Figure 1.
The behavioral task for the robot and system overview.
(A) The task for the robot is to repeatedly produce two different types of behavior: (i) move the object up and down three times at the position L, and (ii) move the object backward and forward three times at the position R. For each series of actions, the robot began from the home position and ended at the same home position. The robot repeatedly generates the same series of actions unless the object was located at the same position. The object position was switched by an experimenter at unpredictable timing. (B) System overview.
Figure 2.
Neural activity and task behavior in normal and disconnected networks.
(A) Flexible switching of behavior through the bottom-up modulation process induced by sensory perturbation. (B) Outwardly normal behavior with intermittent increases of prediction error and aberrant modulation of intentional states (arrows) induced by simulated mild functional disconnection in the hierarchical network. (C) Cataleptic and (D) stereotypic behavior induced by the severe disconnection. Arm: 4 dimensional joint angles. Vision: relative position of the object (x-y axis). A long sideways rectangle indicates the single unit activity over many time steps. Colors of rectangles indicate activation level (cf. color bar). Low and High indicates activity of units in the lower level and the higher level of the network. Intent indicates the activity of parametric bias (PB) units in the higher level, whose activity corresponds to the top-down intention for the task behavior (see Methods). Pred error indicates prediction error accumulating for the past 25 steps.
Figure 3.
Changes in prediction error and robot behavior associated with levels of disconnection.
(A) Prediction error for various levels of disconnection is shown. Bars in the graph correspond to mean values over 30 trials for each parameter setting. Error bars indicate the degree of standard deviation. (B) Changes in robot behavior with various levels of disconnection are shown. Bars in the graph correspond to the occurrence ratio of each behavior type over 30 trials for each parameter setting. Levels of disconnection are determined by the parameter κ (see Method).