Higher order necessary conditions for optimal controls not ranging in the interior
Francesca Angrisani, Franco Rampazzo
TL;DR
The paper extends classical higher-order necessary conditions for optimal controls to affine systems when the minimizing control may lie on the boundary of the control set. By introducing the concept of $i$-balanced controls and leveraging a Lie-bracket–driven variation framework, it derives a higher-order maximum principle that includes $(i,j)$-Goh conditions, $(i)$-Legendre–Clebsch conditions, and, in the single-input case, a third-order condition involving $[g,[f,g]]$. The results are established via a set-separation argument using Boltyanski approximating cones and a detailed construction of finite and infinite families of variations, culminating in multiplier conditions that subsume the classical interior control case. The work broadens the applicability of singular-control theory to problems with target sets that may exclude interior controls, enabling more robust optimality analysis in noninnerable control landscapes.
Abstract
Goh's and Legendre-Clebsch necessary conditions for optimal controls of affine-control systems are usually established under the hypothesis that the minimizing control lies in the interior of the control set $U$. In this paper we investigate the possibility of establishing Goh's and Legendre-Clebsch necessary conditions without this assumption, so that even control sets with empty interiors or optimal controls touching the boundary of $U$ can be taken into consideration.
