Page 66 - Human Inspired Dexterity in Robotic Manipulation

62 Human Inspired Dexterity in Robotic Manipulation

invariant features. First, hand paths in point-to-point movements tend to be
straight and smooth. Second, the velocity profile of the hand trajectory is
bell-shaped [3, 4].
A variety of computational schemes capturing these invariant features can
be found in the literature [5–9]. A large number of these schemes is based on
the optimization theory, where the trajectory of the human arm is predicted
by minimizing, over the movement duration, an integral performance index
subject to boundary conditions imposed on the start and end points. Most
popular models employ the minimization of a squared hand jerk [10] or joint
torque change [11]. These models have been well tested in numerous exper-
iments with human subjects, and have also been successfully applied to the
generation of human-like trajectories for robotic systems.
The optimization models are not just computational tools for fitting
experimental data. They suggest organizing principles for the internal rep-
resentation of movements in the central nervous system (CNS) [12–14]. If
the manifested principles are correct, the corresponding computational
models should also work (i.e., reproduce the invariant features of human-
like movements) not only for the free-space movements but also for move-
ments where the human hand is “embedded” into a dynamic environment.
An example of such an environment is shown in Fig. 5.1. Here, a flexible
object is connected to the hand, and a human subject is requested to produce
a reaching movement and stop the object without excitation of oscillations.
Despite the seeming simplicity, this task requires a lot of skills that must be
acquired by practicing.

Fig. 5.1 Reaching movement in a dynamic environment: rest-to-rest maneuver of two
beads (x 1 and x 2 ) suspended to the hand (x h ).