Project Details
Projekt
Robots Exploring Tools as Extensions to their Body Autonomously
Subject Area
Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
Term
from 2011 to 2018
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 200335523
The body scheme is increasingly appreciated as a core representationto organize interaction and learning of an embodied agent. Whileneuroscientific research is unravelling a remarkable complexity ofbody schemes underlying the actions of biological agents, roboticsstill largely lacks equally sophisticated, adaptive and dynamicallyextensible representations. Associated and largely open challengesare rich representations that marry body morphology, control, and theexploitation of redundant degrees of freedom in representations thatoffer strong priors for rapid learning that in turn support a flexible adaptation and extension of these representations torealize capabilities such as tool use or graceful degradation incase of malfunction of parts of the body tree. This motivates the present project: to develop,implement and evaluate algorithms enabling a robot toautonomously learn various rich extensions of its static (core) bodyschema} in order to support flexible tool use and a coping withsituations that require to anticipate the reaction of novellinkages between the body and arrangements of physical objects(e.g., when pulling a middle piece out of a pile, possiblyusing an available part of the body to stabilize otherparts of the pile).As a major scientific contribution to Autonomous Learning weaim to develop the current, largely kinematics focused body schema representations of robotics into \textit{stronger priorsfor learning and exploration} by creating a rich bodyrepresentation that addresses three key aspects of interactionlearning: 1) enhancing the scope from the body morphology to a representation of body-environment linkages, 2) enhancing the scope from a representation of morphology to a representation that includes control, and 3) enhancing the scope from minimal DOF systems to systems that offer redundant degrees of freedomWe will develop these representations in the context ofa systematically chosen "matrix" of real-world interactionsituations, arranged to pose learning challenges in increasingorder of complexity along the above three dimensions.Thereby, we will build on the previous project, where wehave developed a basic framework for adaptive body schemesemphasising the kinematics level.The project will thus directly contribute to enhance theautonomy of robots for adjusting their physical interactionwith the world to the variety of situations that is characteristic ofmany natural environments. It will alsoadvance the state of the art of representations that cansupport such capabilities, including representations thatcan autonomously extend themselves as a result ofautonomous exploration in robots acting in the real world.
DFG Programme
Priority Programmes
Subproject of
SPP 1527:
Autonomous Learning
Participating Person
Dr. Robert Haschke
