Perception in real and artificial insects : a robotic investigation of cricket phonotaxis

• Main Author: Webb, B. H.
• Published: University of Edinburgh 1993
• Subjects: 591.5
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.663586

The aim of this thesis is to investigate a methodology for studying perceptual systems by building artifical ones. It is proposed that useful results can be obtained from detailed robotic modelling of specific sensorimotor mechanisms in lower animals. By looking at the sensory control of behaviour on simple biological organisms, and in working robots, it is argued that proper appreciation of the physical interaction of the system with the environment and the task is essential for discovering how perceptual mechanisms function. Although links to biology, and concern with perceptual competence, are fields of growing interest in Artificial Intelligence, much of the current research fails to adequately address these issues, as the model systems being built do not represent real sensorimotor problems. By analyzing what is required for a model of a system to contribute to explaining that system, a particular approach to modelling perceptual systems is suggested. This involves choosing an appropriate target system to model, building a system that validly represents the target with respect to a particular hypothesis, and properly evaluating the behaviour of the model system to draw conclusions about the target. The viability and potential contribution of this approach is demonstrated in the design, implementation and evaluation of a mobile robot model of a hypothesised mechanism for phonotaxis in the cricket. The result is a robot that successfully locates a specific sound source under a variety of conditions, with a range of behaviour that resembles the cricket in many ways. This provides some support for the hypothesis that the neural mechanism for phonotaxis in crickets does not involve separate processing for recognition and location of the signal, as is generally supposed. It also shows the importance of understanding the physical interaction of the system's structure with its environment in devising and implementing perceptual systems. Both these results vindicate the proposed methodology.