IEEE Robotics & Automation Magazine - September 2017 - 37

(a)

(b)

Figure 5. (a) The robot has to grasp the ball on the table in front of it. As the ball is far away from the robot but reachable without
moving its feet, the robot bends forward. However, to maintain balance, the left arm moves behind it. (b) The robot has to grasp a
ball in front of it and a ball behind [the ball behind has been intentionally placed at the end position of the left hand from (a)]. It is
possible to identify the differences between these two motions using the Kinetography Laban system.

humanoid robot will be notated, and, as they appear very
similar, the only difference will be the presence of two balls in
the second scenario and only one ball in the first scenario. By
using a high level of detail, the actions of grasping the balls are
only part of a complex movement. However, results in [18]
show that the movements of the two scenarios can be distinguished in the proper task space. The presented method takes
advantage of the knowledge of the task the robot is able to
perform and how the motion is generated from this set of
known controllers to reverse-engineer an observed motion.
The method is based on the projection operation into the null
space of a task to decouple the controllers. In other words,
access to the motor-control space to distinguish similar looking movements is exactly what the Kinetography Laban system, which is designed to describe human movements, is not
able to do [18].
Naturalness of Movements: Implicit Rules
of the Kinetography Laban System
In [6], a simple Kinetography Laban score of the Tutting
Dance sequence, i.e., a dance that mainly involves arm and
hand movements, has been scored according to the Kinetography Laban system and translated in robot motion (Figure 6).
The method is based on the SoT, a robot programming system introduced in [19]. The 27 principal direction symbols
used to describe the Tutting Dance [6] are the starting point
to translate the Kinetography Laban score in the SoT. In
other words, depending on the current configuration of the
humanoid robot Romeo, each principal direction symbol is
translated as a task in the operational space. Indeed, each
direction symbol specifies the main directions and levels
with respect to the point of attachment of the body part to
which the symbol refers [see Figure 2(c)]. As a consequence,
with respect to the point of attachment, it is possible to associate a homogeneous transformation matrix to each symbol
that specifies both the position and orientation of a reference

frame at that direction and level. Based on the current position of the body part and the desired position specified by
the principal direction symbols, a task function is defined as
the error in terms of both rotation and translation between
the current position in space of the reference frame attached
to the free end and the desired one. The SoT software [20] is
then used to determine suitable control signals for the motor
of the robot such that the error becomes zero, while guaranteeing other tasks at the same time (e.g., maintain static equilibrium, maintain the static position of the parts of the body
that are not involved in the movement, and so on). This
results in a dynamic hierarchy of tasks.
Once the whole movement is translated in the SoT, suitable
control signals are sent to the motor of a simulated version of

6

6

5

5

4

4

3

3

2

2

1

1
0

0

(b)

(a)

Figure 6. A comparison between the Kinetography Laban score
of the first part of the Tutting Dance executed by a dancer and
Romeo. The main difference is in the path of the free end of the
arm (here considered to be the wrist). The path for Romeo is
along a straight line from the initial position to the final position
(see Figure 7). There are also several movements of the torso.
Moreover, the arm is a little curved during the movement, and the
gestures are slightly overlapped (the movement in the original
score is a "staccato" movement). (a) The Kinetography Laban score
for a dancer and (b) the Kinetography Laban score for Romeo.

SEPTEMBER 2017

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

37
Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2017
