7 Jan 2025 | Kourosh Darvish, Marta Skreta, Yuchi Zhao, Naruki Yoshikawa, Sagnik Som, Miroslav Bogdanovic, Yang Cao, Han Hao, Haoping Xu, Alán Aspuru-Guzik, Animesh Garg, Florian Shkurti
ORGANA is an assistive robotic system designed to automate various chemistry experiments. It leverages Large Language Models (LLMs) to interact with chemists, understand their goals, and plan and execute complex tasks. ORGANA's key strengths include its ability to perceive the surrounding environment, track task progress, and make informed decisions to optimize experiment efficiency. It supports parallel task execution, provides timely feedback through reports and analysis, and reduces user frustration and physical demand by over 50%. ORGANA has been demonstrated in experiments such as solubility screening, recrystallization, pH measurement, and electrochemistry, showing significant improvements in efficiency and user satisfaction. The system's modular design allows for adaptability to different experiments and hardware setups, and its safety features ensure reliable and robust operation. However, limitations include the need for human intervention in certain scenarios and the complexity of defining PDDL domains for planning robot actions. Overall, ORGANA represents a significant step towards achieving flexible and efficient automation in chemistry labs.ORGANA is an assistive robotic system designed to automate various chemistry experiments. It leverages Large Language Models (LLMs) to interact with chemists, understand their goals, and plan and execute complex tasks. ORGANA's key strengths include its ability to perceive the surrounding environment, track task progress, and make informed decisions to optimize experiment efficiency. It supports parallel task execution, provides timely feedback through reports and analysis, and reduces user frustration and physical demand by over 50%. ORGANA has been demonstrated in experiments such as solubility screening, recrystallization, pH measurement, and electrochemistry, showing significant improvements in efficiency and user satisfaction. The system's modular design allows for adaptability to different experiments and hardware setups, and its safety features ensure reliable and robust operation. However, limitations include the need for human intervention in certain scenarios and the complexity of defining PDDL domains for planning robot actions. Overall, ORGANA represents a significant step towards achieving flexible and efficient automation in chemistry labs.