BieColl - Bielefeld eCollections
Not a member yet
1004 research outputs found
Sort by
Edge and plane classification with a biomimetic iCub fingertip sensor
Get PDFThe exploration and interaction of humanoid robots with the environment through tactile sensing is an important task for achieving truly autonomous agents. Recently much research has been focused on the development of new technologies for tactile sensors and new methods for tactile exploration. Edge detection is one of the tasks required in robots and humanoids to explore and recognise objects. In this work we propose a method for edge and plane classification with a biomimetic iCub fingertip using a probabilistic approach. The iCub fingertip mounted on an xy-table robot is able to tap and collect the data from the surface and edge of a plastic wall. Using a maximum likelihood classifier the xy-table knows when the iCub fingertip has reached the edge of the object. The study presented here is also biologically inspired by the tactile exploration performed in animals
Integrated Model for Generating Non Verbal Body Behavior Based on Psycholinguistic Analysis in Human-Robot Interaction
Get PDF
From Penguins to Parakeets: a Developmental Approach to Modelling Conceptual Prototypes
Get PDFThe use of concepts is a fundamental capacity underlying complex, human-level cognition. A number of theories have explored the means of concept representation and their links to lower-level features, with one notable example being the Conceptual Spaces theory. While these provide an account for such essential functional processes as prototypes and typicality, it is not entirely clear how these aspects of human cognition can arise in a system undergoing continuous development - postulated to be a necessity from the developmental systems perspective. This paper seeks to establish the foundation of an approach to this question by showing that a distributed, associative and continuous development mechanism, founded on principles of biological memory, can achieve classification performance comparable to the Conceptual Spaces model. We show how qualitatively similar prototypes are formed by both systems when exposed to the same dataset, which illustrates how both models can account for the development of conceptual primitives
Towards Modular Spatio-temporal Perception for Task-adapting Robots
Get PDFIn perception systems for object recognition, the advantage of multiple modalities, of combining approaches, and several views is emphasized, as they improve accuracy. However, there are great variances in the implementation, suggesting that there is no consensus yet on how to approach this problem. Nonetheless, we can identify some common features of the methods and propose a flexible system where existing and future approaches can be tested, compared and combined. We present a modular system in which perception routines can be easily added, and define the logic of making them work together based on the lessons learned from different experiments
