Effective robot arm motions: stability and efficiency through natural dynamics

While progress in many fields of robotics has been swift, robot arm movement
in scenarios without contact has changed little in the last decades. This lack of
change is not due a lack of potential for improvement. After all, the human arms
that these robot emulate move in ways that are more robust, energy efficient and
adaptable. This thesis is inspired by human movement skill in these three aspects
to improve the movement of robotic arms in non-contact situations. The six main
contributions of this thesis are divided over those three aspects. The aspects are
studied for two motions, the reaching motion, that is, move from the initial position
to a pre-specified target position and back, and the pick-and-place motion, which
adds picking and placing an object at the initial and target positions respectively.
The first aspect, robustness, is studied from a stability standpoint. The first aspect
is the topic of the first part of this thesis, which contains four of its six main
contributions. Stability, as understood in this thesis, is the property of returning
to a fixed (desired) motion after an initial disturbance two motions. This form
of stability is a minimal requirement for successful task completion. Most current
robot arms rely on fast sensory feedback for their stability. This contrasts with
humans, who rely on skill at choosing motions that are intrinsically stable. Such
self-stable motions have been used by earlier robotics researchers to make robots
that juggle or walk without the need for sensory feedback. However, these robots
perform tasks involving impacts, which can have a large stabilizing effect. No
such impacts are available in reaching motions. A self-stabilizing reaching motion
instead depends on ingenious use potential energy and centrifugal or Coriolis
effects.