Robots are becoming increasingly independent, capable of operating effectively in structured environments — from grasping household objects to opening doors and drawers. While tasks like picking up rigid items or interacting with appliances have seen impressive progress, far less attention has been given to fine manipulationskills: tasks that demand precise, steady, and delicate control. Examples include threading a needle, controlling and moving liquids, inserting delicate components, or manipulating soft tissue — activities that require not just coordination, but careful handling of small forces, complex contacts, and dynamic surfaces. Developing such capabilities is especially challenging, as these skills are both difficult to collect in real-world demonstrations and difficult to simulate due to the intricate interactions, unpredictable collisions, and rich physical complexity of real environments.

In this talk, I will present how robots can acquire fine manipulation skills by learning the right representations for observation and control. I will show how advances in pseudo-physical modeling, differentiable rendering, and perception-driven policy learning — together with the standardization of robotic platforms — are closing the gap between vision foundation models and the demands of precise, autonomous manipulation. Through examples ranging from household tasks to autonomous surgery, I will highlight key challenges, describe emerging solutions, and offer my perspective on what it will take to enable robots to perform highly skilled, sensitive tasks in dynamic and unstructured environments.

Michael Yip is an Associate Professor of Electrical and Computer Engineering at UC San Diego, IEEE RAS Distinguished Lecturer, Hellman Fellow,  Senior Member of the IEEE and the National Academy of Inventors, and Director of the Advanced Robotics and Controls Laboratory (ARCLab). 

His group currently focuses on surgical robots, biomimetic design, and robot learning. His research has received several best paper awards and nominations at ICRA and IROS and RA-L, and he has been recognized by the NSF CAREER award and the NIH Trailblazer award. Several of his research has led to founded startups, and he was named the Faculty Innovator of the Year at UCSD in 2024.

Dr. Yip was previously a Disney Research in 2014 and Amazon Robotics in 2018. He received a B.Sc. in Mechatronics Engineering from the University of Waterloo, an M.S. in Electrical Engineering from the University of British Columbia, and a Ph.D. in Bioengineering from Stanford University.