\begin{array}{c} \;\; - 2g({{\hat x}}(t),{{\hat v}}(t),{{{e}}_x}(t),{{{e}}_v}(t)) =\\ - {\rm{2}}\displaystyle\sum\limits_{i = 1}^N {({{\hat x}_i}(t) + {{\hat v}_i}(t))\displaystyle\sum\limits_{j \in {N_i}} {{a_{ij}} \otimes {{{I}}_{m \times m}}[\alpha ({e_{xi}}(t) - {e_{xj}}(t))} }+ \\ \beta ({e_{vi}}(t) - {e_{vj}}(t)) + {a_{i0}} \otimes {{{I}}_{m \times m}}(\alpha {e_{xi}}(t) + \beta {e_{vi}}(t))] \leqslant \\ \displaystyle\sum\limits_{i = 1}^N {[({{\left\| {{{\hat x}_i}(t)} \right\|}^2} + {{\left\| {{{\hat v}_i}(t)} \right\|}^2})} + \\ \sqrt m \cdot \displaystyle\sum\limits_{j \in {N_i}} {{a_{ij}}[\alpha ({{\left\| {{e_{xi}}(t)} \right\|}^2} + {{\left\| {{e_{xj}}(t)} \right\|}^2})}+ \\ \beta ({\left\| {{e_{vi}}(t)} \right\|^2} + {\left\| {{e_{vj}}(t)} \right\|^2}) + {a_{i0}}(\alpha {\left\| {{e_{xi}}(t)} \right\|^2} + \beta {\left\| {{e_{vi}}(t)} \right\|^2})]] \end{array} $ |
所以
取
$\begin{array}{c} - {\theta _i}(\delta - 1)({\left\| {{{\hat x}_i}(t)} \right\|^2} + {\left\| {{{\hat v}_i}(t)} \right\|^2}) + \sqrt m \cdot \displaystyle\sum\limits_{j \in {N_i}} {{a_{ij}}[\alpha ({{\left\| {{e_{xi}}(t)} \right\|}^2}} + \\ {\left\| {{e_{xj}}(t)} \right\|^2}) + \beta ({\left\| {{e_{vi}}(t)} \right\|^2} + {\left\| {{e_{vj}}(t)} \right\|^2}) + {a_{i0}}(\alpha {\left\| {{e_{xi}}(t)} \right\|^2} + \\ \beta {\left\| {{e_{vi}}(t)} \right\|^2})] \leqslant 0 \\ \end{array} $ |
所以将触发函数
$\begin{aligned} & \qquad\,\, {f_i}({{\hat x}_i}(t),{{\hat v}_i}(t),{e_{xi}}(t),{e_{xj}}(t),{e_{vi}}(t),{e_{vj}}(t))= \\ & \qquad \quad \sqrt m \cdot \sum\limits_{j \in {N_i}} {{a_{ij}}[\alpha ({{\left\| {{e_{xi}}(t)} \right\|}^2} + {{\left\| {{e_{xj}}(t)} \right\|}^2})} + \\ & \beta ({\left\| {{e_{vi}}(t)} \right\|^2} + {\left\| {{e_{vj}}(t)} \right\|^2}) + {a_{i0}}(\alpha {\left\| {{e_{xi}}(t)} \right\|^2} + \beta {\left\| {{e_{vi}}(t)} \right\|^2})]- \\ & \qquad \qquad \quad {\theta _i}(\delta - 1)({\left\| {{{\hat x}_i}(t)} \right\|^2} + {\left\| {{{\hat v}_i}(t)} \right\|^2}) \end{aligned} $ |
对于第
由
于事件触发条件式(8)是分布式的,任意的两个自主体的事件触发时刻可以是不相同的。如果
如果芝诺现象产生则认为我们设计的事件触发函数是不可行的。为了证明系统在运行过程中不会被触发无数次,下面计算任意两次事件触发的最小间隔。
定理2 考虑带有一个领导者的多自主体系统式 (1)、(2),如果领导者作为图
${\tau _q} = \frac{{\sqrt {{\theta _q}(\delta - 1)} \cdot \sup {{\{ \left\| {{A}} \right\|,\left\| {{B}} \right\|\} }^{ - 1}}}}{{\sqrt {\alpha + \beta } (\left\| {{\varGamma }} \right\| + \left\| {{{HD}}} \right\|) + \sqrt {{\theta _q}(\delta - 1)} }}$ | (9) |
式中:
证明 构造
$\begin{aligned} & \frac{{\rm{d}}}{{{\rm{d}}t}}\left(\frac{{\left\| {{{e}}(t)} \right\|}}{{\left\| {{{\xi }}(t)} \right\|}}\right) = \frac{{\rm{d}}}{{{\rm{d}}t}}\frac{{{{({{e}}{{(t)}^{\rm{T}}}{{e}}(t))}^{\frac{1}{2}}}}}{{{{({{\xi }}{{(t)}^{\rm{T}}}{{\xi }}(t))}^{\frac{1}{2}}}}}= \\ & \frac{{{{e}}{{(t)}^{\rm{T}}}{{\dot e}}(t)}}{{\left\| {{{\xi }}(t)} \right\|\left\| {{{e}}(t)} \right\|}} - \frac{{\left\| {{{e}}(t)} \right\|{{\xi }}{{(t)}^{\rm{T}}}{{\dot \xi }}(t)}}{{{{\left\| {{{\xi }}(t)} \right\|}^3}}} \end{aligned} $ |
当
$\varphi (t) \leqslant \frac{{\sup \{ \left\| {{A}} \right\|,\left\| {{B}} \right\|\} \cdot t - {C_1} + 1}}{{ - \sup \{ \left\| {{A}} \right\|,\left\| {{B}} \right\|\} \cdot t + {C_1}}}$ |
当
$\begin{aligned} & \sum\limits_i^n {\sum\limits_{j \in {N_i}} {{a_{ij}}[\alpha ({{\big|\big| {{e_{xi}}(t)} \big|\big|}^2} + {{\left\| {{e_{xj}}(t)} \right\|}^2}) + \beta ({{\big|\big| {{e_{vi}}(t)} \big|\big|}^2} + {{\left\| {{e_{vj}}(t)} \right\|}^2})} } + \\ & \qquad\qquad\qquad{a_{i0}}(\alpha {\left\| {{e_{xi}}(t)} \right\|^2} + \beta {\left\| {{e_{vi}}(t)} \right\|^2})] \leqslant \\ & \qquad\qquad{\left\| {\left[ {\alpha {{\varGamma }},\beta {{\varGamma }}} \right]e(t) + {{HD}}[\alpha ,\beta ] \otimes {{{I}}_{N \times N}}e(t)} \right\|^2} \leqslant \\ & \qquad\qquad\,\,\,\,\,\,{\left\| {\left[ {\alpha {{\varGamma }},\beta {{\varGamma }}} \right] + {{HD}}[\alpha ,\beta ] \otimes {{{I}}_{N \times N}}} \right\|^2} \cdot {\left\| {e(t)} \right\|^2} \end{aligned} $ |
令
$q = \arg \mathop {\max }\limits_{i \in N} {\rm{(}}{\left\| {{{\hat x}_i}(t)} \right\|^2}{\rm{ + }}{\left\| {{{\hat v}_i}(t)} \right\|^2}{\rm{)}}$ |
则
$\begin{aligned} & N\sum\limits_{j \in Nq} {{a_{ij}}\Big[\alpha ({{\left\| {{e_{xq}}(t)} \right\|}^2} + {{\left\| {{e_{xj}}(t)} \right\|}^2}) + \beta ({{\left\| {{e_{vq}}(t)} \right\|}^2} + {{\left\| {{e_{vj}}(t)} \right\|}^2})} + \\ & \,\,\,\,{a_{i0}}(\alpha {\left\| {{e_{xq}}(t)} \right\|^2} + \beta {\left\| {{e_{vq}}(t)} \right\|^2})\Big] \cdot {(N{\rm{(}}{\left\| {{{\hat x}_q}(t)} \right\|^2}{\rm{ + }}{\left\| {{{\hat v}_q}(t)} \right\|^2}{\rm{)}})^{ - 1}} \leqslant \\ & \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\frac{{{{\left\| {\left[ {\alpha {{\varGamma }},\beta {{\varGamma }}} \right] + {{HD}}[\alpha ,\beta ] \otimes {{{I}}_{N \times N}}} \right\|}^2} \cdot {{\left\| {{{e}}(t)} \right\|}^2}}}{{{{\left\| {{{\xi }}(t)} \right\|}^2}}} \end{aligned} $ |
因此
$\sqrt {\alpha + \beta } (\left\| {{\varGamma }} \right\| + \left\| {{{HD}}} \right\|)\phi ({\tau _q},0)=\sqrt {{\theta _q}(\delta - 1)} $ |
解得触发间隔的最小值为
${\tau _q} = \frac{{\sqrt {{\theta _q}(\delta - 1)} \cdot \sup {{\{ \left\| {{A}} \right\|,\left\| {{B}} \right\|\} }^{ - 1}}}}{{\sqrt {\alpha + \beta } (\left\| {{\varGamma }} \right\| + \left\| {{{HD}}} \right\|) + \sqrt {{\theta _q}(\delta - 1)} }}$ |
证毕。
3 计算机仿真在这一部分,将利用计算机仿真来验证控制算法和事件触发函数。仿真实验中共有6个
${{L}} = \left[ {\begin{array}{*{20}{c}} 1&0&{ - 1}&0&0&0 \\[0.5pt] 0&1&{ - 1}&0&0&0 \\[0.5pt] { - 1}&{ - 1}&3&{ - 1}&0&0 \\[0.5pt] 0&0&{ - 1}&2&{ - 1}&0 \\[0.5pt] 0&0&0&{ - 1}&0&{ - 1} \\[0.5pt] 0&0&0&0&{ - 1}&1 \end{array}} \right]$ |
领导跟随矩阵为
${{H}}{\rm{ = diag\{ 1\,\,\,\, 0\,\,\,\, 0\,\,\,\, 1\,\,\,\, 1\,\,\,\, 0\} }}$ |
Download:
|
|
本文考虑多自主体系统在二维平面内的运动的情况。根据在定理1中给出的限制条件,选取
按照给出的控制器来调控系统得到如图2和图3所示的在两个方向上的位置和速度状态,可见多自主体系统逐渐达到一致。
Download:
|
|
Download:
|
|
表1分别给出了在分布式事件触发机制下,多自主体系统在20 s内触发器的触发次数,由表中数据可以看出采用本文的控制协议可以极大降低自主体的采样次数。
图4描述了6个自主体在平面内的位置信息,从图中可以看到随着时间的推移,最后自主体的运动轨迹实现了一致。图5显示了其中一个自主体每个触发时刻的系统误差,从图中可以看出在初始阶段触发间隔小且误差范数的值相当大,且触发时刻密集,直到整个系统渐渐稳定触发器的触发间隔才会变大,幅值不断减小且触发间隔变得稀疏。
Download:
|
|
Download:
|
|
基于以上的仿真结果,本文给出一致性控制算法和分布式事件触发函数可以确保在固定拓扑下多自主体系统实现领导跟随一致性。
4 结束语本文针对二阶多自主体系统设计了一种分布式事件触发机制的控制算法。通过定义跟随者与领导者之间的偏差向量将二阶微分系统转化为二阶闭环系统的稳定性问题,并运用图论、矩阵论和
[1] | OLFATI-SABER R, MURRAY R M. Consensus problems in networks of agents with switching topology and time-delays[J]. IEEE transactions on automatic control, 2004, 49(9): 1520-1533. DOI:10.1109/TAC.2004.834113 (0) |
[2] | CHEN Shun, HO D W C, HUANG Chi. Fault reconstruction and state estimator design for distributed sensor networks in mul-titarget tracking[J]. IEEE transactions on industrial electronics, 2015, 62(11): 7091-7102. DOI:10.1109/TIE.2015.2448685 (0) |
[3] | YANG Hongyong, WANG Fuyong, HAN Fujun. Containment control of fractional order multi-agent systems with time delays[J]. IEEE/CAA journal of automatica sinica, 2018, 5(3): 727-732. DOI:10.1109/JAS.2016.7510211 (0) |
[4] | YANG Hongyong, ZHANG Zhenxing, ZHANG Siying. Consensus of second-order multi-agent systems with exogenous dis-turbances[J]. International journal of robust and nonlinear control, 2011, 21(9): 945-956. DOI:10.1002/rnc.v21.9 (0) |
[5] | WANG Fang, CHEN Xin, HE Yong, et al. Finite-time consensus problem for second-order multi-agent systems under switching topologies[J]. Asian journal of control, 2017, 19(5): 1756-1766. (0) |
[6] | ZHANG Ke, JIANG Bin, COCQUEMPOT V. Distributed fault estimation observer design for multi-agent systems with switch-ing topologies[J]. IET control theory and applications, 2017, 11(16): 2801-2807. DOI:10.1049/iet-cta.2017.0484 (0) |
[7] | HUA Yongzhao, DONG Xiwang, LI Qingdong, et al. Fault-tolerant time-varying formation control for second-order multi-agent systems with directed topologies[C]//Proceedings of the 13th IEEE International Conference on Control and Automation. Ohrid, Macedonia, 2017: 467–472. (0) |
[8] | CHENG Long, WANG Yunpeng, HOU Zengguang, et al. A sampled-data based average consensus protocol for double-integrator multi-agent systems with switching topologies and communication noises[C]//Proceedings of the 10th World Congress on Intelligent Control and Automation. Beijing, China, 2012: 886–891. (0) |
[9] | LI Fengjiao, SHEN Yanjun, ZHANG Daoyuan, et al. Leader-following consensus for upper-triangular multi-agent systems via sampled and delayed output feedback[J]. Neurocomputing, 2018, 275: 1441-1448. DOI:10.1016/j.neucom.2017.09.077 (0) |
[10] | GAO Yanping, WANG Long. Sampled-data based consensus of continuous-time multi-agent systems with time-varying topolo-gy[J]. IEEE transactions on automatic control, 2011, 56(5): 1226-1231. DOI:10.1109/TAC.2011.2112472 (0) |
[11] | DIMAROGONAS D V, FRAZZOLI E. Distributed event-triggered control strategies for multi-agent systems[C]//Proceedings of the 2009 47th Annual Allerton Conference on Communication, Control, and Computing. Monticello, USA, 2009: 906–910. (0) |
[12] | DIMAROGONAS D V, FRAZZOLI E, JOHANSSON K H. Distributed event-triggered control for multi-agent systems[J]. IEEE transactions on automatic control, 2012, 57(5): 1291-1297. DOI:10.1109/TAC.2011.2174666 (0) |
[13] | SEYBOTH G S, DIMAROGONAS D V, JOHANSSON K H. Control of multi-agent systems via event-based communication[J]. IFAC proceedings volumes, 2011, 44(1): 10086-10091. DOI:10.3182/20110828-6-IT-1002.01496 (0) |
[14] |
高永峰. 非线性控制系统的事件触发机制设计[D]. 大连: 大连理工大学, 2017. GAO Yongfeng. Design of event-triggering mechanisms for nonlinear control systems[D]. Dalian: Dalian University of Tech-nology, 2017. (0) |
[15] | XU Wenying, WANG Zidong, HO D W C. Finite-horizon H∞ consensus for multiagent systems with redundant channels via an observer-type event-triggered scheme [J]. IEEE transactions on cybernetics, 2018, 48(5): 1567-1576. DOI:10.1109/TCYB.2017.2707590 (0) |
[16] | CAO Zhengran, LI Chuandong, WANG Xin, et al. Finite-time consensus of linear multi-agent system via distributed event-triggered strategy[J]. Journal of the Franklin Institute, 2018, 355(3): 1338-1350. DOI:10.1016/j.jfranklin.2017.12.026 (0) |
[17] |
刘丹, 胡爱花, 邵浩宇. 自适应事件触发控制的多智能体系统一致性[J]. 计算机工程与应用, 2017, 53(1): 44-48. LIU Dan, HU Aihua, SHAO Haoyu. Adaptive event-triggered control for consensus of multi-agent systems[J]. Computer engi-neering and applications, 2017, 53(1): 44-48. DOI:10.3778/j.issn.1002-8331.1503-0235 (0) |
[18] | JIA Hongwei, ZHAO Jun. Cooperative output regulation of heterogeneous multiagent systems based on event‐triggered control with fixed and switching topologies[J]. International Journal of Robust and Nonlinear Control, 2018, 28(3): 838-858. DOI:10.1002/rnc.v28.3 (0) |
[19] |
张志强, 王龙. 多智能体系统的事件驱动控制[J]. 控制理论与应用, 2018, 35(8):1051−1065. ZHANG Zhiqiang, WANG Long. Event-triggered control for multi-agent systems [J]. Control theory and applica-tions. 2018, 35(8):1051−1065. (0) |
[20] |
杨飞生, 汪璟, 潘泉. 基于事件触发机制的网络控制研究综述[J]. 控制与决策, 2018, 33(6): 969-977. YANG Feisheng, WANG Jing, PAN Quan. A survey of networked event-triggered control[J]. Control and decision, 2018, 33(6): 969-977. (0) |
[21] | DING Lei, HAN Qinglong, GE Xiaohua, et al. An overview of recent advances in event-triggered consensus of multiagent sys-tems[J]. IEEE transactions on cybernetics, 2018, 48(4): 1110-1123. DOI:10.1109/TCYB.2017.2771560 (0) |