Fixed-point methods in optimization problems for control systems (Q6125739): Difference between revisions

The author considers the optimal control problem: \(\Phi (\sigma )=\varphi (x(t_{1}),\omega )+\int_{T}F(x(t),u(t),\omega ,t)dt\rightarrow \inf_{\sigma \in \Omega }\), the state vector \(x(t)=(x_{1}(t),\ldots ,x_{n}(t))\) being the solution to the differential system: \(\overset{.}{x}(t)=f(x(t),u(t),\omega ,t)\), with the initial condition \(x(t_{0})=a=(a_{1},\ldots ,a_{n})\in A\sqsubseteq \mathbb{R}^{n}\), \(u(t)\) being the vector of control functions \( u(t)=(u_{1}(t),\ldots ,u_{m}(t))\in U\sqsubseteq \mathbb{R}^{m}\), and \( \omega (t)\) the vector of control parameters \(\omega (t)=(\omega _{1}(t),\ldots ,\omega _{l}(t))\in W\sqsubseteq \mathbb{R}^{l}\), \(t\in T=[t_{0},t_{1}]\). \(\sigma =(u,\omega ,a)\) denotes an admissible control with values in the set \(\Omega =V\times W\times A\). The sets \(U\), \(W\), and \(A\) are supposed to be closed and convex, and the set \(V\) of admissible control functions consists of \(U\)-valued functions that are piecewise continuous on \( T\). The function \(\varphi (x,\omega )\) is supposed to be continuously differentiable on \(\mathbb{R}^{n}\times W\) and the functions \(F(x,u,\omega ,t)\) and \(f(x,u,\omega ,t)\) and their partial derivatives with respect to \(x\) , \(u\), and \(\omega \) are supposed to be continuous on the set \(\mathbb{R} ^{n}\times U\times W\times T\), the function \(f(x,u,\omega ,t)\) being Lipschitz with respect to \(x\) in \(\mathbb{R}^{n}\times U\times W\times T\). These conditions guarantee the existence and uniqueness of a solution \(x(t,\sigma )\), \(t\in T\), to the above system for any admissible control \(\sigma \in \Omega \). The author introduces the Pontryagin function which involves an adjoint variable \(\psi \in \mathbb{R} ^{n}\) and the standard adjoint system and he draws computations which lead to a nonlinear system which allows improving a given control \(\sigma ^{I}\in \Omega \). This system involves projection operators.\ The purpose of the paper is to propose two algorithms which solve this nonlinear system. The author refers to the contraction mapping principle for the convergence of these iteration processes to solutions of the fixed-point problems for sufficiently small values of the projection parameter \(\alpha >0\). The paper ends with the analysis of two examples of time-optimal problems: Zermelo's problem and orientation of an aircraft.