Page 37 - Human Inspired Dexterity in Robotic Manipulation

Sensorimotor Learning of Dexterous Manipulation 33

condition. We also define the first context experienced by the subjects as
context A, and the following context as B. The intertrial and interblock
duration lasted <10 s, unless otherwise stated for specific experimental
conditions.
Control group (Ctrl): To establish a baseline of manipulation performance,
the first condition consisted of the block sequence A 1 B 1 A 2 B 2 , and no break
was given between blocks. Additionally, the Control group (n ¼ 16) was
recalled after 2 weeks to perform another four blocks following the same
block sequence (Fig. 3.1B).
Retention groups (RT) and Interference groups (IF): These two conditions
were designed to assess the effect of time on the retention and interference
of learned manipulation (Fig. 3.1B). Subjects were divided into six groups
which differed in terms of when the break occurred between blocks and its
duration. Specifically, the retention groups (RT10, RT20, RT60; n ¼ 12,
8, and 8, respectively) were given a 10-, 20-, or 60-min break after learning
the first context A 1 , followed by another three blocks in the sequence of
A 2 B 1 A 3 . In contrast, the interference groups (IF10, IF20, IF60; n ¼ 12,
8, and 8, respectively) were given a 10-, 20-, or 60-min break after learning
both contexts A 1 B 1 , followed by another two blocks (A 2 B 2 ). Note that for
the interference groups, all breaks were given after the object rotation.
Transfer group (TF): The third experiment was designed to quantify the
subjects’ ability to generalize a learned manipulation from Block 1 to the
new context in Block 2 after the effect of interference decayed. A one-hour
break was inserted after a group of subjects (TF60; n ¼ 12) performed
manipulation in context A1 and rotated the object (Fig. 3.1B). Subjects then
performed another three blocks (B 1 A 2 B 2 ) after the break.
The instructions for the activities allowed during breaks were the same as
those used by previous retention studies using force-field reaching tasks
[34,35]. For Conditions RT, IF, and TF, subjects were asked to remain
seated in the chair if the break time was 10 min. For longer breaks, subjects
could leave the room during the break to resume normal activity. All sub-
jects were also asked to avoid activities that involved large forces (e.g., heavy
lifting) during the breaks.
Random group (Rndm): We also tested a group of subjects (n ¼ 12)
who performed the first 16 trials with a pseudo-random presentation of
the two contexts (trial sequence: ABBAABABBBAABBAA), followed
by two blocks of eight consecutive trials in each context (Fig. 3.1B). Similar
to the Ctrl group, breaks between trials lasted <10 s.