[Paper Review] ORGANA: A Robotic Assistant for Automated Chemistry Experimentation and Characterization

Kourosh Darvish, Marta Skreta+10 more|arXiv (Cornell University)|Jan 13, 2024

Innovative Microfluidic and Catalytic Techniques Innovation20 citations

TL;DR

Organa is a modular robotic system that automates diverse chemistry experiments using LLMs for natural-language interaction, parallel task planning, and automated reporting, demonstrated on solubility, pH, recrystallization, and electrochemistry (including electrode polishing in automation).

ABSTRACT

Chemistry experiments can be resource- and labor-intensive, often requiring manual tasks like polishing electrodes in electrochemistry. Traditional lab automation infrastructure faces challenges adapting to new experiments. To address this, we introduce ORGANA, an assistive robotic system that automates diverse chemistry experiments using decision-making and perception tools. It makes decisions with chemists in the loop to control robots and lab devices. ORGANA interacts with chemists using Large Language Models (LLMs) to derive experiment goals, handle disambiguation, and provide experiment logs. ORGANA plans and executes complex tasks with visual feedback, while supporting scheduling and parallel task execution. We demonstrate ORGANA's capabilities in solubility, pH measurement, recrystallization, and electrochemistry experiments. In electrochemistry, it executes a 19-step plan in parallel to characterize quinone derivatives for flow batteries. Our user study shows ORGANA reduces frustration and physical demand by over 50%, with users saving an average of 80.3% of their time when using it.

Motivation & Objective

• Enable flexible, user-friendly automation of diverse chemistry experiments in self-driving labs (SDLs).
• Integrate LLM-based reasoning to translate chemist intent into executable plans and provide feedback.
• Develop parallel task and motion planning with scheduling to improve throughput and resource use.
• Incorporate perception to ground scene objects and monitor experiment progress.
• Automate data analysis and generate comprehensive experiment reports.

Proposed method

• LLM-driven Reasoner converts chemist intent into chemistry task descriptions and structured XDL plans via CLAIRify.
• Temporal-PDDLStream solver enables simultaneous task, motion planning and scheduling with time-varying costs and durative actions.
• Perception pipeline combines GroundDINO, SAM, and depth data to detect and estimate poses of transparent/opaque objects and monitor progress (e.g., turbidity via vision).
• Robotics skills include pick/place, constrained motion planning (PRM* with constraint projections), and a pouring mechanism with PD-based control.
• Automated data analysis estimates electrochemistry parameters with MLE/posterior updates and generates PDF reports summarizing results and uncertainties.

Experimental results

Research questions

• RQ1 Can Organa flexibly automate a diverse set of chemistry experiments with minimal custom engineering for new tasks?
• RQ2 How can LLMs be used to translate human intent into executable, planable chemistry experiments while handling ambiguities?
• RQ3 To what extent can temporal task and motion planning with scheduling improve throughput in automated chemistry labs?
• RQ4 How effective is automated perception in grounding objects and monitoring experiments, including transparent objects?
• RQ5 What is the impact of Organa on user experience and perceived workload in chemist usability studies.

Key findings

• Organa demonstrates autonomous solubility, recrystallization, pH, and electrochemistry experiments with parallel task execution.
• Solubility tests achieved accuracy values of 7.2%, 11.2%, and 12.3% for salt, sugar, and alum, respectively, with an average plan duration of 25.63 minutes for a 7-step plan.
• Recrystallization used pre-heating solvent, executing an 8-step plan with a total time of 44.80 minutes.
• pH experiments used onion-red cabbage indicator with a 6-step plan totaling 3.85 minutes.
• In electrochemistry, Organa automated polishing of a glassy carbon electrode and conducted 3 cycles of cyclic voltammetry across multiple pH conditions to characterize quinone derivatives.
• User study indicates Organa significantly improves user experience and reduces physical workload.