Automation and machine learning provide powerful tools to accelerate the pace of scientific discovery. Self-driving labs, which combine automation and machine learning, can free up researchers to focus on generating hypotheses and answering important scientific questions by reducing human involvement in performing routine tasks.<br/> <br/>Self-driving labs designed for a broad range of tasks struggle to achieve good accuracy when handling very small amounts of material. Consequently, results may not reflect the actual experimental conditions, or the range of explored experimental variables may need to be restricted in order to ensure accurate results. However, tuning the process parameters to find optimal conditions is quite tedious, and not an efficient use of a researcher’s time. By establishing a fully autonomous system for parameter optimization, we can remove the necessity for human involvement and improve research efficiency.<br/> <br/>I will discuss the development of a fully autonomous system – based on a closed loop of multi-objective Bayesian optimization and robotic experiments – for optimizing instrument parameters in order to achieve highly accurate sample dosing. Building an automated lab requires careful planning, particularly when weighing generality versus specificity. For example, the manufacturer specified operating range or accuracy of an instrument may limit what experiments can be done in the future. On the other hand, an instrument with a larger specified operating range – or multiple instruments with different operating ranges – can be very expensive. Using this approach for autonomous parameter optimization, automated instruments with more general capabilities can be successfully used in experiments requiring the capabilities of a specialized instrument.

Can robots learn to have “Good Hands”? Fully autonomous optimization of accurate sample dosing parameters

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.