The dream of having robots living among us is coming true thanks to the recent advances in Artificial Intelligence (AI). The gap that still exists between that dream and reality will be filled by scientific research, but manifold challenges are yet to be addressed. Handling the complexity and uncertainty of real-world scenarios is still the major challenge in robotics nowadays. In this respect, novel AI methods are giving the robots the capability to learn from experience and therefore to cope with real-life situations. Moreover, we live in a physical world in which physical interactions are both vital and natural. Thus, those robots that are being developed to live among humans must perform tasks that require physical interactions. Haptic perception, conceived as the idea of feeling and processing tactile and kinesthetic sensations, is essential for making this physical interaction possible. This research is inspired by the dream of having robots among us, and therefore, addresses the challenge of developing robots with haptic perception capabilities that can operate in real-world scenarios. This PhD thesis tackles the problems related to physical robot interaction by employing machine learning techniques. Three AI solutions are proposed for different physical robot interaction challenges: i) Grasping and manipulation of humans’ limbs; ii) Tactile object recognition; iii) Control of Variable-Stiffness-Link (VSL) manipulators. The ideas behind this research work have potential robotic applications such as search and rescue, healthcare or rehabilitation. This dissertation consists of a compendium of publications comprising as the main body a compilation of previously published scientific articles. The baseline of this research is composed of a total of five papers published in prestigious peer-reviewed scientific journals and international robotics conferences.
Buscar en Dimension