On the second-order optimality conditions for multi-objective optimal control problems with mixed pointwise constraints (Q2679244)

The dynamics of the system is given by \[ \begin{gathered} x'(t) = \varphi(t, x(t), u(t)) \quad (0 \leq t \leq 1) \\ h_0(x(0) = 0, \ h_1(x(1)) = 1,\\ g_1(t, x(t), u(t)) \leq 0, \dots, g_r(t, x(t), u(t)) \leq 0 \end{gathered} \] where the state \(x(t)\) is a \(n\)-dimensional vector function and the control \(u(t)\) is \(l\)-dimensional. The functions \(h_0(x)\), \(h_1(x)\) take vector values, not necessarily of the same dimensions.The multi-objective value function is \[ J(x, u) = (J_1(x, u), J_2(x, u), \dots, J_m(x, u)) \] where \[ J_j(x, u) = \ell_j(x(0), x(1)) + \int_0^1 L_j(t, x(t), u(t)) dt \quad (j = 1, 2, \dots, m). \] The control problem is to minimize \(J(x, u)\) in \(\mathbb{R}^m_+\), the nonnegative orthant of \(\mathbb{R}^m\) defined by \(\xi_1 \geq 0, \xi_2 \geq 0, \dots, \xi_m \geq 0\); order in \(\mathbb{R}^m\) is \(a \leq b \Leftrightarrow b - a \in \mathbb{R}^m_+\). Using the method of critical directions introduced by \textit{A. Ya. Dubovitskiĭ} and \textit{A. A. Milyutin} [U.S.S.R. Comput. Math. Math. Phys. 5, No. 3, 1--80 (1967; Zbl 0158.33504); translation from Zh. Vychisl. Mat. Mat. Fiz. 5, 395--453 (1965)], the authors derive necessary first and second order conditions for a minimum involving two functions \(p(t)\) and \(\theta(t)\), the first satisfying an adjoint equation in \(0 \leq t \leq 1\) with boundary conditions at both ends. The authors show that the second order condition, suitable strengthened, becomes sufficient for a minimum.