Generic uniqueness and conjugate points for optimal control problems

From MaRDI portal





The problem of minimizing\N\[\N\int_0^T L(x(t),u(t))\,\mathrm{d}t + \psi (x(T))\tag{1} \N\]\Nsubject to\N\[\N \dot{x}=g_0(x)+\sum_{i=1}^m g_i(x) u_i, \quad u\in\mathbb{R}^m, \quad x(0)=y\tag{2}\N\]\Nis considered, for \(\mathcal{C}^3\) vector fields \(g_i\) and Lagrangian \(L\) that satisfy usual bounds. \(L\) is also supposed to be uniformly convex w.r.t. \(u\). Necessary optimality conditions for weak local minimizers (in feedback form) are written in full backwards form by using Pontryagin Maximum Principle (with terminal conditions \(x(T)=z\) and \(p(T) = \nabla \psi (z)\), where \(p\) is the adjoint vector). A point \(y\) is called a \textit{conjugate point} if it is the initial term of a minimum for (1), (2) defined through a backwards solution \(x(t,z)\) for which the Jacobian \(D_z(x(0,z)\) does not have full rank. The authors prove that there exists a dense \(\mathcal{G}_\delta\) set \(\mathcal{M}\subseteq \mathcal{C}^4(\mathbb{R}^n)\) such that for every \(\psi \in \mathcal{M}\) the set of conjugate points relative to \(\psi\) is an embedded manifold of dimension \(n- 2\). Moreover, the set of non-conjugate initial points which admit two or more globally optimal trajectories is contained in the union of a locally finite family of embedded manifolds. In particular, the value function is continuously differentiable on an open and dense subset of \(\mathbb{R}^n\). A relevant technical lemma establishes an a priori bound on globally optimal triples \(x,u,p\). The countable family of open dense sets that define \(\mathcal{M}\) is identified through suitable transversality conditions involving \(x\) and \(p\).











This page was built for publication: Generic uniqueness and conjugate points for optimal control problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7295591)