Page 35 - Human Inspired Dexterity in Robotic Manipulation

Sensorimotor Learning of Dexterous Manipulation 31

have to produce a torque to counter-balance the object’s asymmetrical mass
distribution. It has been shown that the learning of this dexterous manipu-
lation task has very limited generalizability: following the object’s physical
rotation, subjects have to re-learn digit-force distribution to generate a tor-
que appropriate to minimize object tilt [27–29]. The question arises about
whether the physical rotation of the target object is the primary factor
that prevents the generalization. Physical rotation of the object may force
subjects to perform a mental rotation of the previously established sensori-
motor memory [18] similar to the mental rotation in visual object recogni-
tion which is considered computationally challenging for the brain [30].
Alternatively, failure to generalize learned manipulation may be an intrinsic
feature of the sensorimotor system not limited to tasks involving changes of
object orientation. We will review the main findings of one of our exper-
imental approaches to demonstrate, both empirically and theoretically, how
context-dependent parallel learning processes could interact during learning
of dexterous manipulation.

3.2.2 Interaction Between Multiple Sensorimotor Processes
Underlies Learning Dexterous Manipulation
3.2.2.1 Subjects and Apparatus
Sixty-four healthy right-handed subjects participated in this experiment.
Subjects were randomly assigned to one of five conditions. We asked sub-
jects to grasp and lift a single L-shaped object (Fig. 3.1A), which had a base

Fig. 3.1 (A) Two alternative presentations, context left (L) and right (R), of the L-shaped
object and their corresponding compensatory torque directions (clockwise and
counterclockwise arrows). (B) Sequence of manipulation contexts, transfer and
retrieval trials, and breaks for each experimental condition. (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.)