| Abstract Scope |
Recent advancements in AI can potentially accelerate progress in materials science by enabling autonomous experiment (AE) methodologies. However, current AE frameworks often struggle to manage the complexity and variability in real-world experimental conditions or to adapt dynamically to evolving research objectives—tasks at which human experts excel. In this talk, I will introduce an approach enabling integrating prior knowledge and human oversight into active learning loops of AE. This transforms pure data-driven AE to multi-stage, knowledge-informed decision-making processes, enabling more refined exploration and facilitating the discovery of complex functionalities. We demonstrate the application of this approach in autonomous thin-film processing and microscopy. However, this approach is applicable to a wide range of AE, offering an adaptive way for advancing AEs.
Acknowledgments: This research was supported by the Center for Nanophase Materials Sciences, which is a US Department of Energy, Office of Science User Facility at Oak Ridge National Laboratory. |