RoboLink – Intelligent Metadata Generation and Machine Learning-Enhanced Experiment Planning for Robotics

RoboLink aims to establish an intelligent, fully integrated workflow that connects a multifunctional robotic testing platform, machine-readable data management, and machine learning–driven experimental planning to accelerate materials discovery. Robotic platform is designed for high-throughput testing of metallic materials in the electrolytes of varying complexity. The project addresses a central challenge in high-throughput experimentation: transforming heterogeneous, complex experimental data into uniform, high-quality, reusable (meta)datasets that can be directly leveraged for adaptive learning and decision-making.

At the core of RoboLink is the coupling of automated robotic experimentation with the electronic lab notebook Kadi4Mat. Robotic platform at Hereon enable autonomous, high-throughput execution of experimental protocols with high precision and reproducibility. Experimental data and associated metadata are automatically extracted from raw files generated by the Autosuite software controlling the robotic platform and integrated into Kadi4Mat in a structured, machine-readable format. This ensures data integrity, traceability, and accessibility, which are essential prerequisites for robust machine learning (ML) applications.

Building on this curated data foundation, RoboLink integrates quantitative structure–property relationship (QSPR) ML models to analyze experimental results, identify promising chemical compounds, and support adaptive experiment planning. The continuous feedback loop between new experimental results, data curation in Kadi4Mat, and ML-driven analysis enables progressively improved predictions of target material properties and more efficient experimental design.

A major goal of RoboLink is to integrate Kadi4Mat in the dataflow of the robotic testing platform. This includes extending backend capabilities for automated validation of metadata completeness and consistency, improving the web interface for intuitive interaction with large datasets, and enhancing Kadi-APY to support high-throughput data submission and advanced metadata querying. These developments enable scalable, automated workflows and lower the barrier for adoption by other research groups.