A Decision-Making Model Based on Basal Ganglia Account of Action Prediction

Abstract

Reinforcement learning offers robotics a framework for the design of sophisticated behaviors. The theory of it is deeply rooted in psychological and neuroscientific perspectives. The basal ganglia, thalamus and cerebral cortex constitute an important neural decision circuit involving in the behavioral choice of reinforcement learning. However, the influence of the changes of various nucleuses in the basal ganglia on decision-making is still unclear. Considering the complexity and difficulty of brain experiments on both animals and humans, it is necessary to establish a neural computing model. This paper builds a decision-making model based on physiology and anatomy of basal ganglia. The feasibility of the model was verified by comparing the behavior data from experiments and simulated by the model as well as readiness potentials and discharges rate of nucleus. Moreover, the simulation data were then used to further analyze the effects of conflict, dopamine, and subthalamic nucleus changes on behavioral accuracy and response. The results show this model opens a window for exploring specific neural mechanisms associated with basal ganglia-related decisions.