StoRMS – Stochastic Robotized Micro-Sampling
Collaborations: HEIG-VD, DEC Engineering, Chemspeed
In automated high-throughput chemistry, dispensing small quantities of solids (sub-milligram to milligram) remains a major bottleneck. Conventional systems rely on repeated collect-and-dispense cycles with feedback from a balance, which is slow for large experiments, prone to cross-contamination and wasteful of material. Given the wide diversity of solid properties, no universal solution exists yet.
To address this, we developed STORMS (STOchastic Robotic MicroSampling), based on two ideas: stochastic sampling and “precise rather than exact” masses. Instead of targeting a fixed mass, a vacuum-assisted coring tool rapidly collects solid samples around the relevant experimental masses, each one precisely weighed, recorded in our database and laser-sealed in a small glass capsule for inert storage. When an experiment is set up, a matching algorithm selects the most suitable capsules for each reaction and adjusts the reaction scale and liquid volumes to preserve the desired molecular ratios. A robot then crushes the capsules directly above the reaction vials. This approach decouples solid sampling from reaction preparation, reducing the time needed to prepare a 96-well plate and enabling continuous 24-hour closed-loop experimentation.
STORMS was developed by Catalysis West Hub – Swiss Cat+ together with Dietrich Engineering Consulting SA (DEC Group, Ecublens), HEIG-VD and Chemspeed Technologies, as part of an Innosuisse project. The project is funded by Innosuisse, with a budget of 786’450 CHF over 3 years, https://www.aramis.admin.ch/Grunddaten/?ProjectID=51768, and is already protected by an European Patent : 22215745.5EP – F. de Nanteuil, E. Marianno, M. Kunz, P.Miéville EPFL-HEIG-VD – Substance preparation method and system.
High-throughput solid microsampling through stochastic robotic automation
Transferring solids with a spatula is a routine task for chemists, yet it remains one of the main bottlenecks in fully automating a chemistry laboratory, as existing solid dispensers only handle a limited range of materials. In collaboration with the Idiap Research Institute, we developed ARMS (Automated Robotized human-Mimicking Solid sampling system), a bimanual robotic workbench inspired by how chemists actually work. An observational study of chemists revealed a simple four-step workflow (characterize the material, scoop, deposit and weigh), along with a set of visual and tactile descriptors that guide their choice of technique. ARMS reproduces this workflow with two robotic arms: one holds a standard laboratory spatula equipped with a force sensor, while the other handles standard vials and moves them between a balance and a cleaning station. The robot learns each scooping and depositing action from about twenty human demonstrations, collected in under an hour and trained in less than a minute, and adapts its movements to each new vial and fill level. ARMS successfully sampled seven of nine test materials, from free-flowing solids such as salt and sugar (with a precision of 1.5–3.3 mg per scoop) to sticky and viscous materials such as honey and grease, paving the way for a general-purpose solid sampler in self-driving laboratories.
ARMS: an Automated Robotized human-Mimicking Solid sampling system