Adaptive Object Classification: a Mobile Robot Case Study. Internal Report 9607

Gillies, E. (1996) Adaptive Object Classification: a Mobile Robot Case Study. Internal Report 9607. Technical Report. Department of Aerospace Engineering, University of Glasgow.

Abstract

An important task for any autonomous agent is to classify any objects that it encounters in its environment. A particular type of object may be useful to the agent, for example, or alternatively, the positions of recognizable objects may be used as local landmarks in a cognitive map of the environment. Biological organisms are adept at classifying new or familiar objects. It is apparent that biological organisms often have self organizing methods of classification and that they specialize in recognizing particular groups of objects. Biological organisms also often incorporate information from their own movements; rather than relying solely on sense information alone. In this paper, these three observations inspire an adaptive scheme in which a mobile robot learns to recognize examples from a group of complex objects. The autonomous agent moves around each object to determine the object shape. The important features of the shape are then extracted by a self-organizing neural layer. Each object is then represented by a small number of feature amplitudes making object classification more simple. The paper describes the scheme applied to a small skid-steer mobile robot equipped with infra-red proximity sensors.

• Item Type: Research Reports or Papers (Technical Report)
• Status: Published
• Glasgow Author(s) Enlighten ID: Gillies, Dr Eric
• Authors: Gillies, E.
• College/School: College of Science and Engineering > School of Engineering > Autonomous Systems and Connectivity
• Publisher: Department of Aerospace Engineering, University of Glasgow
• Copyright Holders: Copyright © 1996 Department of Aerospace Engineering, University of Glasgow
• Publisher Policy: Reproduced with the permission of the Department