Crop protection products are essential for improving crop yields to sustain a growing population. However, formulation design and development remain primarily empirical, often requiring significant resource to design a formulation product. While principal rules and empirical methods such as the hydrophilic-lipophilic balance (HLB) theory, are effective in control experiments, they often struggle to investigate optimal candidate materials for industrial applications. Hence, the development of agrochemical formulation heavily relies on an extensive number of experiments. To accelerate the formulation design process, we will present our level-4 self-driving laboratory (SDL) system for formulation development, powered by machine learning models based on past experimental data. As a level-4 SDL, chemists will design the search space and the SDL will conduct closed-loop experiments which are formed by multiple tasks. This system autonomously designs, conducts, characterises, and refines experiments iteratively. By significantly reducing the workload of formulation scientists, our SDL enhances efficiency and accelerates innovation in crop protection formulation.
