讨论被控对象和二次性能指标,特别是物理意义。这有助于获得在二次型成本泛函中选择各种矩阵的一些优美的数学条件。因此,将从工程角度处理优化问题。
考虑线性时变系统(LTV)
\[\dot{\mathbf{x}}(t) = \mathbf{A}(t) \mathbf{x}(t) + \mathbf{B}(t) \mathbf{u}(t) \tag{3.1.1}\] \[\mathbf{y}(t) = \mathbf{C}(t) \mathbf{x}(t) \tag{3.1.2}\]其成本泛函(CF)或性能指标(PI)为
\[\begin{aligned} J( \mathbf{u}(t) ) = & J( \mathbf{x}(t_0), \mathbf{u}(t), t_0 ) \\ = & \frac{1}{2} [\mathbf{z}(t_f)-\mathbf{y}(t_f)]' \mathbf{F}(t_f) [\mathbf{z}(t_f)-\mathbf{y}(t_f)] \\ & + \frac{1}{2} \int_{t_0}^{t_f} \bigg [ [\mathbf{z}(t)-\mathbf{y}(t)]' \mathbf{Q}(t) [\mathbf{z}(t)-\mathbf{y}(t)] + \mathbf{u}'(t) \mathbf{R}(t) \mathbf{u}(t) \bigg ] dt \end{aligned} \tag{3.1.3}\]其中,\(\mathbf{x}(t)\)为n维状态向量,\(\mathbf{y}(t)\)为m维输出向量,\(\mathbf{z}(t)\)为m维参考或期望输出向量(或为n维期望状态向量,若状态\(\mathbf{x}(t)\)可知),\(\mathbf{u}(t)\)为m维控制向量,\(\mathbf{e}(t) = \mathbf{z}(t)-\mathbf{y}(t)\)(或\(\mathbf{e}(t) = \mathbf{z}(t)-\mathbf{x}(t)\),若状态\(\mathbf{x}(t)\)直接可知)为m维误差向量。\(\mathbf{A}(t)\)为\(n \times n\)的状态矩阵,\(\mathbf{B}(t)\)为\(n \times r\)的控制矩阵,\(\mathbf{C}(t)\)为\(m \times n\)的输出矩阵,
\(\) \(\)