Sylvain Calinon
Country: Switzerland
Affiliations:
Senior Research Scientist at the Idiap Research Institute
Lecturer at the Ecole Polytechnique Fédérale de Lausanne (EPFL)
Biography:
Dr Sylvain Calinon is a Senior Research Scientist at the Idiap Research Institute and a Lecturer at the Ecole Polytechnique Fédérale de Lausanne (EPFL). He heads the Robot Learning & Interaction group at Idiap, with expertise in human-robot collaboration, robot learning from demonstration, geometric representations and optimal control. The approaches developed in his group can be applied to a wide range of applications requiring manipulation skills, with robots that are either close to us (assistive and industrial robots), parts of us (prosthetics and exoskeletons), or far away from us (shared control and teleoperation).
Websites:
Movement Generation and Drawing in Robotics
Despite significant advances in AI, robots still struggle with tasks involving physical interaction. Robots can easily beat humans at board games such as Chess or Go but are incapable of skillfully moving the game pieces by themselves (the part of the task that humans subconsciously succeed in). What makes research in robotics both hard and fascinating is that movement skills are tightly connected to our physical world and to embodied forms of intelligence.
I will present an overview of the research in our group to help robots acquire manipulation skills by imitation and self-refinement. We advocate frugal learning in our research, where frugality has two goals: 1) learning manipulation skills from only few demonstrations or exploration trials; and 2) learning only the components of the skill that really need to be learned!
Toward this goal, I will emphasize the roles of geometric manifolds, manipulability ellipsoids, implicit shape representations and distance fields as inductive biases to facilitate manipulation skill acquisition. For the generation of trajectories and feedback controllers, I will discuss how the underlying cost functions should take into account variations, coordination and task prioritization, where various forms of movement primitives based on Fourier or Bernstein functions can contribute to the optimization process. I will also show how ergodic control can provide a mathematical framework to generate exploration and coverage movement behaviors, which we exploit in robot drawing applications and as a way to cope with uncertainty in sensing, proprioception and motor control.