A new notion of error bounds: necessary and sufficient conditions (Q1996751)

The paper demonstrates how minimal time functions can be treated in the framework of the existence of error bounds for a finite system of convex inequalities. A new notion of local error bounds called generalized local error bounds with respect to a closed convex subset \(F\) of the Euclidean space \(\mathcal{R}^n\) satisfying \(0\in F\) is introduced and studied. The primary goal of this work is to establish several necessary and sufficient conditions for the existence of generalized local error bounds with respect to \(F\). The main results are divided into the following parts. First, necessary conditions for the existence of generalized local error bounds with respect to \(F\) for a finite system of convex inequalities in terms of normal cones and end sets are established. If, in addition, \(F\) is a bounded closed convex set containing the origin as an interior point, then it is shown that under additional stronger assumptions on \(F\), these necessary conditions become sufficient conditions. Further, a generalized invariant-point theorem where the classical distance is replaced by the minimal time function is provided. The obtained results are then used to prove another sufficient condition for the existence of a generalized local error bound with respect to \(F\). The paper is well organized and structured but in some places not easy to understand due to a complicated notation. The main theorems are accompanied with simple examples illustrating the theory.