Figure 6.

The BBD self-generates a learned sequence of arm motions using the sequence generation network model. The BBD was trained to move its arm through four different postures such that its hand traced out a quadrilateral in space. The figure shows the superposition of hand positions recorded during 20 s of training (green line) and during 20 s of self-generated movements after training (red line) for all 5 subjects. Lines are drawn between temporally adjacent data points recorded at 50 ms intervals. One of the five subjects showed some error and consistently “cut” the upper corner, generating a different shape than the other four subjects. The self-generated arm trajectories approximate the training trajectories.