Global optimization on an interval (Q2359788)

The problem is that of finding all global maxima of a continuously differentiable real valued objective function \(F(t)\) in a closed interval \([0, T]\) (this includes possible maxima on the boundary). For this, two auxiliary functions or \textit{adjoint variables} \(x(t),\) \(y(t)\) are introduced, solution of the initial value problems \[ x'(s) = \Phi(f(T - s), x(s)) \, , \quad x(0) = 0 \qquad y'(t) = \Phi(- f(t), y(t))\, , \quad y(0) = 0 \] respectively, where \(f(t)\) is the derivative of \(F(t)\) and \(\Phi(t, x) = 0\) for \(x< 0,\) \(\Phi(t, 0) = \max(0, t),\) \(\Phi(t, x) = t\) for \(x > 0.\) In terms of these functions, the smallest maximum \(t^*\) and the largest maximum \(t^{**}\) are given by \[ t^* = T - \sup\{s \in [0, T]; x(s) = 0\} \, , \qquad t^{**} = \sup \{t \in [0, T]; y(t) = 0\} \] respectively. There also results for the maxima in between (if any), in particular a recursive procedure to find all of them if finitely many, and a necessary and sufficient condition for maxima, namely \(\lambda(t) = 0\) where \(\lambda(t) = \max\{x(T - t), y(t)\}\). The results are applied to monopoly pricing and optimal stopping.