Experiment manager software for an automated chemistry workstation, including a scheduler for parallel experimentation

Abstract

An automation system with the ability to work relentlessly, precisely, strategically, adn autonomously in pursuit of scientific goals will provide a powerful tool for tackling difficult scientific problems. We now present our work over the past seven years aimed at developing the hardware and software architecture for such an automated chemistry workstation. The workstation is designed for microscale experimentation in relatively clean domains of synthetic chemistry. This paper has two themes. First, the software system is presented in a comprehensive manner. An experiment manager software package has been developed that provides for composition of experimental plans, controls all aspects of automated experimentation, and manages the data. Experiment manager software is comprised of open-loop and closed-loop experiment planners bundled together with supporting features for timing, scheduling, material databases, resource management, automated start-up procedures, running display, data handling, configuration options, and maintenance operations. The experiment planners provide a menu-driven user interface, editing features, and a modular set of procedural events with which diverse experimental protocols can be composed. An overarching objective has been to achieve versatility in experiment planning yet still maintain access to the power that parallelism can confer in experimentation. Accordingly, the second theme in this paper is a description of approaches for performing experiments in parallel. Parallelism originates chiefly through simultaneous processing of samples at semi-autonomous hardware modules, at the user interface, and through the use of a scheduler. The scheduler takes as input a set of experimental protocols, establishes a sequence of the protocols, then interleaves the experimental protocols without altering the relative times of the procedures within each protocol. Experimental throughput can be increased by up to ten-fold by this approach. The straightforward planning and implementation of concurrent experimental protocols should yield major increases in research productivity.