ARChemist: Autonomous Robotic Chemistry System Architecture

Authors: Hatem Fakhruldeen, Gabriella Pizzuto, Jakub Glowacki, Andrew Ian Cooper · ICRA 2022

Licences

Open Access
Paper: CC BY 4.0

Everything below draws on the full paper under its CC BY 4.0 licence; its figure is credited to the authors.

TL;DR

ARChemist links a KUKA KMR mobile manipulator, a Franka Panda arm and standard lab instruments through ROS and YAML recipes; solubility screening succeeded in 9 of 10 runs, crystallisation 80% of 5.

Why it matters

The system runs heterogeneous commercial robots in one chemistry workflow, a mobile manipulator for transport and a fixed arm for vial handling, together with off-the-shelf instruments instead of bespoke hardware. Using ROS as middleware lets new robots and instruments be added without changing the core processing logic.

Problem

Most robotic systems in chemistry labs are programmed for a single experiment, often on proprietary architectures or custom hardware such as the ChemPU, and earlier systems do not integrate mobile robots to carry samples to stations out of a fixed robot's reach. Chemists, who are generally not robotics experts, need a system that is easy to program, simplifies data analysis and generalises across experiments.

Proposed solution

ARChemist is a Python architecture with a State Manager (materials, robots, stations, samples and their histories), a Persistence Manager (database storage and parsing of configuration and recipe files) and a Workflow Manager (workflow processor, robot scheduler, parallel station and robot handlers, system monitor and rule-based alert manager). ROS is the communication layer to robots and instruments, with custom drivers written for the lab equipment. Each sample carries a human-readable YAML chemical recipe listing materials, station operations and a state-machine workflow, and a new station or robot is added by supplying its class file and handler, which the system discovers at start-up.

Key insights

Separating process-centred state (materials, stations, robots and their operations) from sample-centred state (recipe, operation history, location, status) lets one installation run different recipes, provided their stations and operations are in the configuration file.
Loose coupling through Python dynamic typing and type introspection means adding a station or robot needs only its class file and handler, with no changes to other modules.
The mobile manipulator carries samples between stations in different parts of the lab, and the authors note it could work in parallel with the fixed arm on other stations.
Robotic runs are slower than a human chemist, but the authors aim to free chemists from repetitive tasks and to run experiments over long periods without stoppage.

Main results

Solubility screening: 10 runs, 90% success rate, average run time 12 minutes; the one failure was the Quantos timing out while taring because of external vibrations.
Crystallisation: 5 runs, 80% success rate, average run time 130 minutes (2 hours 10 minutes); the failure was the Panda misplacing the vial when moving it to the scale.
Solubility workflow: 15 mg of salt dispensed by the Quantos, 2 mL of water by the peristaltic pump, stirring for 1 second at 200 rpm, and RGB images from an Intel RealSense D435i used to compute turbidity and detect dissolution.
Crystallisation workflow: 200 mg of sodium chloride and 2 mL of water, heated at 60 °C with continuous mass monitoring until the mass stabilises.

Limitations

The Workflow User Interface is currently a command-line interface; the authors plan a graphical one.
Only two robots and a small set of instruments (solid and liquid dispensers, balance, hot plate and stirrer) were integrated, and each instrument needed a custom ROS driver.
For safety, the end-to-end system was tested on an open bench with water as the only solvent.
The evaluation is small, 10 solubility and 5 crystallisation runs, with one failure in each.
Machine-learning decision making is not yet integrated; the authors plan to add it.

Field impact

The paper was a finalist for the ICRA 2022 Outstanding Automation Paper award and has 53 citations. It links a video of both case studies (https://youtu.be/mcEYyOBjKpU); the paper does not give a code repository.

Abstract

Automated laboratory experiments have the potential to propel new discoveries, while increasing reproducibility and improving scientists' safety when handling dangerous materials. However, many automated laboratory workflows have not fully leveraged the remarkable advancements in robotics and digital lab equipment. As a result, most robotic systems used in the labs are programmed specifically for a single experiment, often relying on proprietary architectures or using unconventional hardware. In this work, we tackle this problem by proposing a novel robotic system architecture specifically designed with and for chemists, which allows the scientist to easily reconfigure their setup for new experiments. Specifically, the system's strength is its ability to combine together heterogeneous robotic platforms with standard laboratory equipment to create different experimental setups. Finally, we show how the architecture can be used for specific laboratory experiments through case studies such as solubility screening and crystallisation.

Open Source Robotics