Design-OS: A Specification-Driven Framework for Engineering System Design with a Control-Systems Design Case

H. Sinan Bank; Daniel R. Herber; Thomas H. Bradley

cs.CEMar 20, 2026

Design-OS: A Specification-Driven Framework for Engineering System Design with a Control-Systems Design Case

Authors: H. Sinan Bank, Daniel R. Herber, Thomas H. Bradley

Organizations: Department of Systems Engineering Colorado State University Fort Collins, CO 80523

Abstract

Engineering system design -- whether mechatronic, control, or embedded -- often proceeds in an ad hoc manner, with requirements left implicit and traceability from intent to parameters largely absent. Existing specification-driven and systematic design methods mostly target software, and AI-assisted tools tend to enter the workflow at solution generation rather than at problem framing. Human--AI collaboration in the design of physical systems remains underexplored. This paper presents Design-OS, a lightweight, specification-driven workflow for engineering system design organized in five stages: concept definition, literature survey, conceptual design, requirements definition, and design definition. Specifications serve as the shared contract between human designers and AI agents; each stage produces structured artifacts that maintain traceability and support agent-augmented execution. We position Design-OS relative to requirements-driven design, systematic design frameworks, and AI-assisted design pipelines, and demonstrate it on a control systems design case using two rotary inverted pendulum platforms -- an open-source SimpleFOC reaction wheel and a commercial Quanser Furuta pendulum -- showing how the same specification-driven workflow accommodates fundamentally different implementations. A blank template and the full design-case artifacts are shared in a public repository to support reproducibility and reuse. The workflow makes the design process visible and auditable, and extends specification-driven orchestration of AI from software to physical engineering system design.

Figures & tables

Figure 1: Design-OS workflow (abstract): five stages arranged around the designer (central diamond), each executed by an AI agent (robot icon). User interactions (numbered) between stages enable validation and iteration. Each box shows the stage name, key artifacts, and commands.

Figure 2: Two inverted pendulum architectures: Furuta (left) and reaction wheel (right).

ID	Domain	Requirement (shall)	Verif.
REQ-01	Perf.	Damping ratio $\zeta=0.7$	Sim.
REQ-02	Perf.	Natural freq. $\omega_{n}=4$ rad/s	Sim.
REQ-03	Perf.	$\|\alpha\|<15$ deg	Sim.
REQ-04	Perf.	$\|V_{m}\|<10$ V	Sim.
REQ-05	Mech.	State-space from first-principles EOM + linearization	Review
REQ-06	Mech.	Parameters documented with source	Review

Table 1: Requirements by domain for the rotary inverted pendulum.

REQ	Specification	Simulation	Result
01	$\zeta=0.7$	$\zeta=0.700$	Pass
02	$\omega_{n}=4$ rad/s	$\omega_{n}=4.000$ rad/s	Pass
03	$\|\alpha\|<15^{\circ}$	$\|\alpha\|_{\max}=2.9^{\circ}$	Pass
04	$\|V_{m}\|<10$ V	$\|V_{m}\|_{\max}=0.5$ V	Pass

Table 2: Simulation verification of performance requirements.

REQ	Domain flow	Verification
01, 02	Perf $\to$ SW (gain $K$ ) $\to$ Elec ( $V_{m}$ ) $\to$ Mech ( $\tau$ )	Sim. step resp.
03	Perf $\to$ Mech (pendulum $\alpha$ )	Sim. deflection
04	Perf $\to$ Elec (amplifier limit)	Sim. $V_{m}$
05	Mech + Elec $\to$ $A,B$ matrices	Derivation
07	Elec ( $R_{m},k_{m},k_{t},K_{g}$ ) $\to$ $A,B$	Review
11	SW (pole placement) $\to$ $K$	Design rationale

Table 3: Cross-domain traceability: requirement

\to

domain flow

\to

verification.

Domain	SimpleFOC	Quanser SRV02
Mechanical	3D-printed, sys. ID needed [ 32 ]	Aluminum arm + pendulum, known params
Electrical	BLDC + FOC driver, Arduino	DC servo, gearbox, commercial DAQ
Software	C++ + SimpleFOC lib.	MATLAB/Simulink, QUARC
Pendulum	Reaction wheel (2-DOF)	Furuta (2-DOF)
Cost	$\sim$ $100–200	$\sim$ $5k+

Table 4: Platform comparison: same requirements, different implementations.

	Claude Opus 4.6	Gemini 3.1 Pro
Overview	36 artifacts, 29 checkpoints, 24 corrections. Python (model choice). Platforms: Quanser SRV02 + SimpleFOC reaction wheel.	$\sim$ 25 artifacts, 28 checkpoints, 14 corrections. MATLAB (tech-stack). Platforms: Quanser QUBE + ESP32.
Stage 1: Concept Def.	Complete — 0 errors. Mission, roadmap, tech stack correct; open tech stack appropriately deferred.	Complete — 0 errors. Mission, roadmap, tech stack correct; MATLAB explicitly committed.
Stage 2: Lit. Survey	Good — 21 ref. errors caught by /verify-references ; 8-theme coverage; pipeline step self-initiated.	Fair — 55 sources; 2 citation mismatches found by post-hoc /verify-references (2026-03-07); step not self-initiated.
Stage 3: Concept. Design	Good — 1 DOF factual error (1-DOF stated, 2-DOF correct); platform selection and functions correct.	Good — 1 domain error (Control/Math domain missing from decomposition); platforms, functions correct.
Stage 4: Req. Def.	Good — 12 requirements, 4 formal standards ( /discover-standards ); 2 minor corrections.	Poor — 7 requirements; /validate-feasibility gate skipped; /discover-standards not run; 3 specification errors: scope deviation, phantom REQ-HWW IDs, HIL in REQ-IMP-002.
Stage 5: Design Def.	Complete — both platforms ALL PASS; specification-driven order maintained; 1 numerical error (open-loop poles).	Fair — both platforms pass; execution order corrected; tool mismatch between MATLAB artifacts and Python report.

Table 5: AI model output quality by Design-OS stage. Source data: log/report_claude.md , log/report_gemini.md .

Correction	Claude Opus 4.6	Gemini 3.1 Pro
Factual errors	20	1
Missing reqs.	0	1
Halluc. refs.	3	2
Struct. deviat.	1	9
Domain errors	0	1
Total	24	14

Table 6: Human corrections needed per model.

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Stage	Command	Purpose
1: Concept Definition	/plan-project	Establish mission, roadmap, and tech stack through interactive dialogue
2: Literature Survey	/literature-search	Run landscape search and produce lit note, deep-research prompts, and scholar queries
	/focused-literature	Targeted search driven by conceptual design gaps
	/verify-references	Web-verify cited references and log corrections
3: Conceptual Design	/conceptual-design	Develop design objectives (DOs), methodologies, and high-level structure from literature
	/domain-decomposition	Identify system domains and interfaces from literature references

Appendix

Table 7: Design-OS commands by stage.

Explore similar work

GENAI WORKBENCH: AI-Assisted Analysis and Synthesis of Engineering Systems from Multimodal Engineering Data
Feb 27, 2026H. Sinan Bank, Daniel R. HerberModel-Based Systems Engineering
Department of Systems Engineering, Colorado State University, Fort Collins, Colorado, USA
Closed-loop AI achieves certifiable engineering design
Aug 22, 2026Tianyi Yu, Chengxing Tao, Haoxuan Shen +8Large Language Model-Guided Optimization
State Key Laboratory of Ocean Sensing & Ocean College, Zhejiang University, Zhoushan, 316021, China · Zhejiang Province Key Laboratory of Offshore Civil Engineering and Materials & Ocean Academy, Zhejiang University, Zhoushan, 316021, PR China · Institute of Fundamental and Transdisciplinary Research, Zhejiang University, 310058, PR China +3
Automotive Engineering-Centric Agentic AI Workflow Framework
Apr 9, 2026Tong Duy Son, Zhihao Liu, Piero Brigida +4Agentic Workflows Workflow Automation
The authors are with Siemens Digital Industries Software, Leuven, Belgium.

Feb 27, 2026J/K move · Enter open · S save

GENAI WORKBENCH: AI-Assisted Analysis and Synthesis of Engineering Systems from Multimodal Engineering Data

H. Sinan Bank, Daniel R. Herber

Department of Systems Engineering, Colorado State University, Fort Collins, Colorado, USA

Model-Based Systems Engineering

Open paper AskPDF arXiv

Modern engineering design platforms excel at discipline-specific tasks such as CAD, CAM, and CAE, but often lack native systems engineering frameworks. This creates a disconnect where system-level requirements and architectures are managed separately from detailed component design, hindering holistic development and increasing integration risks. To address this, we present the conceptual framework for the GenAI Workbench, a Model-Based Systems Engineering (MBSE) environment that integrates systems engineering principles into the designer's workflow. Built on an open-source PLM platform, it establishes a unified digital thread by linking semantic data from documents, physical B-rep geometry, and relational system graphs. The workbench facilitates an AI-assisted workflow where a designer can ingest source documents, from which the system automatically extracts requirements and uses vision-language models to generate an initial system architecture, such as a Design Structure Matrix (DSM). This paper presents the conceptual architecture, proposed methodology, and anticipated impact of this work-in-progress framework, which aims to foster a more integrated, data-driven, and informed engineering design methodology.

Abstract

Explore similar work

GENAI WORKBENCH: AI-Assisted Analysis and Synthesis of Engineering Systems from Multimodal Engineering Data

Closed-loop AI achieves certifiable engineering design

Automotive Engineering-Centric Agentic AI Workflow Framework