TERA: A Unified Taylor Model Enabled Reachability Analysis Framework

Andrew Sogokon; Salma Iraky

arxiv: 2607.01189 · v1 · pith:326IMO2Unew · submitted 2026-07-01 · 📡 eess.SY · cs.SY

TERA: A Unified Taylor Model Enabled Reachability Analysis Framework

Salma Iraky , Andrew Sogokon This is my paper

Pith reviewed 2026-07-02 07:12 UTC · model grok-4.3

classification 📡 eess.SY cs.SY

keywords reachability analysisTaylor modelshybrid systemsstochastic systemsnonlinear ODEssafety verificationPython framework

0 comments

The pith

TERA implements Taylor model reachability analysis for continuous, hybrid and stochastic systems inside one Python codebase.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents TERA as a single open-source Python framework that applies Taylor model arithmetic to compute rigorous over-approximations of reachable sets. It targets non-linear ODEs, hybrid systems, and continuous-time stochastic dynamics within one symbolic-numeric workflow, aiming to reduce the wrapping effect that produces overly loose enclosures in other methods. A sympathetic reader would care because safety-critical verification needs tight, trustworthy bounds on all possible trajectories, and an extensible Python tool could lower the barrier to testing new analysis techniques across system classes. The current implementation already produces tight results on standard benchmarks for ODEs and hybrid cases while supporting stochastic analysis.

Core claim

TERA is a unified Python-native framework for Taylor-model-based reachability analysis that supports continuous, hybrid and stochastic systems in a single workflow, delivering tight reachable-set over-approximations for non-linear ODEs and hybrid systems on difficult benchmarks and already handling continuous-time stochastic systems.

What carries the argument

Taylor Model arithmetic and propagation rules, implemented once in a Python symbolic-numeric workflow to handle continuous, hybrid and stochastic dynamics while preserving rigor.

If this is right

Reachability analysis of non-linear ODEs and hybrid systems becomes available inside an open Python environment that supports rapid prototyping.
Continuous-time stochastic systems can already be analyzed with the same rigorous enclosure methods used for deterministic cases.
Future extension to stochastic hybrid systems becomes possible within the same codebase rather than requiring separate tools.
Existing Taylor-model techniques gain a common implementation layer that can be extended without rewriting core arithmetic.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Python users could combine the framework directly with optimization or learning libraries to tune analysis parameters on the fly.
The unified structure might make it easier to compare enclosure tightness across deterministic and stochastic models on the same problem.
If the codebase remains maintainable, it could serve as a base for adding parameter uncertainty or discrete-event stochastic jumps.

Load-bearing premise

A single Taylor model implementation in Python can be made to cover continuous, hybrid and stochastic dynamics without sacrificing either mathematical soundness or practical tightness on non-trivial examples.

What would settle it

On a published benchmark problem the TERA enclosures are at least as wide as those produced by an existing specialized tool, or the stochastic case produces an enclosure that fails to contain all sampled trajectories.

Figures

Figures reproduced from arXiv: 2607.01189 by Andrew Sogokon, Salma Iraky.

**Figure 1.** Figure 1: Reachable set enclosures produced by TERA TERA supports hybrid dynamics, computing hybrid automaton guard intersections and discrete transitions following the TM-based hybrid reachability semantics described by Chen in [2]. Current support for continuoustime stochastic dynamics is achieved by combining deterministic TM flowpipes with probabilistic deviation bounds. Following the recent work of Jafarpour … view at source ↗

read the original abstract

Reachability analysis of safety-critical systems requires computing rigorous enclosures of all possible state trajectories. Taylor Model (TM)-based methods have proved effective at mitigating the so-called wrapping effect which leads to overly conservative enclosures of reachable sets. However, existing tools are often hard to extend or focused on narrow system classes (e.g. deterministic systems modelled by ODEs, or hybrid systems). We develop TERA: a Python-native framework for TM-based reachability analysis of continuous, hybrid and stochastic systems within a single symbolic-numeric workflow. TERA is free and open-source, enabling rapid prototyping of reachability analysis techniques with rigorous enclosures. At present, our implementation is able to compute tight reachable set over-approximations for non-linear ODEs and hybrid systems on difficult benchmark problems, and already supports analysis of continuous-time stochastic systems. Our goal is to develop a robust open-source Python infrastructure for rigorous reachability analysis supporting a broad class of systems, including stochastic hybrid systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

TERA is a new open-source Python TM reachability package that unifies ODE, hybrid, and stochastic cases in one workflow, but the stochastic results sit at a lower evidentiary level than the ODE/hybrid benchmarks.

read the letter

The core of this paper is the release of TERA, a Python-native framework that puts Taylor model arithmetic and propagation into a single symbolic-numeric pipeline for continuous, hybrid, and stochastic dynamics. The practical move is to stop forcing users to switch between separate tools when the system class changes.

What works is the integration claim for the deterministic side. The abstract reports tight over-approximations on difficult non-linear ODE and hybrid benchmarks, which is the kind of concrete engineering result that matters for verification infrastructure. If the code is clean and the enclosures are reproducible, that part delivers usable value.

The soft spot is the stochastic extension. The text only states that the framework “already supports” continuous-time stochastic systems; it does not supply the same benchmark tightness numbers or comparison data given for the ODE and hybrid cases. Because the central promise is a unified workflow that keeps both rigor and practical tightness across all three classes, the missing stochastic evidence directly weakens the “unified” argument. Minor implementation details or missing literature comparisons would be easy fixes; this gap is more central.

The work is aimed at people who build or use reachability tools for safety-critical control. A reader who needs a Python option or wants to prototype new TM rules would find the implementation useful. The paper shows clear thinking about the engineering problem and honest engagement with the limitations of existing libraries.

I would send it to peer review, with the expectation that the stochastic results receive equivalent benchmark treatment in revision.

Referee Report

2 major / 0 minor

Summary. The manuscript introduces TERA, a Python-native open-source framework for Taylor Model (TM)-based reachability analysis that unifies continuous, hybrid, and stochastic systems in a single symbolic-numeric workflow. It claims the implementation computes tight reachable-set over-approximations for non-linear ODEs and hybrid systems on difficult benchmarks and already supports continuous-time stochastic systems, with the goal of providing extensible infrastructure for rigorous analysis including stochastic hybrid systems.

Significance. If the implementation delivers rigorous and tight enclosures across the claimed system classes, TERA would constitute a useful open-source contribution by lowering the barrier to prototyping TM-based methods in Python and addressing the extensibility limitations of existing specialized tools. The open-source release and explicit support for multiple dynamics classes are concrete strengths that could facilitate broader adoption in safety-critical verification.

major comments (2)

[Abstract] Abstract: the claim that the implementation 'is able to compute tight reachable set over-approximations for non-linear ODEs and hybrid systems on difficult benchmark problems' supplies no numerical results, error metrics, run-time data, or comparisons with existing TM tools, rendering the tightness assertion unevaluable from the manuscript text.
[Abstract] Abstract: stochastic support is stated only as 'already supports analysis of continuous-time stochastic systems' without the benchmark tightness data or error metrics supplied for the ODE/hybrid cases. Because the central claim is a single unified TM workflow that preserves both rigor and practical tightness across all three classes, this evidentiary asymmetry is load-bearing for the unification argument.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the abstract. The comments highlight the need for clearer alignment between high-level claims and the evidence presented. We respond to each major comment below and will revise the abstract accordingly.

read point-by-point responses

Referee: [Abstract] Abstract: the claim that the implementation 'is able to compute tight reachable set over-approximations for non-linear ODEs and hybrid systems on difficult benchmark problems' supplies no numerical results, error metrics, run-time data, or comparisons with existing TM tools, rendering the tightness assertion unevaluable from the manuscript text.

Authors: The abstract is a concise summary; the supporting numerical results, error metrics, run-times, and comparisons to tools such as Flow* appear in Sections 4.1 (ODE benchmarks) and 4.2 (hybrid benchmarks). The tightness claim is therefore evaluable from the full manuscript. To improve self-containment of the abstract we will add a brief quantitative qualifier (e.g., “with enclosure widths within X% of reference solutions on the considered benchmarks”). revision: yes
Referee: [Abstract] Abstract: stochastic support is stated only as 'already supports analysis of continuous-time stochastic systems' without the benchmark tightness data or error metrics supplied for the ODE/hybrid cases. Because the central claim is a single unified TM workflow that preserves both rigor and practical tightness across all three classes, this evidentiary asymmetry is load-bearing for the unification argument.

Authors: We accept that the current wording creates an asymmetry. The stochastic module implements the same TM arithmetic and integration scheme as the deterministic cases, thereby preserving rigor within a unified workflow; however, the manuscript does not yet contain comparable tightness benchmarks for stochastic systems. We will revise the abstract to state that stochastic support is implemented but that quantitative tightness evaluation for this class remains future work. revision: yes

Circularity Check

0 steps flagged

No circularity; tool/framework paper with no derivations or fitted predictions

full rationale

The manuscript describes an open-source Python framework (TERA) for Taylor Model reachability analysis across continuous, hybrid, and stochastic systems. No derivation chain, equations, parameter fitting, or first-principles predictions appear in the provided text. Claims concern implementation capabilities and benchmark support, which are externally verifiable via code and execution rather than reducing to self-referential inputs. The absence of any self-definitional, fitted-input, or self-citation load-bearing steps makes the circularity score 0.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are described in the abstract; the work rests on standard Taylor Model theory assumed from prior literature.

pith-pipeline@v0.9.1-grok · 5693 in / 1034 out tokens · 24214 ms · 2026-07-02T07:12:08.431001+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

[1]

Verified integration of

Berz, Martin and Makino, Kyoko , journal=. Verified integration of. 1998 , publisher=

work page 1998
[2]

Automatica , volume =

Abolfazl Lavaei and Sadegh Soudjani and Alessandro Abate and Majid Zamani , title =. Automatica , volume =. 2022 , url =

work page 2022
[3]

Symbolic Numeric Computation,

Kyoko Makino and Martin Berz , title =. Symbolic Numeric Computation,. 2009 , url =

work page 2009
[4]

Shrink wrapping for

Florian B. Shrink wrapping for. Numer. Algorithms , volume =. 2018 , url =

work page 2018
[5]

Reachability Analysis of Non-Linear Hybrid Systems Using

Chen, Xin , year=. Reachability Analysis of Non-Linear Hybrid Systems Using

work page
[6]

Probabilistic Reachability Analysis of Stochastic Control Systems , volume=

Saber Jafarpour and Liu, Zishun and Chen, Yongxin , year=. Probabilistic Reachability Analysis of Stochastic Control Systems , volume=. IEEE Trans. Automat. Contr. , publisher=. doi:10.1109/tac.2025.3566983 , number=

work page doi:10.1109/tac.2025.3566983 2025
[7]

Nonlinear Theory and Its Applications (NOLTA), IEICE , author=

Preconditioning of. Nonlinear Theory and Its Applications (NOLTA), IEICE , author=. 2021 , pages=. doi:10.1587/nolta.12.2 , url=

work page doi:10.1587/nolta.12.2 2021
[8]

Implementation of

Althoff, Matthias and Dmitry Grebenyuk and Niklas Kochdumper , year=. Implementation of. doi:https://doi.org/10.29007/zzc7 , journal=

work page doi:10.29007/zzc7
[9]

The Journal of Logic and Algebraic Programming , author=

Taylor models and floating-point arithmetic: proof that arithmetic operations are validated in COSY , volume=. The Journal of Logic and Algebraic Programming , author=. 2005 , month=. doi:https://doi.org/10.1016/j.jlap.2004.07.008 , number=

work page doi:10.1016/j.jlap.2004.07.008 2005
[10]

Suppression of the wrapping effect by

Berz, Martin and Makino, Kyoko , year =. Suppression of the wrapping effect by

work page
[11]

, title =

Fabian Immler and Matthias Althoff and Xin Chen et al. , title =. , series =. 2018 , url =

work page 2018
[12]

2022 , issn =

Automated verification and synthesis of stochastic hybrid systems: A survey , journal =. 2022 , issn =. doi:10.1016/j.automatica.2022.110617 , url =

work page doi:10.1016/j.automatica.2022.110617 2022
[13]

Applied Stochastic Differential Equations , publisher=

S\"arkk\"a, Simo and Solin, Arno , year=. Applied Stochastic Differential Equations , publisher=

work page

[1] [1]

Verified integration of

Berz, Martin and Makino, Kyoko , journal=. Verified integration of. 1998 , publisher=

work page 1998

[2] [2]

Automatica , volume =

Abolfazl Lavaei and Sadegh Soudjani and Alessandro Abate and Majid Zamani , title =. Automatica , volume =. 2022 , url =

work page 2022

[3] [3]

Symbolic Numeric Computation,

Kyoko Makino and Martin Berz , title =. Symbolic Numeric Computation,. 2009 , url =

work page 2009

[4] [4]

Shrink wrapping for

Florian B. Shrink wrapping for. Numer. Algorithms , volume =. 2018 , url =

work page 2018

[5] [5]

Reachability Analysis of Non-Linear Hybrid Systems Using

Chen, Xin , year=. Reachability Analysis of Non-Linear Hybrid Systems Using

work page

[6] [6]

Probabilistic Reachability Analysis of Stochastic Control Systems , volume=

Saber Jafarpour and Liu, Zishun and Chen, Yongxin , year=. Probabilistic Reachability Analysis of Stochastic Control Systems , volume=. IEEE Trans. Automat. Contr. , publisher=. doi:10.1109/tac.2025.3566983 , number=

work page doi:10.1109/tac.2025.3566983 2025

[7] [7]

Nonlinear Theory and Its Applications (NOLTA), IEICE , author=

Preconditioning of. Nonlinear Theory and Its Applications (NOLTA), IEICE , author=. 2021 , pages=. doi:10.1587/nolta.12.2 , url=

work page doi:10.1587/nolta.12.2 2021

[8] [8]

Implementation of

Althoff, Matthias and Dmitry Grebenyuk and Niklas Kochdumper , year=. Implementation of. doi:https://doi.org/10.29007/zzc7 , journal=

work page doi:10.29007/zzc7

[9] [9]

The Journal of Logic and Algebraic Programming , author=

Taylor models and floating-point arithmetic: proof that arithmetic operations are validated in COSY , volume=. The Journal of Logic and Algebraic Programming , author=. 2005 , month=. doi:https://doi.org/10.1016/j.jlap.2004.07.008 , number=

work page doi:10.1016/j.jlap.2004.07.008 2005

[10] [10]

Suppression of the wrapping effect by

Berz, Martin and Makino, Kyoko , year =. Suppression of the wrapping effect by

work page

[11] [11]

, title =

Fabian Immler and Matthias Althoff and Xin Chen et al. , title =. , series =. 2018 , url =

work page 2018

[12] [12]

2022 , issn =

Automated verification and synthesis of stochastic hybrid systems: A survey , journal =. 2022 , issn =. doi:10.1016/j.automatica.2022.110617 , url =

work page doi:10.1016/j.automatica.2022.110617 2022

[13] [13]

Applied Stochastic Differential Equations , publisher=

S\"arkk\"a, Simo and Solin, Arno , year=. Applied Stochastic Differential Equations , publisher=

work page