University of Victoria

Published purpose

Automated chemistry for material design has been a priority for research groups and companies per the requirement of increased productivity, reliable results, safer operation conditions, cost savings, and flexible working time. Due to their features of dexterity and flexibility, robot manipulators (RMs) are widely used for chemistry laboratory automation. However, the safe and high-precision control of RMs for pouring and transmitting reagents and catalysts in synthetic and reaction experiments is still challenging. To fill this gap, this project aims to propose a solution to the control of an RM with guaranteed safety, reliability, and stability for achieving autonomous grasping operation. Specifically, the objective of this work is to develop an intelligent framework based on reinforcement learning (RL), model predictive control (MPC), and visual serving techniques. The proposed RL-based MPC (RLMPC) will improve the control performance of the RM and thus further improve the level of automation in chemistry laboratories.