Samuli Hynninen

Manipulation of fluids and granular materials is vital for a variety of domains in robotics and autonomous work machines. For example, robots performing household chores, such as cooking, waitressing, or assisted feeding, need this skill to flexibly operate with a range of food ingredients and beverages, from flours to hot coffee. Furthermore, granular materials and fluids play an essential role in industries such as mining and construction. In these environments, the possibility of safely and reliably manipulating granular materials creates lucrative opportunities to automate tasks that currently require a human machine operator.

The goal of our research is to develop solutions for the identification and property estimation of granular materials and fluids, as well as for the planning and control of the actual manipulation tasks. We utilize a variety of methods, including but not limited to modern machine learning approaches like neural networks, classical machine learning approaches like support vector machines, and optimization and optimal control. We use different material dynamics models ranging from simplistic equivalent mechanical models, such as the spherical pendulum model for liquid sloshing, to more elaborate methods like the material point method to model the behaviour of the materials under manipulation.

Our research aims for new openings to intelligently automate tasks that have previously required skilled humans. Our research has the potential to yield several benefits, ranging from pure economic benefits, such as decreased human labour costs, to soft benefits, such as increased feelings of independence for people with disabilities who will receive more advanced assistive technology.