Saved in:
Bibliographic Details
Main Authors: Yuceel, Ege, Tchalakov, Teodor, Mitra, Sayan
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.03132
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908934997016576
author Yuceel, Ege
Tchalakov, Teodor
Mitra, Sayan
author_facet Yuceel, Ege
Tchalakov, Teodor
Mitra, Sayan
contents Safety-critical autonomous systems must satisfy hard state constraints under tight computational and sensing budgets, yet learning-based controllers are often far more complex than safe operation requires. To formalize this gap, we study how many distinct control signals are needed to render a compact set forward invariant under sampled-data control, connecting the question to the information-theoretic notion of invariance entropy. We propose a vector-quantized autoencoder that jointly learns a state-space partition and a finite control codebook, and develop an iterative forward certification algorithm that uses Lipschitz-based reachable-set enclosures and sum-of-squares programming. On a 12-dimensional nonlinear quadrotor model, the learned controller achieves a $157\times$ reduction in codebook size over a uniform grid baseline while preserving invariance, and we empirically characterize the minimum sensing resolution compatible with safe operation.
format Preprint
id arxiv_https___arxiv_org_abs_2604_03132
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Minimal Information Control Invariance via Vector Quantization
Yuceel, Ege
Tchalakov, Teodor
Mitra, Sayan
Systems and Control
Robotics
Safety-critical autonomous systems must satisfy hard state constraints under tight computational and sensing budgets, yet learning-based controllers are often far more complex than safe operation requires. To formalize this gap, we study how many distinct control signals are needed to render a compact set forward invariant under sampled-data control, connecting the question to the information-theoretic notion of invariance entropy. We propose a vector-quantized autoencoder that jointly learns a state-space partition and a finite control codebook, and develop an iterative forward certification algorithm that uses Lipschitz-based reachable-set enclosures and sum-of-squares programming. On a 12-dimensional nonlinear quadrotor model, the learned controller achieves a $157\times$ reduction in codebook size over a uniform grid baseline while preserving invariance, and we empirically characterize the minimum sensing resolution compatible with safe operation.
title Minimal Information Control Invariance via Vector Quantization
topic Systems and Control
Robotics
url https://arxiv.org/abs/2604.03132