通过配置系统输入u(t),使u(s)G(s)的极点使系统满足一定特性
\(G(s) = \frac{a}{s+a}\) \(\frac{1}{a}\)是时间常数\(\tau\),对应上升为0.63 \(4\tau\)对应阶跃响应0.98
\(m\ddot x+B\dot x+kx=F\) \(\ddot x+2\omega_n\xi \dot x+\omega_n^2x=\frac{F}{m}\)
阻尼比固有频率:\(\omega_n\sqrt{1-\xi^2}\)
单位化:\(u(t)=\frac{F}{\omega_n^2}\) \(H(s) = \frac{\omega_n^2}{s^2+2\xi\omega_ns+\omega_n^2}\)
零极点图: 极点全部在左,系统稳定 虚轴长度代表振荡周期 实轴长度代表衰减速度 \(\cos \theta\)代表阻尼比
特征多项式系数判断传递函数稳定性
\(D1 = a_1\)
\(D2 = \begin{pmatrix} a_1&a_3\\ a_0&a_2 \end{pmatrix}\)
\(D3 = \begin{pmatrix} a_{1}& a_{3}& a_{5}\\ a_{0}& a_{2}& a_{4}\\ 0& a_{1}& a_{3} \end{pmatrix}\)
(开环->闭环稳定性):分析G(s)的N、P,看闭环系统稳定性 开环传递函数中开环增益K从0-无穷时,闭环特征根的移动轨迹 单位负反馈闭环传递函数 \(\phi(s) = \frac{C(s)}{R(s)}=\frac{G(s)}{1+G(s)}\) G(s)是一个
叠加原理不适用 常规分类: 死区 饱和 间隙-滞环
系统收敛:消耗系统能量 系统发散:从外界获取能量
\(X_{ss}(t)\):ss-steady state \(T_s\)Delay time \(T_r\)Rise time \(M_p\)Max Overshoot \(T_{ss}\)Setting time调节时间 BIBO:输入稳定,输出稳定bounded input-bounded output Real:实轴 Im:虚轴 Proportional:比例 Integral:积分 Differential:微分 bounded input-bounded output:稳定性 \(\forall\)for all :任意 \(\exists\) at least one :存在 \(\left \| \cdot \right \|\)norm:范数
特征值\(\lambda\)有\(\lambda v=Av\) \(\ | \lambda I-A\ | = 0\) 特征值 解法:将\(\lambda\)代回\(( \lambda I - A)* v = 0\) \(\lambda_1 、\lambda_2\)对应特征向量\(v_1 、v_2\) 过渡矩阵:特征向量组成的矩阵 \(P = \begin {pmatrix} v_1&v_2 \end {pmatrix}\) \(AP=A[v_1 v_2] = [Av_1 Av_2]=[\lambda_1v_1 \lambda_2 v_2]= \begin{bmatrix} \lambda_1v_{11} & \lambda_2v_{21}\\ \lambda_1v_{12} & \lambda_2v_{22} \end{bmatrix} =P\Lambda \) 所以有,单位向量矩阵P将A特征值对角化矩阵 \(P^-1AP = \Lambda\)
非线性:\(1/x,\sqrt{x},x^n等\)
卷积:\(x(t) = f(t)*h(t)=\int_0^t f(\tau)h(t-\tau)d\tau\) \(f(t)\)=输入 \(h(t)\)=单位冲激响应 \(L_{卷积}\)=L乘积
\(e^{i\theta}=\cos(\theta)+i\sin(\theta)\)
\(\sin(x) = C\rightarrow x = \pi/2+2k\pi + \ln(C\pm\sqrt{C^2-1})i\) \(Z = a + b i \) \(Re(Z) =a \) \(Im(Z)=b \) \(\left | Z \right | = \sqrt{a^2+b^2}\) \(Z = \left | Z \right | \cdot (\cos\theta+i\sin\theta)= \left | Z \right | \cdot e^{i\theta}\) \(Z_1 \cdot Z_2 = \left | Z_1 \right | \left | Z_2 \right | e^{\theta_1+\theta_2}\) \(Z+\bar Z = 2a\) \(Z- \bar Z = 2bi\)
Normal Distribution正态分布、高斯分布 \(X = (\mu,\sigma^2)\) 漏检False Dismissal 误警False Alarm
状态空间:State-Space,包含输入、输出、状态,写成一阶微分方程的形式 \(\dot x = A x + B u\) \(y = Cx+Du\)
\(a \, of\, \lambda_i \leqslant 0\)实部 判断方法:
存在至少一个特征值实部大于零
plot(x,\(\dot x\)),通过x初值,分析点在轨迹上的移动,判断稳不稳定 matlab绘制实例
% 画解微分方程组的相图
clear;cla;clc;
[x,y]=meshgrid(linspace(-5,5));
streamslice(x,y,0 * x + 2 * y,-3 * x + 0 * y );
xlabel('x');ylabel('y');
特征值和相图的关系
\(\dot x = a x\rightarrow x(t) = e^{at}x(0)\) 同理,多元线性方程 \(\dot x = a x\rightarrow x(t) = e^{At}x(0)\) 其中,状态转移矩阵\(\Phi(t)\)解法
性质: \(\Phi(0) = I\) \(x(t) = \Phi(t-t_0)x(t_0)\) \(\Phi ^{-1}(t) = \Phi(-t)\)
\(x(t) = \Phi (t)x(0)+ \int_0^t\Phi(t-\tau)Bu(\tau)d\tau\) 初始状态x(0)响应+输入项u(t)响应
可控性:\(\forall x(0),x(t_f), \exists t_f < +\infty , u[0,t_f], st. x(0)\rightarrow x(t_f)\) 充要条件:
可观性:\(\forall t \in [t_0,t_f],已知y(t),u(t),可求x(t_0)\) \(rank \begin{bmatrix} C\\ CA\\ CA^2\\ ...\\ CA^{n-1} \end{bmatrix} =n \)
\(f(\lambda) = \sum_{i=0}^{n}a_i\lambda ^i\) \(f(A) = 0 \rightarrow A^n = \sum_{i=0}^{n-1}a_iA^i\)
求解\(\left | \lambda I - A\right |\)的特征多项式 将\(\lambda = A \)代入,得到递推公式,解算\(A^n\)
取\(u=v-kx\),其中,v为参考输入,系统闭环矩阵由A变为A-Bk
状态转移矩阵: 这里要改一下,改成估计量 \(x_t^- = F_t x_{t-1} + B_t u_t\)
状态转移矩阵:\(P_t^-=FP_{t-1}F^T+Q\)
协方差矩阵: \( \begin{bmatrix} \sigma_{11}&\sigma_{12}\\ \sigma_{12}&\sigma_{22}\\ \end{bmatrix} \)
卡尔曼方程≠状态观测器
以小车为例,讲卡尔曼滤波最优状态估计
在上图中,P是观测值\(\hat x\)的方差 R是观测器中,来自预估值的比例
概率函数相乘,多传感器信息融合
Barbalat’s 引理 lemma