Nof S.Y. Springer Handbook of Automation

Подождите немного. Документ загружается.

1485

Service Robot

84. Service Robots and Automation

for the Disabled/Limited

Birgit Graf, Harald Staab

The increasing number of elderly people is re-

sulting in increased demand for new solutions

to support self-initiative and independent life.

Robotics and automation technologies, initially

applied in industrial environments only, are start-

ingtomoveintooureverydaylivestoprovide

support and enhance the quality of our lives. This

chapter analyzes the needs of disabled or limited

persons and discusses possible tasks of new assis-

tive service robots. It further gives an overview of

existing solutions available as prototypes or prod-

ucts. Existing technologies to assist disabled or

limited persons can be grouped into stand-alone

devices operated by the user explicitly such as

robotic walkers, wheelchairs, guidance robots or

manipulation aids, and wearable devices that are

attached to the user and operated implicitly by

measuring the desired limb motion of the user

such as in orthoses, exoskeletons or prostheses.

Two recent developments are discussed in detail

as application examples: the robotic home assis-

tant Care-O-bot and the bionic robotic arm ISELLA.

One of the most important challenges for future

developments is to reduce costs in order to make

assistive technologies available to everybody. On

the technological side, user interfaces need to be

designed that allow the use of the machines even

by persons who have no technical knowledge and

that enable new tasks to be taught to assistive

84.1 Motivation and Required Functionalities 1486

84.2 State of the Art ....................................1486

84.2.1 Mobility Aids ............................. 1487

84.2.2 Guidance Robots ........................ 1489

84.2.3 Manipulation Aids ...................... 1490

84.2.4 Orthoses and Exoskeletons .......... 1491

84.2.5 Prostheses................................. 1492

84.3 Application Example:

the Robotic Home Assistant Care-O-bot..1493

84.3.1 History

of Care-O-bot Development ........ 1493

84.3.2 Key Technologies........................ 1494

84.3.3 Applications .............................. 1496

84.4 Application Example:

the Bionic Robotic Arm ISELLA ................1496

84.4.1 Service Robot Arms

and Drive Technology ................. 1496

84.4.2 The DOHELIX Muscle .................... 1498

84.4.3 The ISELLA Robot Arm .................. 1498

84.5 Future Challenges.................................1499

References ..................................................1499

robots without much effort. Finally, safe manip-

ulation of assistive robots among humans must

be guaranteed by new sensors and corresponding

safety standards.

In the last years, the percentage of elderly people

in our society has grown rapidly. Out of 82.5million

people living in Germany in 2005, according to

numbers from the German Federal Statistical Of-

ﬁce, around 19% were seniors above 65years [84.1].

With this demographic development continuing, by

the year 2050 those above 65years will comprise

33–36% of Germany’s population. Similar numbers

are reported from other industrial nations all over

the world, in particular the USA and Japan. Many

disabilities come with age. In order to provide suf-

ﬁcient support for the growing number of disabled

persons, new solutions to assist these people are re-

quired.

Part I 84

1486 Part I Home, Ofﬁce, and Enterprise Automation

84.1 Motivation and Required Functionalities

Technical aids enable people in need of support and

care to live independently in their accustomed home

environments for a longer time. As not all homes are

suitable for installing intelligent home technologies as

described in Chap.83, mobile service robots able to

navigate and operate in existing homes without mod-

iﬁcations to the environment provide a more ﬂexible

solution. Such service robots not only fulﬁll the user’s

desire for independence and autonomy, but also help to

avoid the high costs of individual treatment in nursing

homes that might otherwise be necessary.

The task of a service robot or other automation

technologies for the disabled and limited is to provide

support and instructional help in a person’s daily life

and to promote self-initiative. The group of target users

for such new technologies includes:

•

Elderly and frail persons

•

Disabled persons (physically, psychically)

•

Severely sick persons (heart attack, plastered leg)

•

Chronically sick persons (diabetes, epilepsy)

•

Severely restrained persons (e.g., pregnancy).

Due to the fact that humans have for a long time

dreamt of an assistant to relieve them of difﬁcult han-

dling tasks, the basic concept of service robots assisting

people in their daily life has been presented and their

functionalities been described long before they were

actually built. The requirements that a service robot

to assist the disabled must cope with and the corre-

sponding technical abilities can therefore be speciﬁed

precisely [84.2, 3]. Some of the most important tasks,

identiﬁed by a study done with elderly people and their

care-givers, are summarized below.

Household Tasks

•

Execute everyday jobs such as serving drinks, set-

ting the table, operating the microwave, simple

cleaning tasks

•

Fetch and carry objects, e.g., books, remote control,

medicine

•

Support in grasping, holding, and lifting objects and

tools

•

Control of the technical home infrastructure, e.g.,

heating system, air-conditioning, lights, windows,

doors, alarm system, etc.

Mobility Aid

•

Support for standing up from the bed or a chair

•

Controlled motion

•

Obstacle detection and guidance to a target.

Communication and Social Integration

•

Media management (videophone, TV, stereo, inter-

active media, etc.)

•

Daytime manager (daily routine, time for medicine,

etc.)

•

Communication with medial and public facilities

(physician, authorities, etc.)

•

Supervision of vital signs and emergency call

functionality.

84.2 State of the Art

In recent decades, and continuing to the present day,

service robots for the disabled and limited is an ac-

tive ﬁeld of research and development, performed in

many countries throughout the world. Main research

is carried out at universities and research institutes,

covering abrought rangeof ideas,approaches, andtech-

nologies. The landscape of companies in this branch

seems to be twofold: on the one hand there are a few

large players, who have been providing orthopaedics

solutions in large numbers and for many years al-

ready, trying to introduce robotics into some of their

high-end products. On the other hand there are many

smaller high-technology companies, often launched by

researchers in this ﬁeld, focusing on a few or a single

product solution.

This section provides a representative overview of

the stateof the art in theﬁeld of assistiverobotic devices

for the disabled and limited. The following examples

may be understood as representatives of many more but

similar developments. For further insights one may re-

fer to proceedings of international conferences such as

those on Robotics and Automation (ICRA), Intelligent

Robots and Systems (IROS), Robotics and Applica-

tions (RA), and others that usually host large sessions

on assistive and rehabilitation robotics. The Interna-

tional Conference on Rehabilitation Robotics (ICORR)

Part I 84.2

Service Robots and Automation for the Disabled/Limited 84.2 State of the Art 1487

Service robots and automation

for the disabled

Stand-alone

devices

Wearable

devices

Robotic walkers

Fixed-platform manipulators

Portable-platform manipulators

Mobile autonomous robots with

manipulation skills

Robotic wheelchairs

Orthoses and exoskeletons

Prostheses

Mobility aids

Manipulation aids

Guidance robots

Fig. 84.1 Classiﬁcation of service robots and automation for the disabled

is focused on this topic alone. Other sources may be

periodicals such as the quarterly International Journal

of Rehabilitation Research (IJRR), the monthly Inter-

national Journal of Therapy and Rehabilitation (IJTR),

the quarterly International Journal of Medical Robotics

and Computer Assisted Surgery, and others.

Existing service robots and assistive devices for the

disabled and limited can be grouped into stand-alone

devices, operated by the user explicitly such as robotic

walkers, wheelchairs, guidance robots or manipulation

aids, and wearable devices that are attached to the user

and operated implicitly by measuring the desired limb

motion of the user such as in orthoses, exoskeletons or

prostheses. A similar classiﬁcation is found in [84.4].

Figure 84.1 summarized this classiﬁcation of service

robots and assistive devices for the disabled andlimited.

The review of existing solutions startsin Sect. 84.2.1

with mobility aids such as robotic walkers or robotic

wheelchairs providing autonomous or semiautonomous

navigation capabilities. Section 84.2.2 describes ex-

isting guidance robots that may be applied for blind

persons or those suffering from dementia and other

mental disorders. From the technical viewpoint they

are quite similar to robotic walkers but the focus of

these systems is on cognitive, not physical support. Sec-

tion 84.2.3 describes existing manipulation aids that

may be either static, such as in desktop-mounted ma-

nipulators, or mobile when attached to a wheelchair or

to another kind of mobile platform.

Finally the group of wearable devices is reviewed.

Section 84.2.4 describes orthoses and exoskeletons for

users who need direct support in moving their limbs.

Prostheses worn by a user to replace amputated limps

are increasingly incorporating robotic features and are

reviewed in Sect. 84.2.5.

84.2.1 Mobility Aids

Robotic Walkers

Robotic walkers are enhanced off-the-shelf walkers

equipped with robot technology such as environment

sensors and drives. Using these technologies, they are

able to provide additional support to their user, ranging

from audio or visual information on the environment

to autonomous or semiautonomous navigation. For

existing robotic walkers, two basic types can be distin-

guished: passive and active robotic walkers.

Passive robotic walkers do not have any driven

wheels and move directly according to the applied

user forces (direct user control). Examples of passive

walkers are the COOL Aide [84.5](Fig.84.2a) and

PAMM-AID [84.6] systems, which usemotorized steer-

ing of the wheels to lead the user around obstacles. The

RT-Walker [84.7] uses special servo brakes to steer the

device in a collision-free direction. In case no obstacles

are detected the user has full control of the device and

can move it, similarly to a conventional walker. Most

passive systems are equipped with brakes to stop them

if they get too close to an obstacle or step. A guid-

ance system for Guido, the commercial successor of

PAMM-AID, has been presented [84.8](Fig.84.2b).

After a target has been set, the robot will plan and fol-

low a path to the target. During guidance the desired

direction indicated by the user input will be ignored.

The successor of the RT-Walker, ORTW-II [84.9], uses

a potential canal method which only allows deviations

up to a certain distance from the optimal path to the

target. The robotically augmented walker developed at

Carnegie Mellon University (CMU) [84.10](Fig.84.2c)

does not use its motors when traveling with a user but

tracks its position and displays the optimal direction

Part I 84.2

1488 Part I Home, Ofﬁce, and Enterprise Automation

a) c)b)

Fig. 84.2a–c Examples of passive robotic walkers: (a) COOL Aide (courtesy of the University of Virginia, USA),

(b) Guido (courtesy of Haptica Ltd., Dublin, Ireland), and (c) CMU robotically augmented walker (courtesy of Carnegie

Mellon University, Pittsburgh, USA)

of travel on a screen to guide its user to the selected

target.

Active robotic walkers are equipped with motorized

wheels. Force or force-torque sensors of different kinds

are used to determine the forces applied to the system

by the user. These forces are used to calculate the de-

sired driving direction and speed of the walker (indirect

user control). Some systems (Silbo [84.11](Fig.84.3a),

Hitachi walker [84.12], RT-walker) are equipped with

angle sensors, enabling the required force input to be

adapted on slopes. All systems are able to detect ob-

stacles and stop in front of them; most of them adapt

the traveling direction to surround obstacles in advance.

The PAMM smart walker [84.13](Fig.84.3b) and the

CMU robotic walker [84.14](Fig.84.3c) are able to lo-

calize themselves in their environment and thus plan an

optimal path to a given target. The shared control sys-

tem of PAMM creates a virtual force, leading the robot

to the given target, which is combined with the real

forces applied to the device by the user. The gains of

a) b) c)

Fig. 84.3a–c Examples of active

robotic walkers:

(a) Silbo (courtesy

of Intelligent Healthcare Labora-

tory, Korea),

(b) Smart walker PAMM

(courtesy of Massachusetts Institute

of Technology, USA), and

robotic walker (courtesy of Carnegie

Mellon University, Pittsburgh, USA)

each control input are set depending on the observed

user abilities. Whereas for PAMM an interface provid-

ing feedback to the user on the planned path has not

been presented, the CMU robotic walker displays the

desired direction of travel on a screen. The user is free

to move in any direction; however, if the deviation from

the planned path is too large, the velocity of the walker

will be reduced to force the user back to the path.

Other robotic walkers focus their works on addi-

tional support functions such as lifting assistance (Hi-

tachi walker, Monimad walker [84.15], MOBIL Walk-

ing & Lifting Aid [84.16], Walking helper II [84.17]),

person tracking (MOBIL Test Bed [84.18]) or moving

out of the way when not used (CMU robotically aug-

mented walker).

Robotic Wheelchairs

Similar to robotic walkers, robotic wheelchairs provide

enhanced safety and/or improved navigation capa-

bilities using mobile robotics technologies. Robotic

Part I 84.2

Service Robots and Automation for the Disabled/Limited 84.2 State of the Art 1489

wheelchairs are based on electric wheelchairs where

control over thedrives is at some point taken over bythe

integrated personal computer (PC). In order to perceive

the environment, robotic wheelchairs are equipped with

environment sensors such assonar sensorsor laserscan-

ners. Existing control systems for robotic wheelchairs

can be divided into two basic types [84.19]: model-

based approaches and behavior-based approaches.

Model-based shared control systems analyze the in-

put of the user and the measured environment data in

order to identify the intended travel direction of the user

and thus adapt the motion of the wheelchair. To iden-

tify the intention of the user, the current sensor input

of the user, e.g., the position of the joystick, is com-

pared with a previously recorded motion model and an

appropriate action is selected. The robotic wheelchairs

NavChair [84.20] or VAHM [84.21] use a method that

generates a histogram of the environment identifying

the optimal direction of travel (minimum vector ﬁeld

histogram, MVFH). The user is responsible for high-

level control of the system, such as route-planning and

some navigation actions, while the machine overrides

unsafe maneuvers through autonomous obstacle avoid-

ance. It can provide addition assistance, e.g., for safely

passing narrow passages such as doors. The MVFH al-

gorithm ensures that the input of the user is considered

at all times and thus the user always feels in control

of the wheelchair. SmartChair [84.22] is another rep-

resentative of model-based shard control systems. This

robotic wheelchair is not only able to adjust the input

of the user for collision-free motion; it further provides

localization and path-planning capabilities to guide the

user to a previously speciﬁed target. SmartChair uses

a potential ﬁeld method to calculate a suitable mo-

tion direction to the given target. MAid [84.23]isan

autonomous wheelchair able to safely pass crowded en-

vironments, however, at the cost of not considering the

input of the user during motion.

Behavior-based shared control systems assume that

the driving behavior of the robotic wheelchair can be

classiﬁed into a few basic tasks such as high-speed

driving, close-quarter manoeuvring, and docking ma-

noeuvres. From the input of the user and the measured

environment data, a suitable motion behavior is se-

lected. The crucial factor in behavior-based shared

control systems is to ﬁnd out which behavior is cur-

rently most appropriate. Wheelsley [84.24] is a robotic

wheelchair based on the well-known behavior-based

subsumption architecture [84.25]. In free space, the

robot moves along with maximum velocity in the di-

rection indicated by the user. If an obstacle in the

a) b)

Fig. 84.4a,b Examples of advanced robotic wheelchairs: (a) Rol-

land III (courtesy of the DFKI-Labor, Bremen, Germany) and

(b) OMNI (courtesy of Forschungsinstitut Technologie und Behin-

derung der Evangelischen Stiftung Volmarstein, Germany)

desired motion direction is detected, obstacle avoidance

behavior is activated to pass by the obstacle. Similar ap-

proaches are applied in the Rolland [84.26](Fig.84.4a),

RobChair [84.27], and Sharioto [84.28] projects. Some

motion behaviors conﬂict with each other, such as

passing a narrow passage and obstacle avoidance. An-

other conﬂict occurs if the user wants to dock to an

object while at the same time collision avoidance is re-

quired. For docking to an obstacle, in [84.28] the user’s

conﬁdence when guiding the wheelchair in a speciﬁc

direction is evaluated. Only if the user moves straight

and without ﬂuctuations towards an obstacle will the

direction be maintained.

Other robotic wheelchairs such as the ofﬁce

wheelchair with high manoeuvrability and naviga-

tional intelligence for people with severe handicap

(OMNI) [84.29](Fig.84.4b) provide advanced assis-

tance through omnidirectional navigation capabilities

and height adjustment. The special mechanics of the

iBOT (http://www.ibotnow.com/) enable stair-climbing.

Even legged chairs such as i-foot, one of the part-

ner robots by Toyota designed to assist people [84.30],

which is able to move along in uneven terrain, have

recently been introduced.

84.2.2 Guidance Robots

Guidance robots are used to assist people with men-

tal weaknesses and diseases or blind people. Guidance

robots for the blind require direct physical contact to

the user. One of the ﬁrst robotic guidance systems

Part I 84.2

1490 Part I Home, Ofﬁce, and Enterprise Automation

a) b) c)

Fig. 84.5a–c Examples of guidance robots: (a) GuideCane (courtesy of University of Michigan, USA), (b) Nursebot

(courtesy of CarnegieMellon University, Pittsburgh, USA), and

is the NavBelt [84.31], a device worn by the user.

Equipped with sonar sensors, NavBeld is able to de-

tect obstacles in front of the user and issue a warning.

GuideCane is the cane-based continuation of this re-

search (Fig. 84.5a). In contains a small mobile robot as

a base that carries the environment sensors and a cane

that the user can hold onto. Using a small joystick, the

user indicates the desired travel direction and is led

along by the robotic base. If an obstacle is detected by

the sensors, the path will be modiﬁed to lead the user

around the obstacle safely.

Other guidance robots are designed to assist the el-

derly, for example, to guide them to speciﬁc locations

in elderly care facilities. In this case, no direct contact

with the user is required. Nursebot [84.32](Fig.84.5b)

is a development that targets this application. Guidance

robots with similar capabilities are also applied in mu-

seums or exhibitions [84.33](Fig.84.5c).

84.2.3 Manipulation Aids

Manipulation aids are applied to assist disabled users

in eating, drinking or object replacement. This Section

describes existing ﬁxed- or portable-platform manipula-

tors as well as existing autonomous mobile robots with

manipulation skills.

Fixed-Platform Manipulators

The professional vocational assistive robot (ProVAR)

[84.34] is a research prototype designed to assist in-

dividuals with a severe physical disability. The system

consists of an industrial manipulator attached to an

overhead track suspended above a worktable and is

able to handle objects for a user who has difﬁcul-

ties in handling these objects himself. The manipulator

is controlled by voice commands or head motion in-

puts. The robot to assist the integration of the disabled

(RAID) [84.35] is a robotic workstation for use by in-

dividuals with little or no upper-limb function in ofﬁce

environments. It is basedaround a custom-designed me-

chanical structureproviding storagefor books, manuals,

paper documents,and otherreference materials.A robot

arm is mounted on a linear track in front of the stor-

age zones. It is able to assist the user, e.g., to transport

books or diskettesor to hold abook and turnpages when

reading it. The arm’s working envelope includes a large

proportion of the desk area but does not reach to theuser

for reasons of safety.

Portable-Platform Manipulators

Portable platform manipulators include manipulators

attached to a mobile base with castor wheels that can

be rolledaround andwheelchair-mountedmanipulators.

Handy1 [84.36](Fig.84.6a) is an example of the ﬁrst

group of assistive manipulators. It consists of a robot

arm and a tray that can be selected for several appli-

cations such as feeding, cleansing, and make-up. The

robot is controlled by a switch input used in conjunction

with a linear scanning control system which is suit-

able for different disability groups. The assistive robot

service manipulator (ARM), which is also known as

Manus [84.37](Fig.84.6b), is a 6+2 degree of freedom

(DOF) robot that assists disabled people with a severe

handicap of their upper limbs. It compensates their lost

arm and hand function. It can be mounted on an electric

wheelchair (or mobile base) and allows numerous daily

Part I 84.2

Service Robots and Automation for the Disabled/Limited 84.2 State of the Art 1491

living tasks to be carried out in the home, at work, and

outdoors. By means of an input device such as a key-

pad (4×4 buttons), a joystick (e.g., of the wheelchair)

or another device attached to a nondisabled body part,

the manipulator can be operated to grasp objects with

its gripper.

Mobile Autonomous Robots

with Manipulation Skills

Mobile autonomous robots with manipulation skills are

a popular topic of current research. Even though re-

search in this ﬁeld has grown signiﬁcantly over recent

years, no commercial products have been placed on the

market so far. Among existing prototypes, two basic ap-

proached can be observed: human-like robots equipped

with legs and arms and wheel-based robots.

Humanoid robots with legs are most popular in

Japan. Most of the currently existing prototypes are not

speciﬁcally designed to assist disabled people. How-

ever, for many robot developers, this application is seen

as one of the most important for future products. The

partner robots developed by Toyota [84.30], for ex-

ample, were ﬁrst demonstrated at EXPO 2005 AICHI,

Japan. During the EXPO they were applied to entertain

the visitors of Toyota’s pavilion by playing different in-

struments. However, this was only the ﬁrst step towards

the creation of robots that can use tools, assist people,

and live in harmony with us. According to the part-

ner robot developers, currently, new robots are being

developed that can provide elderly care to help Japan

cope with its rapidly aging population. The goal of

the ASIMO [84.38](Fig.84.7a) development of Honda

is quite similar: to develop a robot that can duplicate

the complexities of human motion and genuinely help

people. Even though ASIMO is mainly used as a re-

search platform at the moment, someday ASIMO might

help with important tasks such as assisting the elderly

or a person conﬁned to a bed or a wheelchair. HRP-

2 [84.39](Fig.84.7b) by Kawada and the US-American

SARCOS [84.40] humanoid robot are similar platforms

that can be used for experiments to further develop

robotic technologies developments.

One of the ﬁrst wheel-based robots able to assist dis-

abled people with daily tasks is MOVAID by the Scuola

Superiore Sant’Anna in Italy [84.41]. The robot is able

to navigate in homelike environments and to perform

simple manipulation tasks. Another early development

is Hermes [84.42], developed at the Bundeswehr Uni-

versity Munich. Hermes is able to explore unknown

environments, to fulﬁl transportation and manipulation

tasks inspacious human-populatedareas, and to interact

a) b)

Fig. 84.6 (a) Portable-platform manipulators Handy1 (courtesy of

Forschungsinstitut Technologie und Behinderung der Evangelis-

chen Stiftung Volmarstein, Germany) and

(b) ARM (courtesy of

Exact Dynamics BV, Netherlands)

and communicate even with novice users in a natu-

ral and intuitive way. Research on assistive robots is

currently being performed in several projects in Ger-

many such as the Collaborative Research Center on

Humanoid Robots in Karlsruhe [84.43], where several

generations of the ARMAR platform (Fig.84.7c) have

been set up in the last years. One of the most advanced

developments in wheel-based robotic home assistants

is the German Care-O-bot [84.44] which will be de-

scribed in detail in Sect.84.3.

Recent developments of wheeled robots in Japan in-

clude the excellent mobility and interactive existence

as workmate (EMIEW) by Hitachi [84.45] and Smart-

Pal by Yaskawa [84.46]. At the EXPO 2005 AICHI,

Japan, they were introduced in a bar scenario, where

they demonstrated their abilities to serve drinks to visi-

tors. By now, for each robot, the second generation has

recently been introduced. Another interesting Japanese

development is enon (exciting nova on network)by

Fujitsu [84.47], designed for duties such as providing

guidance, transporting objects, and security patrolling.

In Japan, limited sales of enon have already been an-

nounced.

84.2.4 Orthoses and Exoskeletons

Orthoses are understood as orthopaedic devices to sup-

port parts of the body with reduced mobility. There are

those for ﬁxation at reversible or nonreversible defor-

mations and bad postures as well as for recovery (e.g.,

ruff). Others are used to support, e.g., arch supports and

orthopaedic corsets. The latter are said to be passive

if they support only, and active if they force the body

Part I 84.2

1492 Part I Home, Ofﬁce, and Enterprise Automation

a) b) c)

Fig. 84.7a–c Popular humanoid and

wheel-based robots with manipula-

tion skills:

(a) ASIMO (courtesy

of Honda Motor Co., Ltd., Japan),

(b) HRP-2 (courtesy of National In-

stitute of Advanced Industrial Science

and Technology, Japan),

(courtesy of the Collaborative Re-

search Center 588 Humanoid Robots

– Learning and Cooperating Multi-

modal Robots, Karlsruhe, Germany)

to actively correct, e.g., mis-statics of the spine. Any

orthoses usually requires signiﬁcant ﬁne-tuning in or-

der to meet the body conditions as well as the medical

demands of an individual.

Utilizing robotics for orthoses is limited to a few

application examples and there is a fuzzy border to-

wards the prostheses domain, however there is some

research and development (R&D) purely for orthoses.

Most researchers focus on rehabilitation, e.g., of stroke

patients. Sometimes an orthosis may look quite simple

at ﬁrst sight, but there may be a lot of complexity and

careful thoughts behind any combination of materials,

part design, and choice of mechanisms. An earlier ex-

ample is the Wilmer elbow orthosis [84.48], a passive

foot-drop orthosis from the University of Hawaii-

Manoa, USA [84.49], and numerous stationary passive

or motor-powered gravity-compensation rehabilitation

a) b)

Fig. 84.8 (a) Freebal (courtesy of Baat Medical, Hengelo,

Netherlands) and

(b) Hybrid Assistive Leg (HAL, courtesy

of Tsukuba University/Cyberdyne Inc., Japan)

devices for the upper extremities such a dynamic

arm support [84.50], Freebal [84.51](Fig.84.8a), and

iPAM [84.52]. Some orthoses are equipped with elab-

orated spring and lever mechanisms, brakes, and may

also be motor-powered with advanced control algo-

rithms such as a foot orthosis from the University of

Osaka, Japan [84.53], and a knee orthoses from the

Vrije Universiteit Brussels, Belgium, powered by pneu-

matic artiﬁcial muscles [84.54].

There are also quite complex, full-body gait rehabil-

itation systems suchas HapticWalker [84.55], STRING-

MAN [84.56], and LOKOMAT [84.57], which are

composed of a stationary frame, a hanging harness for

the patient to put on, and possibly a treadmill. Current

R&D covers mechanical design as well as extending

the functionalities of existing systems regarding sens-

ing, control, haptic feedback, artiﬁcial intelligence, and

virtual reality.

The term exoskeleton is also often used for powered

orthoses, but it is not only limited to medical appli-

cations. There are those for a single extremity (upper

or lower), or covering one or few joints, e.g., knee

and ankle. And there are those for full-body motion

support, either stationary or mobile. Some examples

are ALEX (stationary, for gait rehabilitation) [84.58],

BLEEX (mobile, for military use to increase pay-

load of a soldier carrying equipment) [84.59], and

the Hybrid Assistive Leg (HAL, mobile, full-body ex-

oskeleton, Fig.84.8b) [84.60].

84.2.5 Prostheses

While orthoses and exoskeletons are attached to ex-

isting parts of the body, prostheses are replacements

of lost extremities. Most sold prostheses of arms and

hands are purely cosmetic, whereas most prostheses

of legs and feet are passive but functional, i.e., they

Part I 84.2

Service Robots and Automation for the Disabled/Limited 84.3 Application Example: the Robotic Home Assistant Care-O-bot 1493

enable patients to walk. However, scientiﬁc R&D as

well as commercial high-end products clearly aim at

coming close to the functionality of the replaced body

part. This means motorization, haptic perception, and

reliable and comfortable control by the patient. Some

advanced commercial prostheses of ﬁngers or the entire

hand are the i-LIMB [84.61] and the Fluidhand [84.62],

and of the leg are the c-leg [84.63]andthePow-

erknee [84.64].

Beside commercial developments there are a num-

ber of prostheses in research. Focus is placed on

powered prostheses and their mechanical and mecha-

tronic design, signal interfaces to the human body for

control and feedback, and advanced motion control.

There are developments for upper [84.65](Fig.84.9)

and lower limbs; some examples of recent lower-limb

developments are powered ankle–foot and knee pros-

thesis as described in [84.66] and [84.67].

Forearm

catalyst

pack

Elbow

catalyst

pack

Elbow

valve

Propellant

cartridge

Elbow

cylinder

Forearm

cylinders

Forearm

valve

bank

Gas

distribution

lines

Fig. 84.9 Steampunk artiﬁcial arm (courtesy of Vanderbilt Univer-

sity, Nashville, USA)

84.3 Application Example: the Robotic Home Assistant Care-O-bot

84.3.1 History

of Care-O-bot Development

Care-O-bot is a mobile robot assistant designed of

Fraunhofer IPA, Germany to assist people in daily

life activities. Three Care-O-bot robots have been

developed so far: The ﬁrst Care-O-bot prototype

(Fig.84.10a) [84.68]was built in 1998, when the idea of

building rehabilitation robots was still new. Care-O-bot

I is a mobile platform with a touchscreen, able to navi-

gate autonomously and safely in indoor environments,

and communicate with or guide people. As a mobile

platform alone, it can be used for transportation and

safeguarding tasks; however, it is unable to execute

complex manipulation tasks as required from a home

assistant. Care-O-bot II (Fig.84.10b) [84.69], built in

2002, is additionally equipped with a manipulator arm,

adjustable walking supporters,a tilting sensor head con-

taining two cameras and a laser scanner, and a handheld

control panel [84.70]. The manipulator arm, devel-

oped speciﬁcally for mobile service robots, provides the

possibility of handling typical objects in a home envi-

ronment. A ﬂexible gripper attached to the manipulator

is suitable for grasping various objects such as mugs,

plates, and bottles. A handheld control panel is used for

instructing and supervising the robot. In addition to all

mobility functions already solved in Care-O-bot I, the

second prototype is able to execute manipulation tasks

autonomously and can be used as an intelligent walking

support. Care-O-bot 3 (Fig.84.10c), built in 2008, is the

latest generation of this successful development series.

It is equippedwith the lateststate-of-the-art components

including omnidirectional drives, a seven-DOF redun-

dant manipulator, a three-ﬁnger gripper, and a ﬂexible

interaction tray that can be used to safely pass objects

between the human and the robot. Its moveable sen-

sor head contains range and image sensors enabling

autonomous object learning and detection and three-

dimensional (3-D) supervision of the environment in

real time.

a) b) c)

Fig. 84.10a–c Care-O-bot prototypes by Fraunhofer IPA,Ger-

many. Care-O-bot I

(a), Care-O-bot II (b) and Care-O-

bot 3

(c)

Part I 84.3

1494 Part I Home, Ofﬁce, and Enterprise Automation

a) b) c) d)

Fig. 84.11a–d Range segmentation for an object (a), object representation: side view of a Tetra Pak (b), detected feature

points

(c), and corresponding feature point cloud (d)

84.3.2 Key Technologies

Autonomous or Semiautonomous Navigation

All Care-O-bot prototypes are able to navigate au-

tonomously to a given target [84.71]. Autonomous

navigation is a key issue in order to execute a fetch and

carry task or to guide the user to a speciﬁc location.

Chapter 16 has additional insights in automation mo-

bility and navigation. A static environment map is used

to plan an optimal path to the given target. Different

approaches for global planning have been implemented

and can be selected according to the geometry, kinemat-

ics, and current operation mode of the robot [84.72].

The generated path is smoothed and eventually modi-

ﬁed in reaction to dynamic obstacles or other external

forces. A laser scanner attached to the robot is used for

continuous and dependable obstacle detection.

Automatic Object Detection

In order to grasp an object, the robot must be able to

detect relevant objects in the environment. This task is

solved by combining a range imaging sensor [84.73]

with a color camera. The recognition algorithm is based

on scale-invariant feature transform (SIFT) descriptors

that are recorded for each object and fed into a learn-

ing algorithm (a one-class support vector machine,

SVM) [84.74]. Using the data of the range imaging

sensor, feature keypoints can be segmented from the

background. A region in space is effectively masked out

in the color image of the scene using the range mea-

sures for the corresponding pixels in the range image.

New objects are taught to the robot by placing them in

front of the sensors and by recording the relevant SIFT

keypoints for the object [84.75]. Figure 84.11a displays

the teaching of new objects using the proposed range

segmentation. Figure 84.11b–d illustrates the learning

process and resulting representation of an object: sev-

eral images of the object are recorded and for each the

feature points are detected. In a second step the feature

points of all images are fused into a 3-D feature point

cloud which again can be used to detect and compute

the position of the object in a given scene.

Object Manipulation

Based on the data from the range imaging sensor and

the identiﬁed location of the object to be grasped,

a collision-free trajectory for moving the manipulator

to the detected object can be computed. To solve this,

the robot and scene are modeled using oriented bound-

ing boxes (OBBs). For the robot a distinction is made

between static components (e.g., the robot’s torso) and

dynamic components (e.g., its manipulators). The dy-

namic components are mapped by articulated models

which are updated with each robot movement. The

model of the scene is obtained by generating corre-

sponding OBB models fromthe point cloudobtained by

the range sensor. The obstacle model is used as the basis

for online collision monitoring. The algorithm consists

of two main phases: determination of potentially col-

liding objects by a rough distance check based on the

velocity vectors of all moving parts, and subsequent

elaborate collision tests for all objects in the determined

potential colliding sets. Figure 84.12ashowstheveloc-

ity vectors of a moving arm. Figure 84.12bshowsthe

manipulator about to collide with the robot torso; the

respective joints are marked in red color. Figure 84.12c

shows successful collision detection for simultaneous

movement of two manipulators.

In addition to the online collision monitoring,

the obstacle model provides the basis to compute

Part I 84.3