You are on page 1of 13

Humanoid Robots

Related terms:

Metamodeling, Robotics, Virtual Reality, Human Robot Interaction, Social Robot,


Autonomous System, Evaluation Methodology, Facial Expression

View all Topics

Humanoid Robots
David J. Bruemmer, Mark S. Swinson, in Encyclopedia of Physical Science and
Technology (Third Edition), 2003

IV.B Human–Robot Interaction


This area includes the study of human factors related to the tasking and control
of humanoid robots. How will we communicate efficiently, accurately, and con-
veniently with humanoids? Another concern is that many humanoids are, at least
for now, large and heavy. How can we ensure the safety of humans who interact
with them? Much work in this area is focused on coding or training mechanisms
that allow robots to pick up visual cues such as gestures and facial expressions that
guide interaction. Finally, this area considers the ways in which humanoids can be
profitably and safely integrated into everyday life.

> Read full chapter

Design of a Humanoid Hand for Hu-


man Friendly Robotics Applications
Naoki Fukaya, ... Rüdiger Dillmann, in Human Friendly Mechatronics, 2001

1 INTRODUCTION
Humanoid robots are expected to exist and work in a close relationship with human
beings in the everyday world and to serve the needs of physically handicapped
people. These robots must be able to cope with the wide variety of tasks and
objects encountered in dynamic unstructured environments. Humanoid robots for
personal use for elderly and disabled people must be safe and easy to use. Therefore,
humanoid robots need a lightweight body, high flexibility, many kinds of sensors
and high intelligence. The successful introduction of these robots into human
environments will rely on the development of human friendly components.

The ideal end-effector for an artificial arm or a humanoid would be able to use the
tools and objects that a person uses when working in the same environment. The
modeling of a sophisticated hand is one of the challenges in the design of humanoid
robots and artificial arms. A lot of research activities have been carried out to develop
artificial robot hands with capabilities similar to the human hand. The hands require
many actuators to be dexterously moved [6,7]. However, the control system of
the humanoid robot becomes more complicated if more actuators are additionally
used for the hand design. This is a key aspect for the artificial arm because a
handicapped person might not be able to control a complex hand mechanism with
many actuators. For this reason, we propose to develop a lightweight hand driven
by a single actuator. To this end we adopted a new mechanism for the cooperative
movement of finger and palm joints.

> Read full chapter

Teleoperation master arm system with


gripping operation devices
Hiroaki Kagaya, ... Susumu Tachi, in Human Friendly Mechatronics, 2001

1 Introduction
A humanoid robot has substantial advantages when working in environments where
human beings live. The main advantage is that a humanoid robot can act as human
beings in such an environment without any previous adjustment for the robot. On
the other hand, human friendly and functional machinery become more neccesary
as robots are used closer to human beings to care.

Based on the needs above mentioned, since 1998 fiscal year, AIST, which belongs
to MITI, has promoted the reserch and development project of “Humanoid and
Human Friendly Robotics System” as a part of the Industrial Science and Technology
Frontier Program(ISTF).
In the first term, from 1998 to 1999 fiscal year, platform systems as a common base
of the research and development has been developed. In the second term, from 2000
to 2002 fiscal year, various kinds of element technologies as for applications, in which
humanoid and human friendly robots are expected to be used, will be developed by
using the platform systems developed.

A teleoperation platform system, which is one of the platform systems, consists of a


humanoid robot and a remote control cockpit system to operate the robot.(Figure 1)
The cockpit system communicates with the humanoid robot which exists at a remote
site. The communication utilizes wireless or optical fibers LAN. The remote control
cockpit system makes an operator possible to operate the humanoid robot with
sense of high reality. The system consists of a master arm system to operate a
humanoid robot arm with reacting force sensation, an audio-visual display system
to provide realistic information as for robot’s views and surrounding sounds, and
a motion-base system to provide the operator with motion sense of the humanoid
robot. We have developed the master arm system and the motion-base system[1].
In the following chapters, the features of the master arm system and the several
experimental results will be shown.

Figure 1. Teleoperation platform system

> Read full chapter

Humanoid robot “DB”


Shin’ya Kotosaka, ... Stefan Schaal, in Human Friendly Mechatronics, 2001

2.2 Humanoid robot research for the practical verification of


motor learning
Recently, several researcher groups have started to investigate “Humanoid” robots
from various aspects. One of the recent study is concerned with trajectory generation
for walking, dancing or jumping movements (9)(10). A full body humanoid robot was
built to generate several kinds of human-like motion. The P3 (11) by Honda Corp. is
a self-contained type humanoid robot. It can fully autonomously walk on moderately
irregular terrain. Humanoid robots are also investigated as communication devices.
Another novel study is the Cog project (12) at MIT and the Humanoid robot project
(13) at the Waseda Univ. Both research groups focus on investigating cognitive be-
havior with their humanoid robots. The collaboration with psychology, neuroscience
and robotics in such projects seems to be very productive from both a technological
and biological point of view (14).

One of the most interesting research areas for neuroscience and robotics research
is the theory of motor learning in humans, and humanoid robots can effectively
be used to validate research hypotheses. What kinds of capability are needed for a
humanoid robot in such a research area? One of the most important properties is
that kinematics and dynamics are similar to humans, e.g., that weight, size, position
of the center of the mass, and hopefully the viscoelastic properties of the joints are
human-like. Equally important is the availability of sensory information to mimic
human proprioception and that joint torques can be produced to realize human
levels of performance, but also human limitations.

> Read full chapter

Recent trends towards cognitive sci-


ence: from robots to humanoids
Sarthak Katiyar, Kalpana Katiyar, in Cognitive Computing for Human-Robot Inter-
action, 2021

Cognition for human–robot interaction


Robots with cognitive abilities are necessary to support human–robot interaction
(HRI) for fulfilling of necessities, capabilities, task, and demand of their human
counterparts.

Cognitive-human–robot interaction is a very active and bourgeoning area of re-


search focusing on human(s), robot(s) and their articulated activities as a cognitive
system for developing algorithms, models and design guidelines to enable the
design of such system. Thanks to phenomenal advances undertaken in human
neuroscience, HRI is moving out from fiction world to real world context. Key to
such HRI is the need for the development of suitable models capable to execute
joint activity for HRI, a deeper understanding of human prospects and cognitive
responses to robot actions (Fig. 2.12).
Figure 2.12. Cognition for human–robot interaction.

The interaction among human–robot occurs in a connected local environment


furthermore data sharing occurs via speech conversation, motion gestures, and
social intent. Interactive object modification is assumed to be accomplished by
the robot while considering the determination, viewpoint, and beliefs of a human
being. So that humans may interact with robot in explicit mode (human–robot
verbal communication) or implicit mode (human–robot nonverbal communication)
(Lemaignan, Warnier, Sisbot, Clodic, & Alami, 2017).

There are plans to extend the capabilities of this framework by developing a unified
human–robot system that work in pro-active (by preparing and suggesting outcome
schedule) and reactive aspect. The HRI should be safe, efficient, comprehensible
with maintain adequate proxemics.

The robot must be skilled to undertake collaborative task, in pro-active (by preparing
and suggesting outcome schedule) and reactive aspect. The robot should be allowed
to function and operate safely and logically, following ethics of the society. The com-
munication and joint action execution between the human and the robot indicates
a need to take action in order to implement cognitive skills in robots (Knoblich,
Butterfill, & Sebanz, 2011; Sebanz, Bekkering, & Knoblich, 2006). The actions are
making a collaborative objective which has been earlier ascertained and consented,
establishment of a realistic ecosystem in which the exteroceptive sensing skills of
a robot is supplemented by conclusion extracted from earlier observations and
creating a state of reality that include a deduced general knowledge, understanding
and behavior for both the robots as well as its human partner involved.

Specifically, the deliberative architecture of a robot is designed to share space


and tasks with humans, and to act and communicate in a praxis that reinforces the
human’s own actions and decisions.
Implementation of cognitive computing in human–robot interaction
Several cognitive components have been already incorporated and embedded into
humanoid robot system. A humanoid robot prototype based on ARMAR consists of
five subsystems from the kinematics control point of view. The right arm, left arm,
head, torso and a movable platform are subsystems of a humanoid that provide 23
degrees of freedom to robots. The separate modules have also existed for the robot
control system. Each subsystem of the robot has its own hardware, software and
control module (Lemaignan et al., 2017).

The orchestration of various autonomous system components in a unified robotic


framework design poses a number of technological challenges but also a sustainable
design challenge in implementation of CC in HRI. Existing studies have relied that
it is easier to achieve human-level interaction if the robot itself relies internally
on human-level semantics (Fong et al., 2005; Trafton et al., 2013). The proposed
technique relies on having substantial information illustration and modification.
The interaction of software components occurs via first-order logic statements
arranged in ontologies and whose semantics are similar as those of modified by
users. A comprehensive overview of architecture is presented here in which an active
knowledge base (Oro) act as a semantic blackboard for assembling most of the
modules. The geometric reasoning module (Spark) generate a considerably high
frequency symbolic claims illustrating the state of the robot environment and its
evolution over time. The knowledge base also stores logical statements for future
reference required by the language processing module (dialogs), the symbolic task
planer (HATP) and execution controller (Shary or pyRobots). the language processing
module outputs and the robot controller activities are stored back as symbolic
statements. An example of this is illustrated by a cup lying on table might be picked
by the Spark and represented in symbolic terms as <CUP 1type Cup, CUP1 is on
TABLE>. Later the robot was subjected to process an additional sentence like “give
me another cup?.” To resolve this issue, the dialog module would then query the
knowledge base and find (?obj type cup,? obj different from CUP 1), and inscribe
assertion like <HUMAN desires GIVE_ACTION45, GIVE_ACTION45 acts on CUP2>
to knowledge base. This would in turn trigger the execution controller Shary to
prepare act. It would first call the HATP planer which uses Oro to initialize the
planning domain (CUP 2 is at? location), and a full symbolic plan transferred to
the execution controller. Finally, the controller must carry out the plan and track
its achievement both for itself and for human. The above-mentioned architectural
design philosophy is capable to nurture the robot’s decision-making components
with models of human conduct and expectations for the purpose of building an
efficient artificial cognition for a robot that can assist and communicate with
humans effortlessly. The main features of the architecture are Beliefs, Desires,
Intentions (BDI) and known as BDI architecture (Weiss, 1999).
The knowledge model architecture of RDF, knowledge manipulation is performed
by a central server (the Oro server). The users of the Oro server are in charge
for making any change in the knowledge as server is pro-active in this regard.
The Oro server depends on Description Logics (OWL) to illustrate and modify
information. The advantages of description logic are good understanding of its
trade off. The OpenCyc (Lenat, Guha, Pittman, Pratt, & Shepherd, 1990), DBPedia,
RoboEarth (Waibel et al., 2011) and WordNet are open access online knowledge
base. Beside above-mentioned architecture, the Prolog/ OWL combination relied
upon by KnowRob (Tenorth & Beetz, 2009), answer Set Programming has been used
in robotics to be practical, the first-order logic and OWL ontologies have basic,
efficient and adequate symbolic structure for practical diligences. The Open Robots
common sense ontology has been designed for fulfilment of our experimental
needs.

Symbol grounding or anchoring is the role of building and sustaining a bidirectional


connection among sub-symbolic (sensor data and actuation) representations and
symbolic representation that can be modified and cleared out (Coradeschi & Saffiotti,
2003; Harnad, 1990). Taking the context of HRI the cognitive ability is of specific
importance. The connection among the knowledge model to the perception and
actuation capabilities of the robot is acquired through symbol grounding. The
geometric reasoning and dialog processing modules are components of symbol
grounding. These components assist robots in building and pushing new symbolic
contents about the world to the knowledge base.

> Read full chapter

Teleportation of human body kinemat-


ics for a tangible humanoid robot con-
trol
Tutan Nama, Suman Deb, in Cognitive Computing for Human-Robot Interaction,
2021

NAO robot platform


SoftBank Robotics formally known as Aldeberan Robotics developed an integrated,
programmable, medium-sized humanoid robot called NAO. NAO (version V6) is a
58 cm, 5 kg robot, communicating with remote computers via an IEEE 802.11 g
wireless or a wired Ethernet link. The nao has 26 DOFs and features a variety of
sensors and joints. The bipadel locomotion of the human body like structure is
created with several geared joints with functional torso, a head, two arms, and two
legs. The head joint has two types of motion HeadYaw and HeadPitch. Each of those
motions is purely manipulated through the LMC in this project work. Each arm of a
nao robot has a total of six joints, namely, Shoulder Pitch, Shoulder Roll, Elbow Yaw,
Elbow Roll, Wrist Yaw, RHand, or LHand, shown in Fig. 12.1. All these arm joints are
successfully manipulated in this project work through a LMC.

> Read full chapter

Applications
Erik T. Mueller, in Commonsense Reasoning (Second Edition), 2015

14.3 Vision
Murray Shanahan and David Randell use the event calculus to implement the
higher-level vision component of Ludwig, an upper-torso humanoid robot. Ludwig
has two arms, each with three degrees of freedom, and a stereo camera hooked up
to a head with two degrees of freedom (pan and tilt). The low-level vision component
uses off-the-shelf edge detection software to map raw images from the camera into
a list of edges. The list of edges is fed to the higher-level vision component, which
is responsible for recognizing shapes. This component is implemented using the
event calculus in Prolog.

The higher-level vision component consists of three layers. The first layer generates
hypotheses about what regions are in view, based on the input list of edges. The
second layer generates hypotheses about what aspects are in view, based on what
regions are in view. The aspects of a shape are the various ways it can appear when
viewed from different angles; for example, a wedge viewed from above appears to
be a rectangle, but a wedge viewed from the side appears to be a triangle. The third
layer generates hypotheses about what shapes are in view, based on the aspects that
are in view over time.

We present here a simplified version of Ludwig’s third layer. The predicate Arc(s, a1,
a2) represents that, by gradually changing the orientation of a shape s with respect
to the camera, it is possible for the appearance of s to change from aspect a1 to
aspect a2. For example, the appearance of a wedge can change from a rectangle
to a rectangle plus an adjacent triangle, but it cannot change immediately from a
rectangle to a triangle—it must first appear as a rectangle plus an adjacent triangle.
The predicate Shape(o, s) represents that object o has shape s. The fluent Aspect(o,
a) represents that the appearance of object o is aspect a. The event Change(o, a1, a2)
represents that the appearance of object o changes from aspect a1 to aspect a2.
We start with state constraints that say that an object has unique shape and aspect:

We have effect axioms that state that if the aspect of an object is a1, the shape of the
object is s, it is possible for the appearance of s to change from aspect a1 to aspect
a2, and the object changes from a1 to a2, then the aspect of the object will be a2 and
will no longer be a1:

Now, suppose we have the following visual knowledge about two shapes Shape1 and
Shape2:

Further, suppose we observe at timepoint 0 that Object1 has Aspect1 and at


time-point 1 that Object1 has Aspect2:

We can then show that the shape of Object1 must be Shape1 and not Shape2:

Note that this inference depends crucially on the sequence of images over time
rather than on a single image. These sorts of inferences are supported by the ability
of the event calculus to reason about time.

> Read full chapter

Avatar embodiment experiences to en-


hance mental health
Laura Aymerich-Franch, in Technology and Health, 2020

Technical features of avatar embodiment technologies


Virtual reality is the most extended and studied form of avatar embodiment. As
mentioned earlier, though, another form of avatar embodiment has been report-
ed, which uses humanoid robots to also provide avatar embodiment experiences
(Aymerich-Franch et al., 2015, 2016, 2017a, 2017b; Cohen et al., 2014b, 2012;
Kishore et al., 2014, 2016a,b).

In order to experience sense of embodiment in the avatar, first-person visual per-


spective from the avatar is combined with multisensory correlations between the
physical and the avatar body. These two aspects are crucial to experience the body of
an avatar as one's own (Lenggenhager, Tadi, Metzinger, & Blanke, 2007).

Regardless of the technology used (i.e., virtual reality or robots), the user is gen-
erally equipped with a head-mounted display, which provides first-person visual
perspective from the virtual or robotic avatar. The user is able to see the limbs and
part of the body of their avatars if they look down, where they would normally see
their real limbs and torso. In addition, full-body identification can be achieved by
reflecting the avatar's appearance in physical and virtual mirrors or other surfaces
(Aymerich-Franch et al., 2016, 2015; Aymerich-Franch, Kizilcec, & Bailenson, 2014;
González-Franco, Pérez-Marcos, Spanlang, & Slater, 2010).

While feedback from other senses are not considered a necessary condition to induce
the illusion of avatar embodiment, it can contribute to enhance the embodiment
experience (Spanlang et al., 2014). Headsets or speakers are used to provide auditory
feedback and haptic devices for force feedback, or object controlling is used for haptic
feedback (Fox, Arena, & Bailenson, 2009; Stone, 2001). Olfaction and gustation are
generally not implemented.

For visuomotor synchronization, head tracking and body movement synchronization


are provided. In virtual reality, the movements of the user's head are followed and
used to update the user perspective in real time (Spanlang et al., 2014). When
the illusion is provided with a robot avatar, head movements are synchronized to
the robot's head movements, and the user receives video feedback from cameras
mounted on the robot's head in real time (Aymerich-Franch et al., 2015, 2016, 2017a,
2017b, 2019; Kishore et al., 2014, 2016a,b).

User movements can be tracked and synchronized to the avatar's movements for the
control of limb and body gestures and to make the avatar walk or move in the space.
In virtual reality, user's body movements are generally tracked and synchronized to
the avatar body movements, and spaces are rendered according to these movements
(Fox et al., 2009; Spanlang et al., 2014). For robot avatars, control of the robot body
movement can be obtained with a motion capture suit (Aymerich-Franch, Kishore,
& Slater, 2019), a joystick (Aymerich-Franch et al., 2015, 2016), a brain–computer
interface (Alimardani et al., 2013; Gergondet et al., 2011), fMRI (Cohen et al., 2012,
2014), or eye-tracking technologies (Kishore et al., 2014).

> Read full chapter

Multimodal open-domain conversa-


tions with robotic platforms
Kristiina Jokinen, Graham Wilcock, in Multimodal Behavior Analysis in the Wild,
2019

1.3.1 Multimodal WikiTalk for robots


After prototyping WikiTalk on a robot simulator as illustrated in [20], the system was
implemented on the Aldebaran NAO robot as described by [9] and [21]. Using a hu-
manoid robot enabled us to include multimodal communication features, especially
face-tracking and gesturing, which were integrated with the spoken conversation
system. This required an event-driven system to manage speech events and action
events, as shown in Fig. 1.5.

Figure 1.5. Multimodal event-driven architecture on the robot.

The robot needs to estimate the user's level of interest in the topic, and the human's
proximity and gaze both are important for this. The robot also integrates nodding,
gesturing and its body posture with its speech during the conversation. Issues in
synchronizing gestures with speech are discussed by [24].

As an example of multimodal gesturing, beat gestures are small vertical hand


movements that have a pragmatic function of emphasizing and giving rhythm
to the speech. These beat gestures are usually synchronized with NewInfos, and
serve a similar role as intonation in distinguishing new, not expected information
from the old, expected topic information. With this multimodal communication
management, the visual gestures emphasize the least known elements so that the
partner will more easily notice the new information and understand it. Interaction
management is closely related to information presentation: planning and generat-
ing appropriate responses, giving feedback, and managing topic shifts.

Assessing the level of interest of the user has two sides: how to detect whether the
human partner is interested in the topic or not, and what the system should do as
a result. Detecting the level of interest is part of the system's external interface, and
deciding what to do about it is part of the system's internal management strategy.
In order to assess the interest level correctly, the external interface should not be
limited to verbal feedback, but should include intonation, eye-gaze, gestures, body
language and other factors. The internal strategy for reacting appropriately must
decide what to do not only if the user is clearly interested or clearly not interested,
but also how to continue when the interest level is unclear, which may be a more
difficult decision.

Information about the evaluation of the Nao robot system based on the recorded
user testing sessions at the 8th International Summer Workshop on Multimodal
Interfaces, Metz, 2012 is given in [9] and [3].

> Read full chapter

Designing Tools that Care


Yanghee Kim, ... Jeffrey Thayne, in Emotions, Technology, Design, and Learning,
2016

Conclusion
Emotion research suggests that, to help learners with various challenges, making
efforts to build learners’ positive affect should be an essential, preliminary step to any
type of instructional support. Tools, such as virtual peers and humanoid robots, are
relatively free from the biases in the real world, so they might provide a safer learning
environment for students who can be marginalized in regular education classrooms
for various reasons. Also, by adding some affective pedagogical strategies, the
relatively simple tool of instructional videos can be utilized to enhance social and
affective qualities in online learning.

Virtual peers, robots, and videos can be effective and affective tools when they are
designed carefully with awareness of affective influence on users. When creating
programs with virtual peers, designers may keep in mind the power of a peer
to positively engage learners and enhance learning experiences. When designing
applications for educational robots, designers should consider how affect opens up
several exciting opportunities for further exploration and innovation. It seems that
affect is more related to the robot’s capacity to replicate human behavior than a
robot’s capacity to replicate a human’s physical abilities. Last but not least, when
incorporating videos into online learning, instructors should first be aware of the
affordance and limitation of the medium, as well as of the research in this area.
Unless instructors take care to ensure that students are engaged on an affective
level, instructors who use online videos may see their students’ positive attitudes
decrease. It should be further explored how online videos can be deliberately de-
signed to increase positive student affect. The case for doing so is strong, but the
research on precisely how to do so is still weak. As our study suggests, the use of
relationship-building strategies may help.
Overall, it is clear that rendering social and relational learning contexts to educa-
tional tools is feasible and crucial to success in technology-based learning. Building
from of our initial findings, subsequent research needs to be done on how those
technologies can mediate and encourage positive learning experiences in both
one-on-one and group settings.

> Read full chapter

ScienceDirect is Elsevier’s leading information solution for researchers.


Copyright © 2018 Elsevier B.V. or its licensors or contributors. ScienceDirect ® is a registered trademark of Elsevier B.V. Terms and conditions apply.

You might also like