Making an Accessible Automation Tool Autonomous

Automated and autonomous experimentation is most effective when it can be easily adopted by a diverse range of researchers. In this light, the company Opentrons has developed a library of low-cost, user-friendly robotic tools that have recently been leveraged in material science and chemistry. The OT-2 offers an extremely low barrier of entry in robotics accelerated research through facile liquid handling and simple job automation software. However, this platform lacks two critical features necessary for closed-loop studies. First, there are no simple metrology tools for process characterization. The number of compatible metrology techniques in literature are limited on biological applications, and Opentrons does not officially support any developed methods. In the typical workflow, human users are required to transfer prepared samples to an ex-situ monitoring system before data can be collected. This inhibits the rate of data generation, produces imprecise reaction time features, and prevents efficient algorithm driven research. Second, the OT-2 system does not have a simple machine facing API. The instrument is controlled through individual jobs that are uploaded onto the machine itself then executed. This can be done easily by a human using the instrument software or SSH commands; however, for a computer to identify and select experiments, additional steps are necessary. Accessing the API indirectly would enable the OT-2 to be paired with offline computational power and to be more easily integrated into larger closed-loop and automated systems and algorithms.<br/>In this work, we present an open-access and user-friendly strategy for adapting the OT-2 robot for closed-loop, autonomous research. The system features a generalizable sample aliquoting tool which has been combined with absorption spectroscopy, machine-facing control of the OT-2 API, and a single user interface for sampling, liquid handling, and experiment selection algorithm specification. We applied the tool toward the exploration of perovskite solar cell ink compositions in order to gain insight into material combination dynamics through active learning and algorithm-based feature prediction.<br/>The aliquot system operates by collecting samples from the OT-2 through a low volume funnel connected to microfluidic tubing. Samples are pulled from the funnel using a positive displacement pump and sent through an optical flow cell where absorption spectra are collected. While this work leverages absorption spectroscopy, alternative techniques, including destructive and non-destructive characterization methods may be easily included. This tool can be used outside of closed-loop systems to facilitate precise time resolved data collection and more efficient experimentation.<br/>The OT-2 is algorithmically controlled through automated experiment scripting, upload, and execution. With each new set of experimental conditions, the algorithm arranges a set of instructions along with the system features in the language specified by the OT-2 Python API and saves them to a new Python script. This submission is then pushed to the OT-2 server along with the command to execute the script. This method automates a process that otherwise requires a human to manually program the experiment script, upload it to the OT-2, and send the execution command. By algorithmically interacting with the OT-2 through the Opentrons supported API, users can easily leverage the actively developed library of modules without manually implementing updates.<br/>The software and techniques presented in this work will enable users without extensive programming or robotics experience to implement their own closed-loop reaction explorations and optimization. Further development of the modules available for sampling and similar user friendly and low-cost autonomous tools will improve the rate of adoption of autonomous technologies by the broader scientific community.