AUTONOMOUS SCIENTIFIC DISCOVERY

Evaluation of the performance of a given molecule, material or device can be resource demanding but remains a tractable problem. The challenge lies in the inverse problem, i.e., the identification of designs with target properties since the configuration space can be enormous. It is desirable to circumvent the trial and error nature of the lab and accelerate innovation by using supercomputers to conduct virtual high-throughput experiments and by using scalable, intelligent algorithms, for example based on machine learning, to automate the scientific process towards uncertainty-aware, autonomous research labs. Development of workflows for integrating information streams from modeling (e.g. multiscale modeling), experiment (e.g. characterization), and other sources (e.g. literature) nevertheless pose challenges like devising effective protocols for information exchange at the interface of knowledge domains.

Schematic of (a) an agentic AI-based system for
(b) autonomous research labs.

ADAPTIVE SYSTEMS FOR MOBILITY AND INDUSTRY

Systems that convert scientific insights into deployable technology at scale are often built around modular architectures that combine physics-based models, real-time data streams, and decision layers. This stack enables adaptive operation, while improving robustness when data are sparse or operating regimes shift, and reduces reliance on purely empirical models that can significantly deviate outside initial training conditions. Such approaches are useful for example as digital twins for battery management and the predictive maintenance of vehicle fleets.

A computing platform architecture for a battery digital twin.

INNOVATION STRATEGY FOR EMERGING TECHNOLOGIES

Many innovations in materials, processes, and devices have been achieved at the intersections of distinct but complementary domains such as nanoscience and nanotechnology, clean energy technology, and computer and information science. To prepare technology roadmaps, allocate research funding, or devise policy instruments that promote innovation and competitive advantage, it is important to understand the type and extent of interdisciplinary research and related technology development efforts across a firm, industry, or national innovation systems. This input also helps businesses, funding agencies, and policymakers plan for transformative technologies such as robotics and artificial intelligence to enable productivity gains, align workforce development, and effectively respond to labor market needs.

OECD framework for measuring national AI compute capacity and
ensuring ongoing monitoring.

See Works for outputs.