Paper Info
Title
An infant-inspired model for robot developing its reaching ability.
Abstract
In robotics community, inspiration from human development theory offers a promising way for robots to efficiently achieve many abilities. Humanoid robot reaching, as an essential ability which forms the foundation of grasping and manipulation, is also possible to take advantages from human. For human infants, reaching is emerging around 3-5 months of age, which is also a fundamental skill for the development and refinement of future high level motion or cognitive behaviors. The issue of how infant reaching is developed has been investigated for decades and several views were established and discussed. Recently, Corbetta et al. proposed that the emergence of reaching is the product of a deeply embodied process, in which infants first learn how to direct their movement in space using proprioceptive and haptic feedback and then map the visual attention onto these bodily centered experiences. With this new sight, this paper addresses the problem of how a robot develops its reaching ability autonomously, and a novel infant-inspired model is proposed. The model is composed of several blocks so that to closely capture the inherent mechanisms of infants in the process of developing the reaching skill. To evaluate the proposed model, a child-sized physical humanoid robot PKU-HR6.0 is employed. Just like an early infant, the robot is assumed without any other abilities. It is only equipped with the proposed model with random initialized parameters. The reaching ability is expected to be developed all by the robot itself. Through iteratively babbling arm motions in its workspace, the robot firstly acquire a reliable sense of its body and movement in space, and then map the sense of the vision onto the proprioception when contingencies happen. Experimental results show that the robot equipped with the proposed model achieves the reaching ability effectively and successfully in a completely autonomous style. The robot is able to reach the objects in different real environments with a high performance, and can even grasp the objects with an average successful rate of 82.50%. Our experimental results also verify that the movement of one joint may induce another, and some sepecific joint may play dominant role in certain category of movements.
Year
Venue
Field
2016
Joint IEEE International Conference on Development and Learning and Epigenetic Robotics ICDL-EpiRob
Robot control,Social robot,GRASP,Computer science,Artificial intelligence,Robot,Haptic technology,Robotics,Mobile robot,Humanoid robot
DocType
ISSN
Citations 
Conference
2161-9484
0
PageRank 
References 
Authors
0.34
0
5
Name
Order
Citations
PageRank
Dingsheng Luo14611.61
Fan Hu222.08
Yian Deng301.35
Wentao Liu453.46
Xihong Wu527953.02