03 / EMBODIED SYSTEMS

AI-Native Robotics

Can AI help design the whole robot—not just its behavior?

Treat a robot as one connected design problem. Use AI throughout circuit-board design, embedded software, control systems, and mechanical construction, rather than adding intelligence only after the hardware is finished.

RESEARCH BRIEF

Can AI help design the whole robot—not just its behavior?

Robots and robotic arms designed with AI—from circuit boards and software to control and mechanical structure.

This HARD Lab direction studies the problem as an end-to-end system rather than as a single model demo. Current related work provides useful building blocks and baselines [1] [2] [3] [4] [5]; the lab’s goal is to connect them into measurable, reproducible research artifacts.

01

AI-assisted circuit, firmware, and mechanical co-design

02

Control, sensing, and hardware integration

03

Simulation-to-prototype iteration and physical evaluation

PROPOSED WORKFLOW

Build, measure, iterate.

01

Co-design

Connect mechanical requirements, circuit design, embedded code, and control objectives in a shared design loop.

02

Simulate

Evaluate candidate designs and behaviors before building or changing a physical prototype.

03

Build & test

Integrate the hardware and software, then use physical measurements to guide the next iteration.

STUDENT ENTRY POINTS

Ways to start contributing.

  • AI-assisted circuit, firmware, and mechanical co-design
  • Control, sensing, and hardware integration
  • Simulation-to-prototype iteration and physical evaluation
USEFUL SKILLSEmbedded systemsElectronicsCAD / MechanicsControl systems

RELATED WORK / 2026-09-10

Selected papers & citations.

Five primary-source papers selected to frame this direction. Links point to the authors’ arXiv records; these are external works, not HARD Lab publications.

Download BibTeX

5 selected papers · newest first by initial submission

01
2026Benchmark preprintarXiv:2608.05539

OmniMech: All-in-one Multimodal Mechanical Benchmark for 3D Reconstruction

Taiting Lu, Runze Liu, Ziwei Dong, Sisong Bei, Jingying Zeng, Mingjia Wang, Zhenghao Li, Kaiyuan Lin, Yi-Shan Wu, Yangshoudu Zheng, Hongxing Pan, Kai Zhang, Guoliang Shi, Ling Ma, Yifan Yang, Jiaying Lu, Qi He, Sung-Liang Chen, Yi-Chao Chen, Yincheng Jin, Mahanth Gowda

Evaluates reconstruction of mechanical designs into executable CAD, including engineering constraints rather than appearance alone.

Why it matters here. A model for checking dimensions and tolerances in generated parts. It is not evidence of a fully autonomous robot-design pipeline.

Abstract PDF
Full citation & BibTeX

Taiting Lu et al. (2026). OmniMech: All-in-one Multimodal Mechanical Benchmark for 3D Reconstruction. arXiv:2608.05539. https://doi.org/10.48550/arXiv.2608.05539

@misc{hardlab_omnimech,
  title = {{OmniMech: All-in-one Multimodal Mechanical Benchmark for 3D Reconstruction}},
  author = {Taiting Lu and Runze Liu and Ziwei Dong and Sisong Bei and Jingying Zeng and Mingjia Wang and Zhenghao Li and Kaiyuan Lin and Yi-Shan Wu and Yangshoudu Zheng and Hongxing Pan and Kai Zhang and Guoliang Shi and Ling Ma and Yifan Yang and Jiaying Lu and Qi He and Sung-Liang Chen and Yi-Chao Chen and Yincheng Jin and Mahanth Gowda},
  year = {2026},
  eprint = {2608.05539},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2608.05539},
  url = {https://arxiv.org/abs/2608.05539},
  note = {External related work; metadata checked 2026-09-10}
}
02
2026Benchmark preprintarXiv:2607.03261

OmniLayout: A Schematic-Coupled Multimodal Benchmark for Constraint-Aware Geometric Reasoning in PCB Layout

Taiting Lu, Kaiyuan Lin, Mingjia Wang, Haolin Ye, Runze Liu, Yuxin Tian, Vahe Melkonyan, Haoyu Wang, Muchuan Wang, Chufan Hong, Yifan Yang, Sung-Liang Chen, Yi-Chao Chen, Yicheng Jin, Mahanth Gowda

Examines PCB-layout reasoning in conjunction with schematic connectivity, component placement, and geometric constraints.

Why it matters here. Supports evaluation beyond a plausible board image: connectivity, placement constraints, and routability must also be checked.

Abstract PDF
Full citation & BibTeX

Taiting Lu et al. (2026). OmniLayout: A Schematic-Coupled Multimodal Benchmark for Constraint-Aware Geometric Reasoning in PCB Layout. arXiv:2607.03261. https://doi.org/10.48550/arXiv.2607.03261

@misc{hardlab_omnilayout,
  title = {{OmniLayout: A Schematic-Coupled Multimodal Benchmark for Constraint-Aware Geometric Reasoning in PCB Layout}},
  author = {Taiting Lu and Kaiyuan Lin and Mingjia Wang and Haolin Ye and Runze Liu and Yuxin Tian and Vahe Melkonyan and Haoyu Wang and Muchuan Wang and Chufan Hong and Yifan Yang and Sung-Liang Chen and Yi-Chao Chen and Yicheng Jin and Mahanth Gowda},
  year = {2026},
  eprint = {2607.03261},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2607.03261},
  url = {https://arxiv.org/abs/2607.03261},
  note = {External related work; metadata checked 2026-09-10}
}
03
2026Research preprintarXiv:2606.01188

pcbGPT: Automatic PCB Schematic Synthesis from Natural Language Requirements

Tobias King, Steven Kehrberg, Michael Beigl, Tobias Röddiger

Uses a Python design representation and tool-assisted validation to synthesize editable KiCad schematics from requirements.

Why it matters here. A direct reference for the electronics branch of AI-native robotics, with generated schematics treated as testable artifacts.

Abstract PDF
Full citation & BibTeX

Tobias King et al. (2026). pcbGPT: Automatic PCB Schematic Synthesis from Natural Language Requirements. arXiv:2606.01188. https://doi.org/10.48550/arXiv.2606.01188

@misc{hardlab_pcbgpt,
  title = {{pcbGPT: Automatic PCB Schematic Synthesis from Natural Language Requirements}},
  author = {Tobias King and Steven Kehrberg and Michael Beigl and Tobias Röddiger},
  year = {2026},
  eprint = {2606.01188},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2606.01188},
  url = {https://arxiv.org/abs/2606.01188},
  note = {External related work; metadata checked 2026-09-10}
}
04
2026Research preprintarXiv:2605.19717

Physics-in-the-Loop: A Hybrid Agentic Architecture for Validated CAD Engineering Design

Elias Berger, Muhammad Usama, Jan Mehlstäubl, Bernhard Saske, Kristin Paetzold-Byhain

Connects agentic CAD design to engineering knowledge and tool-based physical validation.

Why it matters here. Motivates a generate–simulate–measure loop for mechanical parts, rather than accepting a design based on its visual plausibility.

Abstract PDF
Full citation & BibTeX

Elias Berger et al. (2026). Physics-in-the-Loop: A Hybrid Agentic Architecture for Validated CAD Engineering Design. arXiv:2605.19717. https://doi.org/10.48550/arXiv.2605.19717

@misc{hardlab_physics_loop,
  title = {{Physics-in-the-Loop: A Hybrid Agentic Architecture for Validated CAD Engineering Design}},
  author = {Elias Berger and Muhammad Usama and Jan Mehlstäubl and Bernhard Saske and Kristin Paetzold-Byhain},
  year = {2026},
  eprint = {2605.19717},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2605.19717},
  url = {https://arxiv.org/abs/2605.19717},
  note = {External related work; metadata checked 2026-09-10}
}
05
2023Foundational work · arXiv versionarXiv:2311.01455

RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation

Yufei Wang, Zhou Xian, Feng Chen, Tsun-Hsuan Wang, Yian Wang, Katerina Fragkiadaki, Zackory Erickson, David Held, Chuang Gan

Explores generative simulation to produce environments, tasks, and robot-learning experiences.

Why it matters here. Provides a reference for simulation and skill training, not for PCB fabrication or end-to-end mechanical design.

Abstract PDF
Full citation & BibTeX

Yufei Wang et al. (2023). RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation. arXiv:2311.01455. https://doi.org/10.48550/arXiv.2311.01455

@misc{hardlab_robogen,
  title = {{RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation}},
  author = {Yufei Wang and Zhou Xian and Feng Chen and Tsun-Hsuan Wang and Yian Wang and Katerina Fragkiadaki and Zackory Erickson and David Held and Chuang Gan},
  year = {2023},
  eprint = {2311.01455},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2311.01455},
  url = {https://arxiv.org/abs/2311.01455},
  note = {External related work; metadata checked 2026-09-10}
}