Abstract: For the movement balance control, a sensorimotor system cognitive model based on operant conditioning principle is presented, and the cooperation among cerebellum, basal ganglia, cerebral cortex and other nervous organs is studied, so that a cognitive model of sensorimotor system is established. The cognitive model mainly consists of cerebellum, basal ganglia and cerebral cortex. Thereinto, the cerebellum realizes sensorimotor nerve mapping from states to actions by supervised learning; the basal ganglia implements probabilistic choice of actions based on operant conditioning principle according to action forecast evaluation results; moreover, the cerebral cortex receives and sends useful information to the basal ganglia and cerebellum, consequently, the closed feedback loop of a sensorimotor system is formed. This paper presents concrete structure, function and algorithm of the sensorimotor system cognitive model based on operant conditioning principle, and performs some simulation experiments on the two-wheeled robot which reproduces the self-learning ability and movement balance skill like human or animal. The results indicate that this model has the good cognitive characteristics and can make the robot master the movement balance control skill through self-learning.