IEEE Robotics & Automation Magazine - September 2014 - 127

One recent work by some of the authors [13] presents the
validation of the joint reconstruction method through a comparative analysis with the joint angles measured by an optoelectronic motion capturing system. The mean error committed in joint reconstruction is 8.3*e-3 rad (with a standard
deviation of 1.8*e-3 rad).

Robotic Arm
The design of the robotic arm was chosen by evaluating the
ADLs that the system must perform and the corresponding
model of the human arm reachable workspace (ARW) [16].
The human ARW that was used includes three rigid links: 1)
the shoulder link (which represents the clavicula and the
scapula), 2) the upper arm link (which represents the
humerus), and 3) the forearm link (which represents the
radius and the ulna). The model describes the Cartesian coordinates of the wrist (considered the reference point) that correspond to the end-effector coordinates.
Because of the anatomical properties of the arm, bones, and
muscles, the lower and upper limits of the arm joint angles are
strictly related. Referring to the upper-limb kinematic chain
[Figure 3(a)], shoulder flexion-extension ^q 2h limits can be
defined as linear functions of the shoulder adduction/abduction ^q 1h, while the limits of the shoulder intra-/extrarotation
can be stated as functions of both shoulder abduction-adduction ^q 1h and flexion-extension ^q 2h . The limits of the other
joint angles are considered constant and independent. Anthropometric data, such as the limits of the joint angles and lengths
of the shoulder, upper arm, and forearm links, were derived
from [16]. Finally, the human ARW was computed considering all of the joint angles (from the lower to the upper limit).
To analyze the workspace needed for six typical ADLs: eating with a spoon, combing one's hair, picking up or hanging
up an object, drinking water from a glass, and washing one's
face/brushing one's teeth, the arm kinematics of the tested
healthy subjects were reconstructed during the execution of
ADL tasks. The experiments were conducted at the Bioengineering Institute of UMH in a room in which no external
stimuli could disturb the subject. Each subject performed
three repetitions of each ADL task. The ADL tasks were carried out in either a standing or sitting position, depending on

Physiological Signals
Psychophysiology is the branch of psychology that is concerned with the physiological bases of psychological processes. The most common measures of psychophysiology
used in human-robot interaction studies include:
1) The cardiovascular system: heart rate variability, respiratory
sinus arrhythmia, cardiac output, interbeat interval, and
blood pressure.
2) Electrodermal activity: skin conductance and galvanic skin
response (GSR).
3) The respiratory system: breaths per minute and respiration
volume.
4) The muscular system: electromyography.
5) Brain activity: electroencephalography and brain imaging
methods [15].
Pulse, GSR, skin temperature, and respiration rate
(breaths per minute) were used to monitor the emotional
state of the user during the execution of a drinking task
assisted by the robotic device. Pulse, GSR, and skin temperature sensors were placed on the fingers of the user's
left hand. The user's pulse was measured using a plethysmographic pulse sensor placed on the distal phalanx of
the thumb, GSR was recorded using two electrodes placed
on the medial phalanxes of the second and third finger,
and skin temperature was measured using a thermistor
located on the palmar surface of the distal phalanx of the
fifth finger. The user's respiration rate was computed by
employing a thermistor placed beneath the
nose; its larger part covered the mouth, while
its smaller part was slightly bent so that it
Patient Monitoring System
could not enter the nose. All signals were sampled at 256 Hz, using the g.US-Bamp USB BioControl System
signal Amplifier, from g.tec Medical EngineerGenerator
of
ing GmbH, and they were processed through
Modulation Functions
the MATLAB/Simulink software.
To remove the noise and artifacts from the
acquired physiological signals, they were preproControl of the Acoustic
Low-Level Interaction
cessed; then, aiming to determinate user's emoand
Visual
Feedback
Control System
tional state, different features were extracted from
the acquired signals.
Force Sensor
To measure the interaction forces and torques
between the user and the robotic arm, a force/
torque sensor (from ATI Industrial Automation)
was added at the end-effector. Furthermore, an
electromagnetic system was employed to attach
the user's arm and rapidly release it from the
robotic arm in case of an emergency.

Graphical Interface

Robot Arm

Patient
Figure 5. The block scheme of the overall system-the control system is reported in
the dotted block, and its component modules are specified.

september 2014

IEEE ROBOTICS & AUTOMATION MAGAZINE

127

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2014