The Governors of the University of Alberta

Published purpose

Automation in chemistry labs expedites experimentation and minimizes errors, facilitating more cost-effective material and drug discovery. The project aims to automate chemistry labs by developing transferable, adaptable, and generalizable robotic systems using Reinforcement Learning (RL). The project is designed to tackle the challenges of integrating robotic arms into chemistry labs, focusing on minimizing data requirements and enhancing real-world applications of RL. Key objectives include developing a model-based RL method that generalizes across different robotic platforms, creating robust and transferable RL agents, and ensuring the real-time adaptability of robots to evolving lab tasks. The project addresses the unique challenge of manipulating transparent objects and aims to implement high frequency control for agile and responsive robot behavior. By extending the Remote-Local Distributed (ReLoD) system to model-based RL, the project aims to refine the real-time learning and control of lab robots, setting the stage for widespread adoption of automated, efficient, and intelligent lab systems.