Talking to instruments: device-integration standards for self-driving laboratories

@misc{gottstein2026talking,
  title = {{Talking to instruments: device-integration standards for self-driving laboratories}},
  author = {Willi Gottstein and Lukas Bromig and Matthias Arnold and Sina Barazandeh and Joseph S Brown and Keith A. Brown and Ivano E. Castelli and Patrick Courtney and Basita Das and Martin Fitzner and Tobias Ginsburg and Aldair E. Gongora and Sang Soo Han and Annica Heyne and Georg Hinkel and Kedar Hippalgaonkar and Sergei V. Kalinin and Nicholas Kleiner and Christian Knittl-Frank and Jose Lugo-Martinez and Benji Maruyama and Pascal Miéville and Ankush Kumar Mishra and Timothy Noël and Ozlem Ozcan and Noah Paulson and Ilja Porohovoj and Mark Z. Porter and Zekun Ren and Harmen M. van Rossum and Bastian Ruehle and Samuel Vitor Saraiva and Elia Savino and Philippe Schwaller and Woosuck Shin and Joshua W. Sin and Eric T. Sletten and Arthur Sloan and Tomasz M. Stawski and Casey Stone and Ryo Tamura and Dean Thomas and Tejs Vegge and Robert Waelder and Ilya Yakavets and Hyuk Jun Yoo and Naruki Yoshikawa and Sergio Pablo-García},
  howpublished = {ChemRxiv},
  year = {2026},
  month = oct,
  doi = {10.26434/chemrxiv.15009990/v1},
  url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009990/v1},
  note = {CC BY},
}

TY  - UNPB
TI  - Talking to instruments: device-integration standards for self-driving laboratories
AU  - Willi Gottstein
AU  - Lukas Bromig
AU  - Matthias Arnold
AU  - Sina Barazandeh
AU  - Joseph S Brown
AU  - Keith A. Brown
AU  - Ivano E. Castelli
AU  - Patrick Courtney
AU  - Basita Das
AU  - Martin Fitzner
AU  - Tobias Ginsburg
AU  - Aldair E. Gongora
AU  - Sang Soo Han
AU  - Annica Heyne
AU  - Georg Hinkel
AU  - Kedar Hippalgaonkar
AU  - Sergei V. Kalinin
AU  - Nicholas Kleiner
AU  - Christian Knittl-Frank
AU  - Jose Lugo-Martinez
AU  - Benji Maruyama
AU  - Pascal Miéville
AU  - Ankush Kumar Mishra
AU  - Timothy Noël
AU  - Ozlem Ozcan
AU  - Noah Paulson
AU  - Ilja Porohovoj
AU  - Mark Z. Porter
AU  - Zekun Ren
AU  - Harmen M. van Rossum
AU  - Bastian Ruehle
AU  - Samuel Vitor Saraiva
AU  - Elia Savino
AU  - Philippe Schwaller
AU  - Woosuck Shin
AU  - Joshua W. Sin
AU  - Eric T. Sletten
AU  - Arthur Sloan
AU  - Tomasz M. Stawski
AU  - Casey Stone
AU  - Ryo Tamura
AU  - Dean Thomas
AU  - Tejs Vegge
AU  - Robert Waelder
AU  - Ilya Yakavets
AU  - Hyuk Jun Yoo
AU  - Naruki Yoshikawa
AU  - Sergio Pablo-García
PY  - 2026
DA  - 2026/10/06
PB  - ChemRxiv
DO  - 10.26434/chemrxiv.15009990/v1
UR  - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009990/v1
AB  - Self-driving laboratories (SDLs) are rapidly outgrowing the single-instrument prototype, scaling into multi-instrument and multi-site platforms that run autonomous discovery across chemistry, materials science, biotechnology, pharmaceutical re-search and adjacent domains. What now decides whether an SDL can scale, share its integrations and federate experiments is no longer the algorithms or the robots, but the software layer that orchestrates heterogeneous devices. Today that layer is the bottleneck: a fast-growing ecosystem of open-source orchestrators in which most have built device integration on their own bespoke conventions instead of adopting a formally governed standard. The result is predictable and costly. Integrations do not transfer between laboratories, orchestrators cannot interoperate and many new SDLs rebuild their device-control layer from scratch. Crucially, this is an engineering coordination problem, not a fundamental scientific one and therefore among the most tractable bottlenecks in autonomous science. We map the device-integration standards landscape onto the open-source orchestrator ecosystem, compare the leading options and give scenario-driven guidance on when to use each. We then translate the requirements of autonomous operation into a concrete deliverable: a set of testable engineering criteria and a weighted procurement scorecard that defines an autonomous-ready control interface and gives institutional buyers the leverage to demand it; these requirements do not change as orchestration turns agentic: whether an instrument is driven by a program or by a language-model agent through a model-facing protocol, it is only as safe and reliable as the interface it exposes. Adopted at scale, these requirements turn device integration from a burden repaid lab by lab into shared infrastructure,
ER  -