The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9685

Title:Robot Control Using the Halperin Neuro-Connector Model
Date: 1996
Abstract:This study involves the implementation of the biologically inspired Halperin Neuro-connector model for learning as a controller on a transputer based autonomous robot. The Halperin net was designed to explain the social behaviour of Siamese fighting fish, and has been adopted as a potential robot controller by robotocists because of its mathematical simplicity and fast learning rate. The net uses a variation of Hebb's rule of neuronal coactivation, [Hebb 49], to determine strengthening at a synapse. It is the coordination of the ends of firing of the neurons that is important in this model. The net can learn very quickly, but the down side of this is that it is very sensitive to timing of the neuron firing, to such an extent that it has been difficult in practice to chain together sequences of behaviours which in theory and simulation are possible. This study specifically addresses the issue of behavioural chaining on a mobile robot. The effect of the number of behaviours and the timing of inputs during the behaviour chaining are discussed.

[Search These Pages] [DAI Home Page] [Comment]