Publication Summary and Abstract
Prescott, T. J. Gurney, K. Gonzalez, F. M., Humphries, M. & Redgrave, P. (2003), Action selection in a robot model of the basal ganglia: The role of simulated dopamine, Society for Neuroscience, Program No. 915.5.
Manipulations of tonic dopamine (DA) levels in the basal ganglia affect action selection (behavioural switching). DA antagonists impair instrumental and spontaneous behaviours and can induce akinesia or bradykinesia, while DA agonists can cause increases in behaviour switching or lead to patterns of behavioural stereotypy. To further the understanding of basal ganglia function we have successfully embedded a high-level computational model of basal ganglia circuitry within the control architecture of a mobile robot engaged in a simulated foraging task. A key assumption in this model is the differential sensitivity of sub-types of striatal neurons to DA modulation. Specifically, DA is assumed to enhance cortical inputs to neurons with predominantly D1-type receptors, whilst reducing the effectiveness of inputs to D2-type cells. With simulated tonic dopamine (simDA) at optimal mid-range levels, the robot selected actions appropriate to different combinations of extrinsic (perceptual) and intrinsic (motivational) variables, and generated integrated sequences of 'purposive' behaviour. When the level of simDA was reduced or increased results were observed that parallel outcomes seen in animal studies of DA modulation. First, lowering simDA caused bradykinesia and difficulty in initiating behaviours. Behaviours were executed normally only when they carried a high level of urgency or 'salience'. Second, an increase in simDA caused either an increase in behavioural switching, or parallel selection of two incompatible behaviours, resulting in a mixture of two activity patterns. Where either the early interruption of a selected behaviour, or the 'distortion' caused by dual selection prevented behaviour completion, the robot could become stuck in a 'behavioural trap'. These correspondences between animal and robot behaviour provide further validation of our functional model of basal ganglia architecture.
Copy of the article (PDF)
Copy of the article (Powerpoint)