Robot manipulation in human environments: Challenges for learning algorithms

Conference Article


Dagstuhl Seminar (DROPS)





Doc link


Download the digital copy of the doc pdf document


Manipulator robots are widening their range of activities in factories, as well as finding increased application in human-centered domains such as healthcare, education, entertainment and services. For robots to become handy co-workers and helpful assistants, quick and user-friendly ways to endow them with flexible manipulation skills are needed. At the Perception and Manipulation Lab of IRI (CSIC-UPC), we are addressing several of the learning challenges arising in this context. Namely, manipulator robots should be easy to teach by non-experts and acquire skills from demonstrations, they need to be intrinsically safe, able to appropriately deal with forces and to perceive and manipulate deformable objects, tolerant to noisy perceptions and inaccurate actions, and they must exhibit a high adaptability to non-predefined and dynamic environments, as well as the capability of learning to plan.


intelligent robots, learning (artificial intelligence), manipulators, service robots.

Author keywords

robot learning, manipulation of deformable objects

Scientific reference

C. Torras. Robot manipulation in human environments: Challenges for learning algorithms, 14081 Dagstuhl Seminar, 2014, Dagstuhl, Germany, in Robots Learning from Experiences, Vol 4-2 of Dagstuhl Reports, pp. 99-101, Schloss Dagstuhl.