Autonomous digital laboratories require the synchronized orchestration of very different types of hardware and software. To manage this complexity, we designed a simple yet efficient set of control layers that can be generalized and applied to laboratories with different configurations. Our architecture follows a bottom-up, multilayer design: laboratory instruments are grouped by manufacturer under dedicated middleware, which in turn communicates with a centralized scheduler tightly coupled to our database. Chemists populate the scheduler database through the HCI. Another key component is the Data Processor, which is connected to the scheduler and extracts raw data from proprietary file formats to populate the Mat-file. Decision-making algorithms interact with the scheduler database to guide experiments, and results are compiled into reports for the chemist. Data are then transferred to an open database hosted with SDSC and SWITCH, making them publicly accessible.
|
IN COLLABORATION WITH SWISS DATA SCIENCE CENTER |
 |
Generating high-quality experimental datasets requires every process and analysis to be contextualized and linked to the samples produced during an experiment, including failed attempts. To achieve this, we developed the Matryoshka file: a structured ZIP container that gathers, in an open and human-readable form, all the metadata and data generated from synthesis to analytical characterization, with each sample traced through persistent identifiers. A major challenge lies in extracting relevant information from the proprietary file formats of instrument suppliers. We address this by progressively converting these outputs into the standardized Allotrope ASM-JSON format and mapping them to a dedicated ontology. In collaboration with the Swiss Data Science Center (SDSC), we also developed HT-CHEMBORD, a research data infrastructure and web interface through which these FAIR, machine-readable data can be searched, queried and downloaded for further analysis and AI applications.
A FAIR research data infrastructure for high-throughput digital chemistry
Orchestrating a fully automated laboratory requires coordinating many instruments, robots and control software from different suppliers. To pilot this equipment, we developed GLAS (Git-based Lab Automation Scheduler), an open-source, Python-based architecture. In our laboratory, several independent instances of GLAS operate below the main laboratory scheduler, each one dedicated to a group of equipment: for example, one instance manages all mobile and fixed robots for sample transfer, while another pilots the OmniFire, which handles fraction collection, sample preparation and injection into analytical instruments. Each instance receives tasks from the laboratory scheduler and executes them directly on the instruments through their own APIs or protocols, keeping complexity close to the equipment. This design allows tasks to run in parallel and makes it easy to add new equipment by deploying a new instance. A web interface displays running tasks and equipment status in real time, and GLAS is freely available on GitHub for other laboratories to adapt.
GLAS: an open-source easily expandable Git-based scheduling architecture for integral lab automation