Exploring adaptive behaviour control with an animat inspired by the Siamese fighting fish (Betta splendens)

grantor: University of TorontoA computational neuroethological simulation (virtual animat) was implemented that loosely emulates the physical form and behavioural abilities of the Siamese fighting fish, a popular animal subject for aggression behaviour research. The animat is equipped with tactile sensors and visual sensors as well as a number of action primitives. The behaviour of the animat is generated by a neurobiologically inspired behaviour control system. The animat was employed in two different domains of adaptive control. In the first, the animat was used to demonstrate how design, genetic algorithms, hebbian association learning, could be used to produce a robust obstacle avoidance competency. In the second, the animat was used to demonstrate a novel implementation of Halperin's machine motivation model of behaviour selection and reinforcement. The animat was equipped with one control system to show simplified Pavlovian learning and another to modulate its display to a moving rival versus a stationary one.M.Sc