Developing a robot visual system using a biologically inspired model of neuronal development

✍ Scribed by T. Elliott; N.R. Shadbolt

• Publisher: Elsevier Science
• Year: 2003
• Tongue: English
• Weight: 847 KB
• Volume: 45
• Category: Article
• ISSN: 0921-8890
• DOI: 10.1016/j.robot.2003.08.001

✦ Synopsis

Animal, especially vertebrate, visual systems exhibit extraordinary levels of plasticity in response to manipulations of the visual environment in both the developing and mature states. This plasticity is thought to reflect an underlying activity-dependent competition between afferent neurons at the anatomical and physiological levels. Such plasticity allows an animal to tune its visual system in a manner conditioned by the animal's individual morphology and its particular environment. We implement a well-characterised, biologically inspired model of developmental synaptic plasticity on a Koala robot equipped with a binocular visual system. We use this system to simulate the development of certain, well-studied anatomical characteristics of the vertebrate visual system. We find that our results are exquisitely sensitive to the statistical structure of the visual environment. This structure also has a bearing on the success or otherwise of models proposed to deal with plasticity.