ERC grant to teach robots a softer touch
Enabling robots to handle soft and deformable objects is one of the major challenges in robotics research. Now, Noémie Jaquier, Assistant Professor in Robot Learning and Control, has been awarded an ERC Starting Grant for a project that takes a new approach to the problem.
Grasping, moving and shaping soft and changing objects such as cables, clothes, bags and food is an everyday task for us. We do it without thinking, thanks to the brain and body’s ability to quickly perceive an object’s shape, plan a movement and adjust the grip.
“For robots, however, this is extremely difficult. Being able to grasp and control objects that can change shape safely and reliably is one of the major challenges in robotics,” says Noémie Jaquier .
Now, researchers in the GEOMANIP project, led by Noémie Jaquier, will develop a new way for robots to understand and handle soft and deformable objects. The approach takes into account both the geometric properties of objects and how they behave according to the laws of physics. The researchers will combine knowledge of how materials and objects change with modern AI and deep learning.
Work more accurately
“This will give robots a better ability to perceive, plan and control their movements – and make them more autonomous when picking up, moving and working with soft and changing objects.”
The goal is to develop a general method for robotic manipulation that works with many different types of objects, enabling robots to work more accurately and reliably.
“With this research, we want to take a step towards robots that can work with deformable objects as naturally as humans do.”
Bring ideas to life
The ERC funding, awarded by the European Research Council, amounts to EUR 1.5 million and runs for five years.
“I feel incredibly grateful for the opportunity to bring the ideas behind this project to life. The funding gives me the freedom, resources and time to pursue the research I truly believe in. I can build a team and explore a completely new way for robots to understand and handle deformable objects.”
Text: Christer Gummeson ( gummeson@kth.se )