NSF Workshop on Self-Driving Laboratories

NSF SDL 2023 Workshop

This invitation-only, NSF-sponsored workshop on Self-Driving Laboratories (SDLs) aims to develop a 10-year research roadmap for automation of scientific experiments.

SDLs have the potential to make new practices of science possible, accelerating scientific progress. In particular, advances in low-cost sensors, actuators, robotic systems, and control systems, have lowered the barrier to entry to laboratory automation such that fully automated labs are leaving the realm of science fiction, while advances in artificial intelligence are enabling large degrees of autonomoy. However, whereas science overall is a strongest-link problem, with the best solution pushing the field forward, SDLs are a weakest-link problem, whereby all components of the system need to function for scientists to be able to rely on them as infrastructure. Therefore, advancing SDLs requires the identification, development, dissemination, and reproduction of core infrastructural components that can be used broadly to support scientists in creating and operating self-driving laboratories. In so doing, we likely draw from best practices in data science, human-machine interaction, manufacturing and quality control, open-source ecosystems, and laboratory science methods.

The workshop will bring together approximately 20 experts in the field to develop a 10-year roadmap for automation of scientific experiments. It will take place in Atlanta, Georgia, starting with dinner on the evening of November 13, and ending on November 15 at 1 PM. An NSF grant will provide travel support for participants in the workshop.

Schedule

Note! The Monday dinner is held at the Hilton Garden Inn Atlanta Midtown, but most workshop participants are staying at the Sonesta Select (Atlanta Midtown Georgia Tech), at 1132 Techwood Dr NW, Atlanta, GA 30318, which is a short walk away.

Monday

Tuesday

GA Tech Global Learning Center, room 335, 81 4th St NW Atlanta, GA 30332

As a first step towards defining a 10-year research roadmap for automation of scientific experiments, it is important to understand the current state of the art in relevant areas.

Thus a first goal is to collect representative examples of SDLs of different types and scales, from research prototypes to substantial operations. For each, we’d like to capture such details as: science domain(s) targeted/supported, or educational mission; examples of applications/workflows operated/targeted; location, US vs international; academia vs labs vs industry; instruments involved, degree of automation and autonomy, technology readiness level; scale and scalability; and whether general- or special-purpose. We are also interested in experiences with integrating across multiple labs.

We also encourage attendees to collect insights from these examples as to what works well and less well, and to identify methods that show promise for generalization and scaling.

Supporting infrastructure for interoperability, connectivity, and data sharing/protection will be critical to successful autonomous laboratory systems. This theme focuses on the infrastructure needed to ensure flexible self-driving laboratories that have the ability to interface with new hardware and analytic tools, as well as data architectures. This will enable teams using these laboratories to selectively share data, protocols and hardware capabilities in a secure and efficient manner. In this session, we would like to discuss the protocols, standards and capabilities that are necessary to interface self-driving laboratory hardware, software and facilities that will maximize system interoperability, secure data sharing and upgradability to future capabilities. In particular, we would like to focus on those concepts that are crosscutting between application domains such that efforts on one domain are applicable to other domains. Questions to be addressed include: How should hardware elements be interfaced at a system level (e.g., open standards). How should software elements be interfaced at a system level (e.g., open source code, or at least published protocols for software interface capability). How should data be curated, stored, and shared in a secure and selected manner ensuring both protection and accessibility of information.

While science is organized around disciplines, self-driving labs are a cross-cutting method that can be used in many scientific contexts. Self-driving lab infrastructure is highly valuable for many different scientists, yet contributing to self-driving labs’ infrastructure is not incentivized in the same ways as other science practice. Furthermore, the diversity of experimental approaches and priorities can create challenges for cross-disciplinary collaboration, even if scientists share infrastructural goals.

In this session, we would like to discuss what could be done to support cross-disciplinary collaboration towards shared infrastructural goals for self-driving labs. What are examples of successful cross-disciplinary infrastructural projects and communities, and how can we learn from them? What makes self-driving lab infrastructure different, and how could we address these challenges? What are existing communities that collaborate on self-driving labs, e.g. through replication, tailoring, or extension?

Beyond development, maintenance is key to infrastructural longevity. What are key maintenance challenges, and what can be done to support a community that faces them? What policies and strategic funding could support long term sustainability?

Wednesday

GA Tech Global Learning Center, room 335, 81 4th St NW Atlanta, GA 30332

The workshop organizers are Ian Foster (UChicago & Argonne), Nadya Peek (UW), and Tom Kurfess (GATech).