|
| Personal | Forskning | Utbildning | Publikationer | Sök |
| OBS: Arkiv – underhålls ej mera |
Robotics in rehabilitation provide considerable opportunities to improve the quality of
life for physically disabled people. However, practical results limited, mainly due to the
need for developing new robotics concepts where people are where they are working in
separeted areas. This paper reveals some of the developments needed and presents two
projects currently underway at Lund University. The first one is concerned with
end-effector design for a robotic workstation for officebased tasks, while the second is
concerned with a mobile robotic system for use in medical an chemical laboratories by
disabled people. Both projects show promising results. There is also a need for further
research in developing new robotic systems for use in rehabilitation with new mechanical
features, as well as programming and control suitable for any user.
REHABILITATION is an activity which aims to enable a disabled person to reach an
optimum mental, physical, and/or social functional level. Thus, rehabilitation robotics
deal with advancing robotic´s technology to provide physically disabled people with tools
to improve their quality of life and productivity of work [Man machine interaction].
Examples within this application area include vocational tasks, such as, manipulative
operations in a structured environment (paper handling in office-based work, test
procedures in laboratory-based work, etc.) and daily living activities in structured and
nonstructured enviroments, such as game playing, educational tasks, eating, and personal
hygiene [Evaluation of Prototype and Improvement to RAID workstation]. This implies the
use of robots in a way that is quite different from industrial applications where robots
normally operate in a structured enviroment with predefined tasks, separated from human
operators. Furthermore, industrial robots are operated with specially trained personnel
with a certain interest in the technology. This may not be the case in rehabilitation
robotics. Thus, rehabilitation robotics have more in common with service robotics which
integrate humans and robots in the same task, requiring certain safety aspects and special
attention to man-machine interfaces for people with low programming interest or people
with physical problems operating a specific programming device. Therefore, more attention
must be paid to the user´s requirements, as the user is a part of the process in the
execution of various tasks. Although a home-based service robot for general purpose use is
needed, we have selected two application areas with emphasis on structured enviroments,
such as those normally found in vocational workplaces. This enables us to concentrate on
functionalities defined or evaluated by users rather than novel robotics research which
may be difficult to develop to a stage necessary for practical evaluation by disabled
users within a limited timeframe.
However, it is recognized that there is a need for research and development in robotics
to focus on developing more flexible systems for use in unstructured enviroments. Such
developments of importance in rehabilitation robotics concern, among others, the following
topics:
These areas will be further described in the following sections.
Robotics for use by the disabled is an application area where robots, from a home-based perspective, integrate robots and humans both in a common workspace and in the execution of the same work task. Therefore, the mechanical design of robots for rehabilitation must consider different specifications, compared to those used in industrial applications, which may affect design aspects of the mechanical structure. Examples of differences are:
set-up at new workplaces;
Nevertheless, most robots used in rehabilitation today have similarities with
industrial robots, such as the RT-series robots and SCORBOT, which were developed for
educational purposes. An example of an adaptation of a robot for rehabilitation purposes
is HANDY1, which is used to assist in eating [3], and DeVar, which uses a PUMA robot for
assisting the disabled at home-based or vocational workplaces [4]. However, new designs
are on the way which will include the use of compact and flexible arms, as well as new
drives/actuators. Examples of this include the wheelchair-based Manus robot [5], the Tou
soft (flexible) assistant arm [6], the pneumatically driven Inventaid arm [7], [8], and
the compliant actuator Digit Muscle [9], Wheelchair-mounted manipulators show an increase
in interest not only through the manipulator itself, but also through the enhancements of
the wheelchair control by providing it with sensors and control systems like any mobile
robotic base [10]. The development of flexible arm/link systems will also have a great
impact on gripper systems, which need a high degree of flexibility in terms of
maneuverability and dexterity. Despite these developments, much work is needed in the area
of mechanical design, specifically the introduction of composite materials in the arm
structure with inbuilt strain gauges wich may be used as flexible links with feedback of
the deflection and redundant kinematics for optimal reachability.
A basic goal in rehabilitation robotics is to be able to use the robot for a task that
is done only once. This is in contrast with most industrial uses of robots, where robots
are used in preprogrammed repetitive tasks. Another difficulty is that robots for
rehabilitation may be used by "anyone", in contrast with industrial robots,
which are operated by skilled people with, in most cases, an interest in robotic´s
technology. Thus, many tasks in rehabilitation robotics can be said to be unique in the
sense that a motion for a certain task, e. G., picking up a newspaper or opening a door,
can not be preprogrammed. This indicates that there is a need for manual or direct control
of the robot in a way to a telemanipulator. Also needed is an increased use of sensors to
guide the robot and increase the performance in autonomous tasks, and interface devices to
program and control the robot arm. It should also be noted that direct control of the
robot arm puts a high cognitive load on the user and that physically disabled persons may
have difficulties operating joysticks or push buttons in delicate movements. Thus, there
is an obvious need for a certain degree of autonomy of the robotic system, such as
automatic grasping., which includes recognition of a specified object in front of a
sensor. A positive factor in this context is that there is a human operator working with
and supervising the robot. Therefore, if a task fails, to a limited extent, the user will
be able to correct the situation.
Programming and manual control of the robot relates to a high degree with MMI which,
for disabled people, not only put certain demands on programming languages, but also on
input devices by which the user can interact with the system. Generally speaking, robot
systems should be developed to allow any input devices connect to the standard set of
devices, such as keyboard emulation, mouse emulation, and serial communication through
RS-232/422 interface. As more severely disabled people need individual adaptation, such
work is normally done at rehabilitation centers. However, in the RAID project described in
this paper, the joystick used to control the electrical wheelchair is interfaced with the
control language of the robot and mouse control function on the PC computer. This is, for
most users, a good solution, as most of them can control the wheelchair with the same
control device.
Taking into account both the need for an interactive programming method, as well as
different interfacing devices depending on the individual disability, several attempts
have been made to provide programming and control methods which resemble the interactive
use of most modern graphical software for personal computers. As an example, most robot
languages today for industrial robots are robot-oriented in a way that the are specially
adapted for a specific robot and all operations are made upon the robot itself, e.g.,
motion types, poses, I/O. If the task is repetitive, it does not matter so much if the
robot program is defined through poses or frames which are related to the robot or
attached to objects in the environment. However, if the task is frequently redefined by
moving objects in the work space of the robot, such as paper an book handling, page
turning, etc., it is preferable to adopt an object-oriented approach. This means that the
tasks are defined by manipulating objects and that the robot must adapt its motions and
logics to fulfil the program description.
Much work in the area of rehabilitation robotics is therefore directed toward
controllers or control languages, e.g., MASTER [11], which allows the user to interact in
the performance of a task, e.g., by directing the robot through manual control, as well as
advanced sensory interfacing and object or task level description which frees the user
from concentrating on how the robot will operate to execute the tasks. An example of an
object-oriented language is CURL [12], which provides a flexible programming environment
through direct (manual) control, object manipulation, and selection/definition of
procedures. An interesting development in this area is RoboGlyph [13], which uses a set of
icons which graphically represent different robot actions on the screen like a storyboard.
This is in line with new developments of the CURL language which, by using drag and drop
techniques, make use of the possibilities of the graphics. A workstation could, for
example, represent a book shelf and the reader board. By dragging a book (document) from
the shelf to the reader board, the system will activate appropriate procedures to execute
the task. Another direction in the development of languages with high-level
characteristics are event-based controller languages and reactive planners which are based
on the state of the system and activate a certain action or procedure [14]-[17].
The EPI-RAID project is concerned with the development of a robotized computerized
office workstation RAID which was developed during an earlier project. The project is part
of the European Community TIDE program. (The partners in the EPI-RAID project are:
Armstrong Projects Ltd., UK.; Cambridge University, UK.; Oxford Intelligent Machines Ltd.;
CEA/DTA/UR, France; and HADAR, Sweden and Lund University, Sweden.)
The robotized system is intended primarily for vocational use in an office environment
(see Fig. 1). The selected application areas include CAD (Computer-Aided Design) and other
computer office tasks such as desktop publishing, graphics layout, and wordprocessing,
which are applications full of handling tasks for the robot and creative work for the
user.
During the initial work on the end-effectors, it was evident that we should design the
end-effectors with as high a degree of flexibility as possible in order to minimize tool
changing operations. The technical solution is based on two end-effecctors, called
"book gripper" and "page turner".
The two end-effectors are outlined in Figs. 2 and 3. The book gripper is designed to
handle books, catalogs, and manuals with varying thickness and geometrical size (maximum
weight 2 kg, maximum width 75 mm) between the book shelf and the reader board.
The book gripper is based on a pneumatic clamping device. The movements of the
gripper´s "thumb" are controlled by a double-acting pneumatic cylinder
(diameter 16 mm). The gripper will hold a book with a force of 30 N, if the air pressure
is set to 0,6 MPa (6bar). The grasped book is supported by a small shelf to reduce the
maximum clamping force needed. The approximate friction coefficient of the surface of the
"thumb" is one and the weight of the book gripper is 0,8 kg.
The design of the book gripper resulted mainly from the user requirements requiring
that the books be stored in a normal upright position and that the book shelf look as
normal as possible. These requirements are met with the exception that the books must be
stored with space between each object. The width of these spaces must be at least 100 mm,
which is the width of the book gripper when it is open. A photoelectric switch detects if
a book is in the gripper.
The page turner is designed to open books at any point and page-turn forward or
backward from that point. The page turner can also grasp papers and move them between the
printer, the reader board, the storage racks, and the input and output trays. The page
turner is also designed to handle disks, and drinks served on a specially designed tray.
The three main parts of the page turner are a "knife", suction cup, and
clamping device placed close to the suction cup. The "knife" is a plastic plate
with the size of a human hand. It is used for opening books and turning multiple pages
simultaneously. The suction cup and clamping device are used for single page turning. The
bellow-type suction cup lifts a single page when it reaches the page surface. A push
button is mounted next to the suction cup and detects when the suction cup has reached the
page surface. The activated push button stops the approaching movement of the robot arm.
The page is then lifted and grasped with the clamping device, which is connected to a
double-acting pneumatic cylinder.
Some arrangements have been made at the reader board to prevent small books from moving
when they are opened and to prevent unwanted movements when pages are turned in small
books with stiff pages. A big suction cup, placed in a hole in the reader board, will
prevent small books from moving. A "finger" has been added to the lower part of
the reader board to press the pages and prevent unwanted page movements. The
"finger" is connected to a double-acting pneumatic cylinder, which is controlled
by the robot. The "finger" is removed for a short time during the page turning
process.
The "knife" is also used when handling papers (up to approximately 50 pages)
and disks, and serving drinks. The clamping force is produced by a single-acting pneumatic
cylinder (diameter 6 mm). The clamping device is activated toward the knife, which is used
as a supporting surface for the papers, the disks, and the drink tray. A force of 15 N
will hold the objects, if the air pressure is set to 0,6 MPa (6bar). The approximate
friction coefficient of the surface of the clamping device is one and the weight of the
page turner is 0,7 kg. A photoelectric switch detects if an object is in the gripper.
The end-effectors are mounted on a robot tool changer, which makes it possible for the
robot to change end-effectors automatically. The tool changer also increases the
flexibility of the RAID workstation. New handling tasks which perhaps would require a
separate gripper, could then be added more easily. The possibility of adapting RAID to
individual needs is an important user requirement.
1 Book Gripper: The time to move a book from the book shelf to the reader board
is 60 s. It is expected that this figure can be reduced by 40% during an optimized work
cycle. Grasping a book from the shelf has not caused any problems. When positioning soft
catalogs at the reader board, the robot has to make some extra movements to prevent the
pages from being folded. In addition, to grasp a book from the reader board has caused
some problems with varying positions of the book in the gripper. However, this does not
cause any problems when returning the book to the shelf again, except for catalogs that
have a tendency to fold.
2 Page Turner: When opening a book it is only possible to reach an accuracy of +10
pages. To reach a specific page, the user then has to turn one page at a time. The cycle
time for turning one page is 15 s. In order to test the performance of the page turner at
higher speed, the page turner was mounted and tested on an ABB Irb1000 industrial robot.
The cycle time obtained with full functionally of the page turner was 3 s and
approximately 100 pages could be turned without errors. Furthermore, in case of an error,
the robot could still proceed with the operation by turning backward and forward. The
errors during page turning were: 1) failure to lift and turn a page, 2) two or more pages
turned at one time, or 3) an uncompleted page turn. In all cases, a subsequent page turn
without human interaction corrected any problems caused by the error.
At this stage, it is not possible to have one task program for all books. Our approach
is to make one program for each book type with respect to size. Furthermore, the tilt
angle of the reader board has to be specified. It is anticipated that the angle can be a
parameter in the program. Page turning at the beginning and end of books causes some
problems because the corners are not in the same position. Some user interaction may be
needed during robot execution. Stiff pages get slightly folded in their upper corners due
to a clamping device on the page turner. Some noise is produced by a vacuum ejector and
pneumatic valves during operation. An electric vacuum pump was tested but rejected by the
user.
Disk handling tasks have proven successful. Straight line interpolation and good robot
repeatability is needed during this operation. However, the page turner is not ideal for
this task due to geometrical constraints.
A special tray was adapted to the page turner in order to serve refreshments. No
problems have occurred.
The first RAID prototype workstation has been evaluated by a group of potential users
[18]. RAID was well accepted since it addresses the need to find a vocation. The overall
impression of the workstation was positive, both regarding size and appearance.
The users´major concern was reliability of the robot tasks, e.g., turning pages in a
pile of paper sheets and returning them into the storage compartment. Occasionally, the
sheets were not aligned and fell on the floor. The users divided errors into two
categories, recoverable and unrecoverable. A stapler not feeding a staple every time was
considered as a typical recoverable error. This task could be repeated by trying a second
time. Paper sheets falling on the floor was considered as an unrecoverable error and was
not accepted.
The end-effectors were found to have a high reliability in the paper and document
manipulation tasks. However, the reliability of the tasks is not a function of the
end-effector itself, but also include the robot and peripherals. Therefore, necessary
improvements were identified concerning the robot (motion control) and peripherals
(document storage). An improved version of the RAID workstation is now undergoing
evaluation at three rehabilitation centers in Sweden, France and the U.K.
The user input device, integrating the wheelchair joystick with the computer, was part of the RAID prototype workstation. This resulted in a drastic decrease in typing speed compared to the users´normally used input devices. Thus, the input device should not be a part of the RAID workstation but should be the responsibility of the rehabilitation center responsible for the installation. Only preprogrammed tasks were
used during the evaluation. Large buttons representing different robot tasks were
representing different robot tasks were presented to the user. The user interface was
found to be easy to use and understand.
Based on the results of the user trials, the RAID workstation will be further developed
in a second stage with increased reliability and autonomy. Thus, the mechanical
functionally of the end-effectors will be redesigned concerning integration with the
necessary sensors. Much work be devoted to increasing the degree of flexibility and
autonomy so the workstation can operate in a less structured environment, and to
developing process models for generic tasks, for example, grasping a book and turning to a
specific page for different types of books.
Modularity of the workstation will also be improved to allow the user to specify the
hardware and software components, e.g., the number of compartments in the bookshelf and
automatic recognition of book sizes. In this context, users till be involved in the
development of the workstation.
A mobile robot system is being developed to work in laboratory environments (typically
chemically, medical and biological) for people with disabilities. This will increase the
working possibilities, which are normally limited to office-type work. We have found three
different workplaces which are suitable to robotize:
The system is intended for workplaces with varying workloads at different locations
during normal work hours. It has been found that this is common at hospital laboratories,
where tests may come in batches which demand different routines and equipment at the
laboratory. Thus, a mobile robot may be well suited for this kind of work place which uses
different equipment and procedures which may take up to a few hours for each working
session.
The robot task can be divided into two different problems:
It is preferential to change as little as possible in the environment at the laboratory. Thus, the size of the mobile base has to be small enough to move around in a normal laboratory, as well as move through a door case, etc.
The robot tasks have been analyzed in order to adapt grippers, special tools, and to
specify the working procedure for each task. From a user point of view, it is important to
use the robot for manipulative tasks and leave descision making and analysis work to the
human.
The mobile robot system consists of the following parts:
The robot is mounted on the mobile base which is equipped with eight ultrasonic sensors
(see Fig. 4). The sensors are used to detect obstacles and guide the robot into position
for a new task. Safety aspects are taken care of by the ultrasonic sensors (software
routine) and bumpers on the LabMate (hard wire). The on-board computer holds all necessary
information for path planning and programs for different robot tasks, but can, if
necessary, receive new information via a wireless modem.
C. programming and Control
As with the RAID workstation, the WALKY mobile robot system is design to integrate with
the users´own input devices, such as voice control or mouse emulation devices. These are
normally connected to a computer via a serial line (COM-port on a PC) and these are
therefore not used by the system for the purpose of interfacing with user devices.
In most cases, it is assumed that all working positions, equipment, walls, etc., are
fixed and a map is created which is similar to simple objects (rectangles, circles, etc.),
in a CAD-system. When the user wants to tell the system where to go, he or she picks a
location on the map on the screen(assuming the system knows where it is) and invokes a
path planning routine to generate a path between the two locations. In general, each
object is associated with paths around it, which will be evaluated through a search
routine to check if there is an object in between the start and stop locations. The method
used is a combined depth-first and breath-first search and picks the best-first solution
to the problem. In case of unknown obstacles during run-time, a local pathplanning routine
will take over and either guide the system back to an earlier position or around the
obstacle. The pathplanner can be overridden by manually inserting the solution on the map
as via-points.
Investigations at laboratories connected to the academic hospital in Lund shows that
there exist several possible workplaces which are suitable for WALKY. The different tasks
have been analyzed and simulated for robot trials. The path planner for the mobile base is
tested in an environment similar to the final one. In order to cope with nonfixed objects,
suchas chairs, boxes, etc., the mobile base is equipped with a set of eight ultrasonic
sensors for reactive planning. By utilizing the existing eight ultrasonic sensors in
different configurations, trials on wall following and detection of various obstacles
(table leg, chair, book shelf, etc.) show it is possible, in a partly known environment,
to use ultrasonic sensors for collision avoidance and guiding the mobile system. Results
from trials show that small objects on the floor, doorsteps, table edges, etc. are
difficult to detect. Therefore, it is necessary to increase the number of sensors in order
to ensure a reliable system. Trials will be made at laboratories during 1995, and further
developments will be defined out of the results from these trials. Future work will be
directed toward increasing the level autonomy for more unstructured environments, such as
home-based activities, and two or more disabled sharing the same robot station for
vocational tasks similar to those described for the RAID station.
Rehabilitation robotics is an emerging field with many connections to service robotics.
However, special attention must be paid to the specific needs of individual users and
their physical handicaps. Thus, every individual case must be studied with care in order
to design and build a system which can be utilized by the user in an efficient way. As
described in this paper, much research has been devoted toward mechanical design,
including mobility and end-effectors, as well as programming and control. Much of this
work is based on experience from industrial robotics. Although results are promising, it
is important to recognize the need for research and developments which are free from the
influences of industrial robotics and instead to look for functional specifications within
service and rehabilitation robotics and how these can be transformed into technical
solutions. This work, which is a part of new research currently underway a Lund
University, will include advances in robotics design, including the use of reinforced
composite materials and event-based programming with a model representation to generate
autonomous functionally. The utilization of such systems for rehabilitation and their
personal service to humankind may well be the starting point of a revolution similar to
the one which began the PC computer came on the market.
REFERENCES
[1] TIDE, "1993-1994 workplan", Commission of the European Communities, 1993.
[2] S.A. Napper and R.L. Seaman, "Applications of robots in rehabilitation", Robotics Autonom. Syst., vol. 5, pp. 227 - 239, 1989.
[3] J. Hegarty, "Rehabilitation robotics: The user´s perspective", in Proc. 2nd Cambridge Wkshop. Rehab. Robotics, Cambridge, U.K., Apr. 1991.
[4] M. van der Loos, J Hammel and L. Leifer, "DeVar transfer from R&D to vocational and educational settings", in Proc. 4th Int. Conf. Rehab. Robotics, Wilmington, DE, pp. 151 - 155, June 14 - 16, 1994.
[5] T. Öderud, J.E. Bastiansen and S. Tyvand, "Experiences with the MANUS wheelchair mounted manipulator", in Proc. ECART 2, Stockholm, p. 29.1, May 26 - 28, 1993.
[6] A. Casals, R. Villá and D. Casals, "A soft assistant arm for tetraplegics", in Proc. 1st TIDE Congr.,Brussels,pp 103 - 107, and Studies in Health Technol. Inform., vol 9, IOS Press, Apr. 6 - 7, 1993.
[7] R.D. Jackson, Robotics and its role in helping disabled people", Eng. Sci. Educ. J., pp 267 - 272, Dec. 1993.
[8] J. Hennequin and Y. Hennequin, "Inventaid, wheelchair mounted manipulator", in Proc. 2nd Cambridge Wkshp. Rehab. Robotics, Cambridge, U.K., Apr. 1991.
[9] S.Greenhill,, "The digit Muscle", Ind. Robot, vol. 20, no. 5, pp. 29 - 30, 1993.
[10] S Levine, D. Bell and Y. Koren, "An example of a shared-control system for assistive technologies", in Proc. 4th Int. Conf. Comput. for Handicapped Persons, Vienna, Austria, Sept. 1994.
[11] J.-M. Détriché and B. Lesigne, "The robotized system for helping the disabled persons", in Proc. IARP 2nd Wkshp. Med. Healthcare Robotics, Newcastle, U.K., pp. 19 - 41, 1989.
[12] J.L. Dallaway, R.M. Mahoney, R.D. Jackson, and R.G. Gosine, "An interactive robot control environment
for rehabilitation applications", Robotica, vol. 11, pp. 541 - 551, 1993.
[13] D.S. Lees and L.J. Leifer, "Experimental evaluation of a graphical programming environment for service robots", in Proc. 4th Int. Conf. Rehab. Robotics, Wilmington, DE, pp. 19 - 23, June 14 - 16, 1994.
[14] H. Neveryd and G. Bolmsjö, "Mobile robot system for disabled", in Proc. ECART 2, Stockholm,
[15] ---------, "WALKY, a mobile robot system for the disabled", in Proc. 4th Int. Conf. Rehab. Robotics, Wilmington, DE, pp. 137 - 141, June 14 - 16, 1994.
[16] G. Nikoleris, K. Brink, M. Olsson and G. Bolmsjö, "Simplified programming of robots-Sport", in Proc. Robotics Wkshp., Linköping, Sweden, p. B2, June 2 - 3, 1993.
[17] M. Lindström, "Intelligent control of a mobile robot base", Master´s thesis, Dep. Production and Materials Eng., Lund University, Lund, Sweden, May 1993.
[18] C. Danielsson and L. Holmberg, "Evaluation of the RAID workstation", in Proc.
RESNA94´Annu. Conf., Nashville, TN, pp. 451 - 453, June 17 - 22, 1994.
Gunnar Bolmsjö received the M.Sc. and Ph.D. degrees in mechanical engineering
from Lund University, Lund, Sweden, in 1981 and 1986, respectively.
During 1986, he was at the Department of Production and Materials Engineering, Lund
University, and since 1987, he has been a Professor of Robotics at the same department.
His research interests include robotics in rehabilitation and industrial processes, such
as arc welding and grinding. Research includes control structures, including task level
programming, multisensor feedback, process models, and simulation.
Håkan Neveryd received the M.Sc. degree in mechanical engineering at Lund
University and a mechanical engineering at Lund University and a teacher degree at Växjö
University, Sweden.
He has worked as a Teacher at technical colleges in the South of Sweden. Since 1990, he
has been working at CERTEC at Lund University within the area of robotics and disabled
people. Since 1993, he has been the Director of CERTEC.
Håkan Eftring received the M.Sc. in mechanical engineering at The Royal
Institute of Technology, Stockholm, Sweden, in 1985, and received the licentiate degree of
engineering in manufacturing systems in 1990 at the same institute.
Between 1985 and 1992, he was working at The Royal Institute of Technology and The Swedish Institute of Production Engineering Research within the area of CAD/CAM, robotics, and FMS technologies. Since 1992, he has been working at CERTEC at Lund University with robotics, end effectors, and sensors in rehabilitation robotics.