FIND ME ON

GitHub

LinkedIn

Free endpoint problem

🌱

Control

We wish to find Admissible Controls that defined on [t0,t1][t_{0},t_{1}], starting at x(t0)=x0x(t_{0})=x_{0} s.t. we minimize the cost J(u)=t0t1(xu(t)L(t)xu(t)+u(t)u(t))dt+xu(t1)Qxu(t1)J(u)=\int\limits _{t_{0}}^{t_{1}}(x^{\top}_{u}(t)L(t)x_{u}(t)+u^{\top}(t)u(t)) \, dt + x_{u}^{\top}(t_{1})Qx_{u(t_{1})} where the index uu in xux_{u} is used to emphasize that the trajectory depends on the choice of control, L:[t0,t1]Mn(R)L:[t_{0},t_{1}]\to M_{n}(\mathbb{R}) is a Continuous function, and QMn(R)Q\in M_{n}(\mathbb{R}). Note that in this problem we are not concerned with where the trajectory is going to at time t1t_{1}, but only w.r.t. minimizing the cost.

Consider the LTVC system x˙=A(t)x(t)+B(t)u(t)\begin{align*} \dot{x}&= A(t)x(t)+B(t)u(t) \end{align*}and suppose that K()K(\cdot) is a differentiable matrix function on [t0,t1][t_{0},t_{1}] s.t. for K(t)=K(t),t[t0,t1]K(t)=K^{\top}(t),\,\forall t\in[t_{0},t_{1}]. Then x(t1)K(t1)x(t1)x(t0)K(t0)x(t0)=t0t1[u(t)x(t)][0B(t)K(t)K(t)B(t)K˙(t)+A(t)K(t)+K(t)A(t)]dt\begin{align*} &x^{\top}(t_{1})K(t_{1})x(t_{1})-x^{\top}(t_{0})K(t_{0})x(t_{0})\\ &=\int\limits _{t_{0}}^{t_{1}}\begin{bmatrix}u^{\top}(t) & x^{\top}(t)\end{bmatrix}\begin{bmatrix}0 & B^{\top}(t)K(t)\\ K(t)B(t) & \dot{K}(t)+A^{\top}(t)K(t)+K(t)A(t)\end{bmatrix} \, dt \end{align*}

Consider a LTVC system x˙(t)=A(t)x(t)+B(t)u(t)\dot{x}(t)=A(t)x(t)+B(t)u(t)where A,BA,B are Continuous functions of time and x(t0)=x0x(t_{0})=x_{0}. If K:[t0,t1]Mn(R)\exists K:[t_{0},t_{1}]\to M_{n}(\mathbb{R}) with K(t)=K(t),t[t0,t1]K(t)=K^{\top}(t),\,\forall t\in[t_{0},t_{1}] which satisfies the Riccati Differential Equation: K˙(t)=A(t)K(t)K(t)A(t)L(t)+K(t)B(t)B(t)K(t)K(t1)=Q\begin{align*} \dot{K}(t)&= -A^{\top}(t)K(t)-K(t)A(t)-L(t)+K(t)B(t)B^{\top}(t)K(t)\\ K(t_{1})&= Q \end{align*}the solution to which we denote by tΠ(t,Q,t1)t\mapsto \Pi(t,Q,t_{1}), then there exists an optimal control u:[t0,t1]Rmu^{*}:[t_{0},t_{1}]\to \mathbb{R}^{m} for the Free endpoint problem with the cost J(u)=t0t1(xu(t)L(t)xu(t)+u(t)u(t))dt+xu(t1)Qxu(t1)J(u)=\int\limits _{t_{0}}^{t_{1}}\left(x^{\top}_{u}(t)L(t)x_{u}(t)+u^{\top}(t)u(t)\right) \, dt + x_{u}^{\top}(t_{1})Qx_{u}(t_{1}) In particular, the optimal control is u(t)=B(t)Π(t,Q,t1)x(t),t[t0,t1]u^{*}(t)=-B^{\top}(t)\Pi(t,Q,t_{1})x(t),\quad t\in[t_{0},t_{1}] a continuous function, and the optimal cost is Jmin=x0Π(t0,Q,t1)x0J_{min}=x^{\top}_{0}\Pi(t_{0},Q,t_{1})x_{0}

Linked from