A robotic model of the Ecological Self

This paper discusses an integrated model of a robot's sensory and perceptual capabilities based on one of the earliest forms of self-knowledge that humans develop, knowledge of the Ecological Self. The Ecological Self is a cohesive model of the body and senses learned through the experience of using them together. This unified model allows kinematic and sensory data to be combined, producing an intersensory perception grounded in both inputs. Taking inspiration from this Ecological Self, but building on modern engineering practices, this model allows a robot to learn the kinematics of its end-effector by witnessing its motion in its visual field. This property of adaptation through self-observation also allows the model to adapt to changes in the robot's kinematic structure, as in the case of tool use. A final refinement is performed over the combined visual-kinematic model and is demonstrated to improve not only the accuracy of the kinematic model, but also the robot's stereo vision calibration. This refinement is inspired by the hypothetical process by which infants learn about their selves. The system is demonstrated to require fewer than 200 motion samples to fully train, to predict end-effector position within 2.29mm (SD=0.10) and 2.93 pixels (SD=3.83), to learn the lengths of the linkages in the robot's arm to within 1.1mm, and to adapt to tool use after only 52 samples.

[1]  Susan J. Hespos,et al.  Differential rooting response by neonates: Evidence for an early sense of self , 1997 .

[2]  Craig M. Shakarji,et al.  Least-Squares Fitting Algorithms of the NIST Algorithm Testing System , 1998, Journal of research of the National Institute of Standards and Technology.

[3]  Andrew Zisserman,et al.  Multiple View Geometry in Computer Vision (2nd ed) , 2003 .

[4]  Zhengyou Zhang,et al.  A Flexible New Technique for Camera Calibration , 2000, IEEE Trans. Pattern Anal. Mach. Intell..

[5]  Alejandro Hernández Arieta,et al.  Body Schema in Robotics: A Review , 2010, IEEE Transactions on Autonomous Mental Development.

[6]  Alexander Stoytchev.  Toward Video-Guided Robot Behaviors , 2007 .

[7]  Minoru Asada,et al.  Is it my body? Body extraction from uninterpreted sensory data based on the invariance of multiple sensory attributes , 2004, 2004 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (IEEE Cat. No.04CH37566).

[8]  P. Rochat.  The infant's world , 2001 .

[9]  John M. Hollerbach,et al.  The Calibration Index and Taxonomy for Robot Kinematic Calibration Methods , 1996, Int. J. Robotics Res..

[10]  Manuel Lopes,et al.  Body schema acquisition through active learning , 2010, 2010 IEEE International Conference on Robotics and Automation.

[11]  Kurt Konolige,et al.  Calibrating a Multi-arm Multi-sensor Robot: A Bundle Adjustment Approach , 2010, ISER.

[12]  Minoru Asada,et al.  Unique association between self-occlusion and double-touching towards binding vision and touch , 2007, Neurocomputing.

[13]  P. Rochat,et al.  Oropharyngeal Control of Hand-Mouth Coordination in Newborn Infants , 1988 .

[14]  J. Denavit,et al.  A kinematic notation for lower pair mechanisms based on matrices , 1955 .

[15]  Bernhard P. Wrobel,et al.  Multiple View Geometry in Computer Vision , 2001 .

[16]  Aude Billard,et al.  Online Learning of the Body Schema , 2008, Int. J. Humanoid Robotics.

[17]  Wolfram Burgard,et al.  Body schema learning for robotic manipulators from visual self-perception , 2009, Journal of Physiology - Paris.