Page 43 - Human Inspired Dexterity in Robotic Manipulation

Sensorimotor Learning of Dexterous Manipulation 39

Fig. 3.2 Experimental results from the Ctrl group. Trial-by-trial compensatory torque
(T com ) from blocks 1–4. The task torque (T target ; thin black dashed line) is defined as
positive for the first context (A) and negative for the second context (B). (Modified
from Q. Fu, M. Santello, Retention and interference of learned dexterous manipulation:
interaction between multiple sensorimotor processes, J. Neurophysiol. 113 (1) (2015)
144–155.)

When switching back to context A in Trial 1 of Block 3 after eight trials
in context B, subjects were not able to perfectly recall the manipulation that
was previously learned in context A, as indicated by the underestimation of
the T com (113.91 15.76 N mm; Fig. 3.2). Therefore, the RI quantifying
the extent of performance degradation was significantly negative (Fig. 3.3A;
RI ¼ 174.60 15.76 N mm; one sample t-test, P < .001). Furthermore,
we found that subjects also failed to perfectly recall context B in the first trial

Fig. 3.3 Experimental results from Ctrl group. (A) Retrieval index (RI) plotted against the
duration of the break between blocks. (B) Comparisons of first trials and transfer trials
from the Ctrl vs. transfer (TF60) groups. Data are averages of all subjects. *P < .05.
(Modified from Q. Fu, M. Santello, Retention and interference of learned dexterous
manipulation: interaction between multiple sensorimotor processes, J. Neurophysiol. 113
(1) (2015) 144–155.)