Speeding Up “Self-Driving” Labs

A new design for self-driving labs can generate ten times the data from these autonomous platforms while exploring new chemistry and materials at least three times as quickly as older versions.

---

▲ Although this is an exaggerated depiction of what a futuristic laboratory might look like, a new method of designing self-driving labs developed by engineers at North Carolina State Univ. is pushing the boundaries of what’s possible for automated labs today.

Self-driving labs combine robotics with machine learning to make discoveries in chemistry or materials science. These platforms can autonomously run chemical reactions in a closed loop, making decisions based on previous experimental outcomes about which experiments to try next in order to reach a certain target. The reactions can be done in batches or as a continuous flow, but until now, the labs have been designed to wait until the experiment reaches a steady state before sampling data.

“If the reaction is ten minutes, you’re going to wait twenty to thirty minutes before you do one measurement,” says Milad Abolhasani, a professor of chemical and biomolecular engineering at North Carolina State Univ. “We wanted to remove that 20–30-minute wait time and do continuous experiments in a dynamic mode so we can push the data generation capabilities.”

Machine learning algorithms are only as good as the data that feeds them, Abolhasani adds, so generating large amounts of data on chemical processes is important for boosting the autonomous decision-making of these...