Set-based state estimation of nonlinear discrete-time systems using constrained zonotopes and polyhedral relaxations
Rego, Raffo, Terra, Scott · eess.SY,cs.SY,math.OC · 2026-09-08 · 原文
This paper presents a new algorithm for set-based state estimation of nonlinear discrete-time systems. A key step in such algorithms is to propagate a set (often a zonotope or constrained zonotope) through a nonlinear function. Existing methods accomplish this through conservative linearization procedures that are known to lead to severe overestimation in many cases. Here, we propose an alternative that avoids linearization using the so-called factorable representation of nonlinear functions. We use a recursive polyhedral relaxation technique based on this representation that is well-established in the global optimization literature but has not previously been used for set-based estimation. This technique is combined with constrained zonotope (CZ) technology to avoid the limitations of recursive computations with polyhedra in halfspace representation. The resulting state estimation method is fully automated, has attractive computational complexity (with one caveat discussed herein), and can provide significantly tighter enclosures than those resulting from linearization procedures in many cases. Numerical examples highlight the advantages of this approach relative to existing CZ me
讲义
讲义·推断 依据「原文」自动生成的结构化摘要（推断），非原文表述；以原文为准。
1. 人话版
This paper presents a new algorithm for set-based state estimation of nonlinear discrete-time systems.
A key step in such algorithms is to propagate a set (often a zonotope or constrained zonotope) through a nonlinear function.
2. 领域脉络
本文类目：eess.SY、cs.SY、math.OC，属于其所在研究脉络的最新进展。
3. 机制拆解
Existing methods accomplish this through conservative linearization procedures that are known to lead to severe overestimation in many cases.
Here, we propose an alternative that avoids linearization using the so-called factorable representation of nonlinear functions.
4. 证据与数字
摘要未给出量化结果——留意原文的实验与数据。
5. 反例与边界
We use a recursive polyhedral relaxation technique based on this representation that is well-established in the global optimization literature but has not previously been used for set-based estimation.
This technique is combined with constrained zonotope (CZ) technology to avoid the limitations of recursive computations with polyhedra in halfspace representation.
6. 跨领域连接与意外收获
横跨 3 个类目（eess.SY、cs.SY、math.OC），关注其在你兴趣板块间的迁移面。
7. 可复用方法
把本文机制与你手头项目对照，找一个两周内能验证的最小实验。
8. 术语表
精读时把不熟的术语记入此处，作为下次回忆的锚点。