Self-Driving Laboratory

Researchers in a traditional laboratory come up with a hypothesis, conduct experiments, and eventually use the evidence to prove or adjust their initial idea. In a self-driving laboratory (SDL), this whole process is automated end to end, giving a virtual scientist (a computer) the opportunity to gather physical experiment data. We look into autonomous experimentation, robotic specimen handling, closed loop test data feedback, and the integration of testing into existing manufacturing machines.

Circular flow chart showing the steps in a self-driving lab and where humans are required for sample handling and data transfer. — Experimental optimization in a lab often relies on human intervention to transfer specimens from manufacturing to testing and feed results back into the optimization model. Self-driving laboratories aim to automate all these steps, to achieve a closed loop feedback cycle of design, manufacturing, testing, and optimizaion. (ETH Zürich)

Self-driving laboratories extend human-centered discovery by enabling computer systems to plan, execute, and interpret experiments while researchers define the scientific questions and orchestrate on a higher level. Over the past years, these systems have evolved from specialized automation tools into broader discovery platforms that autonomously execute closed-loop design–make–test cycles. Our goal is to accelerate discovery by reducing downtime, improving reproducibility, and scaling experimentation across multiple machines.

Our research is focused on additive manufacturing, especially fused deposition modeling (FDM). Our current interests include (this list is not exhaustive):

Material mechanics: Investigating how materials behave mechanically within 3D-printed structures, e.g., under compression.
Process optimization: Automatically tuning printing parameters using sensor feedback and evaluating how these parameters influence the properties of the finished part.
Optical inspection: Using image-based measurements to assess printed objects and provide feedback for subsequent design and manufacturing decisions.
Functional materials: Integrating electrically conductive polymer composites and flexible polymers into printed structures, with automated testing of their mechanical and functional performance.
Software interfaces for autonomous experimentation: Developing APIs and integration tools that enable autonomous agents to control manufacturing and testing equipment and interpret experimental results.

Timeline showing when specimens were produced in an experimental campaign of n=17 specimens in a self-driving laboratory. — Experiments are running sequentially on a single machine, and for each step we have initial cleaning of the printer, then actual printing, followed by cooling and finally the test. (ETH Zürich)

For more information please contact Tom Stein or Mark Zander.