Large language models for chemistry robotics

Abstract

This paper proposes an approach to automate chemistry experiments using robots by translating natural language instructions into robot-executable plans, using large language models together with task and motion planning. Adding natural language interfaces to autonomous chemistry experiment systems lowers the barrier to using complicated robotics systems and increases utility for non-expert users, but translating natural language experiment descriptions from users into low-level robotics languages is nontrivial. Furthermore, while recent advances have used large language models to generate task plans, reliably executing those plans in the real world by an embodied agent remains challenging. To enable autonomous chemistry experiments and alleviate the workload of chemists, robots must interpret natural language commands, perceive the workspace, autonomously plan multi-step actions and motions, consider safety precautions, and interact with various laboratory equipment. Our approach, CLAIRify, combines automatic iterative prompting with program verification to ensure syntactically valid programs in a data-scarce domain-specific language that incorporates environmental constraints. The generated plan is executed through solving a constrained task and motion planning problem using PDDLStream solvers to prevent spillages of liquids as well as collisions in chemistry labs. We demonstrate the effectiveness of our approach in planning chemistry experiments, with plans successfully executed on a real robot using a repertoire of robot skills and lab tools. Specifically, we showcase the utility of our framework in pouring skills for various materials and two fundamental chemical experiments for materials synthesis: solubility and recrystallization. Further details about CLAIRify can be found at https://ac-rad.github.io/clairify/.

Toward Making Polymer Chemistry Autonomous

“…However, the translation of natural language experiment descriptions into low-level robotics languages presents significant challenges. 54 Recent advancements have utilized large language models to generate task plans, but reliably executing these plans in the real world using embodied agents remains a considerable hurdle. To enable autonomous chemistry experiments and reduce the workload of chemists, robots must interpret natural language commands, perceive the workspace, autonomously plan multistep actions and motions, consider safety precautions, and interact with various laboratory equipment.…”

“…The integration of natural language interfaces into autonomous chemistry experiment systems aims to simplify the use of complex robotics systems, making them more accessible to nonexpert users. However, the translation of natural language experiment descriptions into low-level robotics languages presents significant challenges …”

The Future of Material Scientists in an Age of Artificial Intelligence

“…Progress in this area is already underway, exemplified by initiatives from companies like Hugging Face which have showed that it is possible to use LLMs as a controller to manage existing AI models to solve sophisticated AI tasks in different modalities and domains. [8] Examples in material science where LLMs are connected to robotic experimentation are still few but include recrystallization experiments, [107] successful performance of catalysed cross-coupling reactions, [108] and synthesis of humidity colorimetric sensors. [109] The reports currently available on this topic has the character of initial proofof-concepts but provide an indication of where we may be heading.…”

Density Functional Theory-Guided Synthesis of Cu-N-TiO2 for Overall Water Splitting by Breaking the Scaling Relationship

“…Another way to screen target materials is through high-throughput experiments, which deepen the understanding of experimental work and improve catalysts through automated experiments. 17 However, this approach is challenging to implement for scholars lacking automated laboratories, as any successful step in synthesis and testing takes at least 1−2 days, and evidently, a large number of experiments are impractical for manual operation. Therefore, we aim to rely solely on DFT, conducting accurate and rapid screening work based on established research foundations.…”

“…The main reason is that machine learning requires a significant accumulation of early stage data, and the total time consumption may even exceed that of high-throughput experiments. , Its lack of specificity has long been pointed out by scholars. Another way to screen target materials is through high-throughput experiments, which deepen the understanding of experimental work and improve catalysts through automated experiments . However, this approach is challenging to implement for scholars lacking automated laboratories, as any successful step in synthesis and testing takes at least 1–2 days, and evidently, a large number of experiments are impractical for manual operation.…”