Id: arxiv/2608.07908 Url: https://arxiv.org/abs/2608.07908 Tags: math.OC, eess.SY Score: 0 Title: Computing the Maximal Controlled Invariant Set for Neural Network Control Systems Domain: math.OC Source: arxiv Authors: Tianxiao Ye, Hang Zhang, Xiangru Xu Reviews: 0 Subject: math.OC Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-08-08 Page: https://delphi.science/r/arxiv/2608.07908 This article studies the computation of the maximal controlled invariant set (MCIS) for neural network control systems (NNCSs) with a nominal plant model and a neural network residual model. An interval-based method is developed to construct a control-affine inclusion function for the NNCS dynamics and to exploit the induced hyperplane arrangement in the input space, which reduces the verification of controlled invariance from an infinite control search to a finite set of representative evaluations. Based on this finite characterization, verification-guided algorithms iteratively construct…