A new challenge - robotics in the rehabilitation of the neurologically motor impaired.

A New Challenge--Robotics in the Rehabilitation of the Neurologically Motor Impaired

Many people suffer disruption of sensory or motor pathways that results in chronic motor impairment with loss of function of one or more limbs. This loss of function can lead to an inability to carry out many of the normal activities of everyday living. Physical therapists have traditionally met the challenge of treating neurologically impaired patients with motor disabilities by providing a mixture of exercise, to develop muscle strength and endurance, and skill training, to build up sensorimotor integration and coordination. Such therapy may involve the supervised use of machines that provide controlled exercising of single joints (for strength and mobility [1]) or of multiple joints (for coordination [2]). A significant new development in this area is the use of sensors for providing augmented feedback [3] or for tailoring the level of exercise to a patient's movement capability. [4] A further potential benefit of such equipment is that it can provide objective records of the progress of individual patients.

The application of technology to rehabilitation extends beyond equipment that assists in the restoration of motor function to devices that are capable of substituting for a patient's impaired movement if it is thought that further improvement in the patient's motor function is unlikely. The aim is to confer as much independence on the patient as possible so that control is typically the responsibility of the patient, rather than the therapist. This responsibility usually entails the patient using controls that regulate external power sources by remote means. Thus, for example, a powered wheelchair has been developed for control with minimal effort using a miniature joystick that signals left-right and forward-backward movement. [5] A mechanical arm with electric motors driving each joint has also been designed for control by a selector switch linked to a menu-drive computer display. [6]

In substituting for impaired function, the user's development of skill is likely to benefit if fine details of movement execution are delegated to the machine and the user can concentrate on the goals of action. [7,8] In providing automatic control of machine movement, as in therapy devices, sensor systems are becoming increasingly important. Nisbet et al, [9] for example, described the use of a wheelchair (ie, CALL Centre smart wheelchair) that is equipped with sensors and a microprocessor that can detect the proximity of potential obstacles and take automatic avoiding action. Another example of the use of sensors for automatic correction is a mechanical prosthetic hand equipped with acoustic sensors that respond to any slipping of a held object. [10] This information is fed back to a microprocessor that is programmed to compensate for the slip by a small increase in grip force.

Devices that move or manipulate things under the control of a programmed sequence of instructions constitute robots. The traditional use for a robot is in industrial assembly operations. A robot is defined by the Robot Institute of America as "a reprogrammable multifunctional manipulator designed to move material, parts, tools or specialized devices through variable programmed motions for the performance of a variety of tasks." [11](p1950) However, perhaps the most significant development in the field of robotics, not captured by this definition, is the use of sensors that permit robots to operate in an uncertain environment. The development of a capacity in robots to cope, if required, in autonomous fashion with unexpected events will be a crucial element in developing robotic domestic assistants that will require only intermittent supervision.

Many neurologically impaired patients permanently lose the ability to perform various necessary everyday functions. This loss has to be made up by providing either human assistance or assistance from some other source including, possibly, robots. With their remaining physical abilities, many patients might be able to command a robot to carry out those tasks that they themselves would be unable to perform unaided. Such use of robots holds the potential of offering patients the advantages of increased independence, the feeling of control over their environment, and vocational possibilities at the same time as being cost-effective in that such use could decrease the need for assistance from other people.

Given the potential benefits that the development of suitable robots might bring to the disabled, and the current extensive use of industrial robots, it might be thought that the technology would be very easily transferred by engineers to realize a rapid application of robotics in rehabilitation. However, the requirement that such robots work in contact with people and under supervisory control places them in contrast to most robots in industrial settings, in which the aim is to minimize human intervention. In rehabilitation, the most important issue becomes the relation between the user and the robot (ie, the human-machine interface [HMI]).

There are published reports on the development of robotic devices for use by motor-impaired individuals that touch on the HMI issue. Many of these reports have appeared in the series of proceedings of meetings held under the auspices of the Association for the Advancement of Rehabilitation Technology (RESNA). (*1) However, a survey of these reports reveals two, possibly related, problems. The first concern is that there is a tendency to treat each project in terms of specifics, such as the controls or display for a particular machine, with little attempt to formulate findings in terms of principles that might be generalized to systems based on different hardware. With rapidly developing technology, there is the risk of repeatedly making the same mistakes if questions are not formulated in sufficiently general fashion. The second concern is that in the reported research, there is often an apparent lack of input regarding the user's behavior. Therapists and experimental psychologists are trained to observe and analyze patterns of behavior in ways that could be expected to provide insights that the system designer might miss or might not be disposed to investigate in a rigorous manner. Ergonomists concerned with human factors could be expected to make a valuable contribution in this as well as other areas of rehabilitation. [12]

In this article, we take the position that work on robotics has the potential to make such an important contribution to rehabilitation that it deserves discussion from a broad, overview perspective rather than in terms of the progress made on local research projects. Our thesis is that the greatest progress will be made by taking an integrated approach that not only takes account of the engineering challenges to meet the nature of the motor impairment, but also recognizes the full psychological environment of the motor-impaired user of a robotic device. We first briefly summarize the nature of robots and consider some of the issues raised by allowing people to intervene in their control. We then turn to the use of robots by patient groups with constrasting needs. Finally, we consider problems in evaluating benefits and assessing equipment safety.

Robots

Robot manipulators come in many different shapes, according to the environment within which they must work and the movements required of them (Fig. 1). The various links and joints that make up the arm and gripper of a robot may be more or less anthropomorphic. For example, in a project involving very young, developmentally delayed children, Hoseit et al [13] described the robot they used as moving like a human arm in its articulation. The robot had five joints between its base, body, upper arm, forearm, hand, and two fingers. The particular movements allowed by the joints defined six degrees of freedom for this particular robot, such that it could rotate on its base, extend and flex at the shoulder and at the elbow, extend and flex ("pitch") and supinate and pronate ("roll") at the wrist, and open and close the hand. (The rather limited hand function reflects a trade-off between the greater flexibility of use of a gripper modeled on the human hand and increased reliability of an end-effector that has fewer movable parts. In some cases, the limitations of the end-effector may be offset by endowing the robot with the ability to reconfigure itself by allowing it to switch between different grippers for different jobs).

Robots also vary widely in size and speed of movement. The larger and heavier robots used in industry can often move so fast that they pose a considerable danger to people working around them, and it is essential to close off the workspace when they are operating. In contrast, lightweight, portable robots, such as those designed for education, usually move relatively slowly and so pose little risk in the event of an unexpected collision. Robots used in rehabilitation are sometimes custom-made for a specific purpose. [14,15] However, the use of commercially available robots, perhaps in a modified form, is more common. Some research projects have involved the use of commercially available robots (eg, PUMA, UMI RTX) designed for industrial applications. For example, van der Loos et al [16] describe the use of PUMA, and Harwin et al [17] report on a project using a UMI RTX. Other projects have been based on smaller, educational robots such as the Microbot [18] or the Cobra. [19]

Any desired movement of the end-effector of a robot, in terms of either its location or its orientation in three-dimensional space, involves adjusting the configuration of its links using the motors that drive each joint. Determining how the angle of each joint should be changed relative to all the other joints to achieve a given repositioning of the end-effector is known as the "inverse kinematics problem." Solving the inverse kinematics problem is not a trivial task and entails extensive computation by the computers associated with any robotic system. Moreover, in an unpredictable environment it is possible, for example, for links to be unexpectedly perturbed so that the motors do not exactly reproduce required changes in joint angle. For reasons such as this it is desirable that the robot receive feedback from a suitably distributed array of sensors. Some of these sensors may provide feedback about individual joint angles. For object manipulation, however, information about the force being exerted on the object can be particularly useful. In general, the processing of information from multiple sensors cannot be handled in an isolated fashion, but must be integrated (in a process often termed "sensor fusion") in order, for example, to resolve conflicting sources of information. This integration may require relating sensory input to a preexisting knowledge base or to information derived from a vision-based analysis of the general scene. It also entails considerable computation.

User Interface

In developing computer programs to control a robot, it is perhaps unreasonable to expect a software engineer to envisage all possible ways of using the robot's manipulator. It would certainly be nice if the user could give instructions to a robot in terms of higher-level goals for action, such as "get some water," or even if the robot could respond to implied instructions, such as "I'm thirsty," and the user could delegate control of the fine details of requisite movements to the robot. However, it is realistic to expect there will be occasions when direct, moment-to-moment control is required over the individual movements of a robot. [20] The user may possess "privileged" information such as bodily sensations derived from contact with the robot when the robot is being used to accomplish some aspect of personal care. It is then appropriate to let the user specify the robot's movements, possibly with a confirmatory check by the robotic system before the movement is implemented. [21] Subjecting a robot to operator intervention moves the focus toward the nature of the user's controls and the presentation to the user of information about the remote-controlled device. We now turn to a brief overview of some of these considerations. For a more extended discussion, the reader should refer to a review of the growing literature on remote-controlled (teleoperated), semiautonomous devices, such as that of Sheridan. [22]

In order to develop a set of controls suited to the physical capacities of any particular user, many choices must be made. n[23] These choices include the number of controls, the range of the control (for example, whether it operates over a continuum of values or whether it switches between discrete values, the latter being more resistant to inaccurate movement), and the physical dimension along which the control varies (for example, whether it registers force or position). Another important consideration is the relation of the control to the controlled variable on the robot. In general, it is found that controls are easier to use if the spatial relations of the control. correspond as nearly as possible to those of the controlled variable. [24] However, it is worth noting that control of individual joint movements is not generally to be recommended. Control defined in terms of end-effector position and orientation, which may be viewed as higher level, is usually found to be easier for both the physically handicapped and the able-bodied user. [25,26]

The selection of suitable controls from a set of options, such as those shown in Figure 2, is important. Two commonly encountered classes of problems arise from the selection of controls that are awkward to use. The first class of problems concerns performance. Controls that are difficult to use are likely to increase the time required for the robot to carry out a task and thus to render its performance less accurate. An inaccurate performance by the robot could create an unsafe situation in which the user is at risk or there is potential damage to the robot or objects in the environment. The second class of problems caused by unsatisfactory controls concerns comfort. If the disabled person is very limited in his or her manner of movement and excessive use is required of the control, he or she might become fatigued or even suffer repetitive strain injuries. With this point in mind, and also given that variety of exercise is desirable in combating the decreases in bone density, muscle length, muscle strength, and joint integrity that often accompany disuse, it may be beneficial if the controls are chosen to draw on the disabled user's remaining range of movement. This consideration may be seen as contrasting with the common preoccupation in designing controls for the nondisabled person, in which the goal often appears to be to minimize the amount of physical work required.

Because a computer is invariably required to control robot movement, it makes sense to use its information-handling capabilities to provide the user with a graphics display. One use for such a display might be to provide information on the current state of controls. Even if the user has unrestricted vision of the robot's actions, the various possible levels of interaction with the robot mean that the function of a finite set of controls is likely to keep changing. A display could serve as an external memory aid for the way each control relates to the set of actions that are currently available. The display might also be used to provide information to the user about the progress of the task at hand. This information might be at the level of actions completed so far, or, if the user is engaged in direct control of individual movements, it might be feedback in schematic terms about the disposition of the robot arm. Another use for a display arises if movement sequences can be programmed by the user and saved in the computer memory for later execution. The display is then required for symbolic manipulations to allow programming of the action sequences. Alternatively, a schematic view of the manipulator might be shown in the "teach" mode, in which the user can set up a sequence of positions for subsequent use. [27] In either case, such programming functions would be assisted by allowing a symbolic "preview" display.

In selecting a set of controls and determining the nature of the display, the system designer should be aware that there may be interactions with the user's intellectual organization of the activity. There is a distinction in mechanical terms between the movements required to use a keyboard and those required to use a mouse, such that control sequences may have more parallel elements (in the case of simultaneous keystroke combinations on a keyboard) or be more serially organized (in the case of sequential moves of the mouse). This distinction may influence the way people think or remember how to interact with the system. [28,29] A goal for design would therefore be that, with developing familiarity, using a robotic system should become part of one's way of thinking. The system would then be more likely to be thought of as an extension of one's self. It would seem reasonable to suppose that natural language styles of communication with a robot, rather than selection from a formal, constrained command set, would facilitate this goal. [30] Because speech is already a commonly used mode of asking others for assistance, speech would certainly appear to be a natural choice for robot control. There is growing expertise in automatic speech recognition for the disabled, [31] and it is certainly encouraging to see the appearance of at least one commercially available, voice-controlled robotic work station. [32]

Patient Differences

The range of motor impairments that may be seen in neurologically impaired patients is very great. Both the nature and the extent of the impairment may vary. Cerebellar damage can result in incoordination without slowing or weakening of voluntary movements. Weakness and fatiguability are often associated with damage to the corticospinal pathways. When damage to the corticospinal paths arises as a result of a stroke affecting the middle cerebral artery, the motor deficits are often accompanied by damage to sensory pathways. Because proprioceptive disturbances, by themselves, disrupt fine-precision control, this damage can be particularly disabling.

The neural damage resulting from a stroke is almost invariably restricted to one side of the brain. With the resulting deficits in movement lateralized to the opposite side of the body, the "recovery" of a stroke patient may be a matter of learning to substitute use of the arm or leg o nthe uninvolved side. However, such substitution is not possible if the disease process results in functional deficits on both sides of the body. Thus, for example, bilateral lesions of the basal ganglia occurring in Parkinson's disease result in symptoms (which can include slowness of movement and fatiguability) that are typically bilateral. With damage to the spinal cord (usually resulting from mechanical trauma), the interruption of sensory and motor paths to and from the brain results in symmetric loss of function of all parts of the body innervated by nerves branching distally with respect to the point of damage.

Clearly, the provision of robotic assistance for patients with motor impairment must be adaptable to take account of the extent of involvement (unilateral/bilatera, lower/upper limbs) and its nature (incoordination, weakness. The determination of what would be appropriate to provide a patient should not be governed only by the physical requirements. Account should also be taken of the patient's own perceived needs, perhaps in the form of a "wish list." [14] Even though all items on the list might not be achievable, the list could serve the further purpose of providing a basis for any concurrent psychological counseling. Fengler and Cameron [33] have reported realistic expectation in a preliminary study of five severely disabled new users of a menu-driven robot. In planning the potential role for robotic aid in a management program for a particular patient, account must also be taken of whether the functional deficits are likely to remain stable or to deteriorate progressively or whether there is a likelihood of improvement through recovery. We will now discuss the contrasting needs of different patients for each of these conditions.

Stable Conditions

Because of the profound loss of movement that is directly related to the level of the lesion, patients with spinal injuries are the group most commonly encountered in the application of robotics to rehabilitation. The lesions of the spinal cord secondary to such injuries result in stable functional deficits. To overcome these deficits, robots have been used for a variety of activities of daily living as well as for leisure and vocational tasks. Seamone and Schmeisser [34] described their evaluation of 20 quadriplegic adults' use of a robot controlled either by movement of individual joints or through preprogrammed motion sequences. Among the tasks assayed was picking up the telephone and holding it to the ear and using a spoon to eat from a bowl. Hall et al [35] report a study of high-level quadriplegic patients who used a voice control over a robot with preprogrammed sequences for tasks such as washing the face, feeding, and operating a two-way speaker phone. Both studies used table-mounted robots in a work-station context in which the robot was fixed in a specially designed desklike environment. However, because the spinal-injured patient is invariably wheelchair bound, the wheelchair provides an obvious platform for mounting a robot with movements that are consistent with, and easily related to, the body. [19,36] Such a mounting also has the advantage of not restricting the user to a particular location (as is the case with table-mounted robots), which may have an isolating effect on the individual.

Progressively Deteriorating

Conditions

In patients whose motor impairment increases progressively (eg, patients with motoneuron disease, Duchenne's muscular dystrophy, or multiple sclerosis), the degree of assistance required will increase with time. Simultaneoulsy, the types of usable control will become more constrained; merely increasing the sensitivity of a given control may not be sufficient. Rather, a different type of control may be needed, one that recognizes the patient's decreasing capacity for numerous or complex control movements. At first sight, increasing limitations on movement might appear to conflict with the need for greater assistance. However, it is exactly this situation in which a robot, able to learn and adapt to a user's requirements and also offering a varying hierarchy of control, might be especially useful. As the patient's skill develops, more complex actions might be offered as command options, whereas if offered earlier, these actions would only have confused an unskilled user. Naturally, an important issue for this approach is the extent to which changing to command interface can be met so that the user perceives consistency of usage. A balance of responsibility for identifying when to make the changes must also be apportioned between therapist, patient, and even perhaps the robot itself.

Improving Conditions

A major goal of physical therapy intervention in conditions such as stroke or cerebral palsy, in which the patient is expected to improve, is the relearning of normal movement while preventing secondary changes from disuse. The patient should be an active participant in attempting to achieve challenging, realistic goals; practicing with accuracy and consistency; receiving feedback about performance; receiving manual, verbal, and visual guidance when necessary; and trying to generalize newly learned skills to his or her everyday environment. [37] In terms of rehabilitation robotics, using external means to replace lost function is unlikely to be the aim as long as the therapist judges that function is continuing to improve. However, in such cases, there is a potential role for robots as training devices. This role is illustrated by a pilot study of Kristy. [38] A robot arm was placed in front of the patient, and a light on the gripper switch of the robot arms was used to indicate different positions to which a patient should reach. Immediate feedback was provided by the light going off when touched, and quantitative information regarding the speed and accuracy of the movement was recorded by the robotic system to provide the patient with knowledge of results. It should be noted that no knowledge of qualitative aspects of the movement was provided; the patient was required to monitor this. Nonetheless, the subjective reports of therapists and patients in this study indicated the robotic system was helpful as a treatment tool.

So far, our emphasis in this article has been on adult rehabilitation. Most of what has been stated assumes a level of intellectual functioning sufficient to control a semiautonomous robot. Although it might be supposed that with suitable training most patients would be able to learn the necessary skills, it is conceivable that difficulties in using a robot might arise, for example, in people who are disabled from birth. These individuals may lack perceptual abilities that are dependent on having had active experience with objects in the real world. However, one could take the contrasting view, that a robot manipulator might provide just the means for multiply handicapped individuals to build up a better cognitive representation of the world. It is a widely held belief that cognitive development is based on a child's active manipulation of the physical environment. Work by Hoseit et al [13] with very young (chronological age, [is less than] 36 months), developmentally delayed children suggests a robot can provide the means to motivate them to explore the environment and to facilitate understanding (eg, of cause-and-effect relationships). A subsequent report from the same group of researchers [34] indicated that children can become interested in using a robotic arm, as long as the actions being performed directly involve the child (eg, bringing an object to them) and allow the child to appreciate the relationship between switch activation and resultant behavior. However, for the children with serious intellectual delay, it may be appropriate to limit the degree of flexibility afforded by the controls. [40]

Another, more comprehensive approach to using a robot as a teaching aid for older, severely physically handicapped children is described by Gosine et al. [41] The applications developed by these authors may be summarized under six categories of activity: early developmental education (teaching concepts such as color, size, sequence), interactive game play (for example, fox and hounds), painting, interactive chemistry experiments, cooking, and feeding. On the basis of informal observations, the authors concluded that the active participation afforded by the robot facilitated learning and warranted further development of the approach. More detailed behavioral observations from another study, [42] in which a 13-year-old cerebral-palsied child was provided with a low-cost robotic arm for feeding, suggest further potential benefits in the area of movement control. Over a period of a month, the child learned to use the robot to feed himself in the time it had previously taken his mother to carry out the same task. More significantly, improvements were observed in hand and head movement control, with a decline in the frequency of involuntary movements. In further developments in this area, an important issue will be the appropriate balance between the robot providing automated execution of component tasks as opposed to allowing the child direct access over the robot's movements, with the possible benefit of "learning by doing." [43]

Performance Evaluation 

Questioning the extent to which a disabled person's needs are being met by robotic assistance must be fundamental to any development program for rehabilitation robotics. Perhaps the most obvious means to assess robotic systems in rehabilitation is to ask the user to provide an evaluation, using ratings in questionnaires and the like. Hammel et al [44] report satisfaction among a group of high-level quadriplegic patients with a robot's performance of a number of tasks including brushing teeth and feeding. Significantly, these authors report that the users indicated a preference for the robot versus the family/attendant in performing these functions. (Elsewhere, [45] a more complex approach that seeks to determine the semantic structure underlying users' evaluative comments of robotic assistance has been outlined). In addition to soliciting the patient's comments, there is relevance in the therapist's observations, particularly if these observations are collected in a structured fashion. [46,47] The therapist's observations might be used to confirm the patient's observations. In addition, the therapist might be expected to be more attuned to the need to maintain muscle strength and endurance and to prevent changes associated with disuse. These changes may be exacerbated if the robot takes over functions that the patient can perform independently or with minimal assistance.

Although subjective report methods offer face validity to an evaluation exercise, their susceptibility to unintentional bias [48] should foster the increasing use of objective performance measures. The results of studies of human computer interaction with nondisabled populations have suggested that objective evaluations would be more widely used if a set of standard (benchmark) tasks were to be developed. [49] Indeed, the evaluation of telerobotic systems commonly involves performance assessment with a standard set of tasks such as Fitts' peg task, in which the time take to transfer pegs between cloe-fitting holes is recorded. This would be a useful approach to the evaluation of rehabilitation robotics, provided the tasks be designed with sufficient generality to be appropriate to a broad range of control-display arrangements. Ideally, the tasks would afford not only global outcome measures of performance, but also detailed measures of coordination and fluency in using the robotic system. With either type of measure, it will be important in future work to support claims made of the useability of particular systems with statistical tests that make due allowance for the inherent variability of human performance.

Another approach to evaluation in this area that would be worth considering in rehabilitation robotics is the use of graphics simulation. In order to speed up development, the user controls a computer display that recreates the characteristics of a telerobot without the need to implement is hardware. [50] Indeed, the general point may be made that, given the difficulty in carrying out performance evaluation tests with the small numbers of patients available at research centers, more effort might be put into testing robot systems with nondisabled users. This approach may be seen as analogous to a study of the attention demands of using an artificial limb that was conducted using healthy subjects. [51]

A primary area for concern in any assessment of robot performance in rehabilitation applications must be safety, particularly in settings in which the manipulative devices are designed to function with the use within its range of actions. [52] A disabled person may be critically dependent on the continued functioning of a robot; its failure might be dangerous. The disabled population may be less able to take evasive action and are likely to be using the robot to perform tasks that involve them directly, such as transporting hot liquids to their mouth. Various strategies are currently being developed to reduce risks in such situations. Lees et al, [21] for example, describe a desk-top robotic assistant that moves in small, incremental steps as it approaches the user's face. Possible sources of hazard are control errors (by the computer software); human errors; mechanical hazards; or problems with the electronics, hydraulics, or pneumatics of the system. [53] Clearly, it woul be desirable for there to be some consideration of standards for mobile robots by the International Organization for Standardization. [54]

Conclusions

Rehabilitation is aimed ultimately at increasing a person's independence. This goal can be accomplished either by regaining the ability to move or by using external means to achieve movement. In this discussion of the application of robots to the rehabilitation of neurologically impaired patients, two roles for robots have been considered. First, the robot can serve as a training aid, in which the person is regaining the ability to move. Second, the robot can be a substitute for human movement. Research is well underway into investigating how robots might be implemented in rehabilitation in these two ways. Most of this research has been initiated by the designers of the robots; to date, the involvement of those who might contribute understanding to the end- users' problems, such as physical therapists and psychologists, has been somewhat lacking. When there has been such involvement, it has tended to be at later, operative stages, in which the role is one of advising on minor alterations to a robot to fit the user, rather than as a key input to the initial design. We consider it imperative that those who understand patients' problems actively participate in and provide direction for the application of robots to rehabilitation. We believe such initiatives would contribute significantly to ensuring that the patients' interest remain the primary consideration.

Acknowledgments

We acknowledge the comments of Susan Goodrich, Ray Gosine, Patrick Haggard, William Harwin, and Robin Jackson.

(*1) RESNA Press, 1101 Connecticut Ave NW, Ste 700, Washington, DC 20036.

References

[1] Isolated Joint Testing and Exercise: A Handbook for Using Cybex I and the UBXT. Ronkonkoma, NY: Cybex, Div of Lumex, Inc; 1983.

[2] Hocherman S, Dickstein R, Pillar T. Platform training and postural stability in hemiplegia. Arch Phys Med Rehabil. 1984;65:588-592.

[3] Winstein CJ, Gardner ER, McNeal DR, et al. Standing balance training: effect on balance and locomoation in hemiparetic adults. Arch Phys Med Rehabil. 1989;70:755-762.

[4] Godall RM, Pratt DJ, Rodgers CTG, Murray-Leslie CM. Enhancing postural stability in hemiplegics using externally applied forces. Int J Rehab Res. 1987:10:132-140.

[5] Wilshere ER. Equipment for the Disabled: Wheelchairs. 5th ed. Oxford, Englad: Oxfordshire Health Authority; 1982.

[6] Hillman MR, Orpwood Rd. A robot workstation for the disabled. In: Proceedings of the First International Workshop on Robotic Applications in Medical and Health Care. Ottawa, Ontario, Canada: National Research Council; 1988:7.1-7.6.

[7] Johnson P. The acquisition of skill. In: Smyth MM, Wing AM, eds. The Psychology of Human Movement. London, England: Academic Press Inc (London) Ltd; 1984: chap 8.

[8h Gentile AM. Skill acquisition: action, movement and motor processes. In: Carr J, Shepherd R, eds. Movement Science: Foundations for Physical Therapy in Rehabilitation. London, England: William Heinemann Medical Books Ltd; 1987:93-117.

[9] Nisbet PD, Odor JP, Loudon IR. The CALL Centre smart wheelchair. In: Proceedings of the First International Workshop on Robotic Applications in Medical and Health Care. Ottawa, Ontario, Canada: National Research Council; 1988:9.1-9.10.

[10] Kyberd PJ, Chappell PH, Nightingale JM. Sensory control of a multifunction hand prosthesis. Biosensors. 1988;3:347-357.

[11] Knight AL. Robots and automatic production machinery. In: Parmeggiani L, ed. Encyclopaedia of Occupational Health and Safety, Volume I. 3rd ed. Geneva, Switzerland: International Labour Organisation; 1983:1950-1952.

[12] Rice VB, Sind PM. The role of human factors in rehabilitation. Human Factors Society Bulletin. 1990;33:4-5.

[13] Hoseit P, Liu KM, Cook AM. Development and use of a robotic arm system with very young, development delayed children. In: Foulds R, ed. Interactive Robotic Aids -- One Option for Independent Living: An International Perspective. New York, NY: World Rehabilitation Fund; 1986:48-50.

[14] Cameron W. Manipulative appliance development in Canada. In: Foulds R, ed. Interactive Robotic Aids -- One Option for Independent Living: An International Perspective. New York, NY: World Rehabilitation Fund; 1986: 24-28.

[15] Hillman MR. Development of a robot workstation for the disabled. In: Proceedings of the First International Workshop on Domestic Robots and Second International Workshop on Medical and Health Care Robotics. London; England: Department of Trade and Industry; 1989:355-359.

[16] van der Loos HFM, Michalowski SJ, Leifer LJ. Design of an omnidirectional mobile robot as a manipulation aid for the severely disabled. In: Foulds R, ed. Interactive Robotic Aids--One Option for Independent Living: An International Perspective. New York, NY: World Rehabilitation Fund; 1986:61-63.

[17] Harwin WS, Ginige A, Jackson RD. A potential application in early education and a possible role for a vision system in a workstation-based robotic aid for physically disabled persons. In: Foulds R, ed. Interactive Robotic Aids--One Option for Independent Living: An International Perspective. New York, NY: World Rehabilitation Fund; 1986:18-23.

[18] Anderson LL. Use of a small robotic arm in the workplace by a severely physically disabled individual. In: Proceedings of the 8th Annual Conference on Rehabilitation Technology. Memphis, Tenn: RESNA Press; 1985:7-9.

[19] Zeelenberg AP. Domestic use of a training robot-manipulator by children with muscular dystrophy. In: Foulds R, ed. Interactive Robotic Aids--One Option for Independent Living: An International Perspective. New York, NY: World Rehabilitation Fund; 1986:29-33.

[20] Gosine RG, Harwin WS, Jackson RD. CURL: A programming environment for a rehabilitation robot. In: Proceedings of the First International Workshop on Robotic Applications in Medical and Health Care. Ottawa, Ontario, Canada: National Research Council; 1988: 6.1-6.10.

[21] Lees DS, Crigler R, van der Loos HFM, Leifer LJ. A third generation desktop robotic assistant for the severely physically disabled. In: Proceedings of the International Conference of the Association for the Advancement of Rehabilitation Technology (ICAART 88). Washington, DC: RESNA Press; 1988:450-451.

[22] Sheridan TB. Telerobotics. Plenary presentation for the 10th IFAC World Congress on Automatic Control; July 27-31, 1987; Munich, Federal Republic of Germany.

[23] Dymand EA, Potter R, Griffiths PAG, McClemart EJW. A week in the life of Mary: the impact of microtechnology on a severely handicapped person. J Biomed Eng. 1988;10: 483-490.

[24] Poulton EC. Tracking Skill and Manual Control. New York; NY: Academic Press Inc; 1974.

[25] Wallersteiner U, Stager P. Lawrence P. A human factors evaluation of teleoperator hand controllers. In: Mason CA, ed. Teleoperation and Control: Proceedings of the International Symposium (IFS Publications). Berlin, Federal Republic of Germany: Springer-Verlag; 1988:291-296.

[26] Kwee HH, Duimel JJ, Smits JJ, et al. The Manus wheelchair-borne manipulator: system review and first results. In: Proceedings of the First International Workshop on Domestic Robots and Second International Workshop on Medical and Health Care Robotics. London, England: Department of Trade and Industry; 1989:385-393.

[27] Apkarian J, Naumann S, Masanic C, Milner M. Proceedings of the First International Workshop on Domestic Robots and Second International Workshop on Medical and Health Care Robotics. London, England: Department of Trade and Industry; 1989:361-370.

[28] Buxton W. Chunking and phrasing and the design of human-computer dialogues. In: Kugler H-J, ed. Information Processing 86. Amsterdam, the Netherlands: Elsevier Science Publishers BV; 1986.

[29] Buxton W. There's more to interaction than meets the eye: some issues in manual input. In: Norman DA, Draper SW, eds. User-Centered System Design: New Perspectives on Human-Computer Interaction. Hillsdale, NJ: Lawrence Erlbaum Associates Inc; 1986: 319-337.

[30] Michalowski S. Crangle C, Lian L. Experimental study of a natural-language interface to an instructable robotic aid for the severely disabled. In: Proceedings of the 10th Annual Conference on Rehabilitation Technology. San Jose, Calif: RESNA Press; 1987:773-775.

[31] Noyes JM, Haigh R, Starr AJ. Automatic speech recognition for disabled people. Applied Ergonomics. 1989;20:293-298.

[32] Rash W. A helping hand. Byte. December 1989:129-130.

[33] Fengler M, Cameron WM. Clinical testing of a low cost robotic arm for the severely physically disabled. In: Proceedings of the First International Workshop on Robotic Applications in Medical and Health Care. Ottawa, Ontario, Canada: National Research Council; 1988:10.1-10.3.

[34] Seamore W, Schmeisser G. Evaluation of the APL/JHU robot arm workstation. In: Foulds R, ed. Interactive Robotic Aids--One Option for Independent Living: An International Perspective. New York, NY: World Rehabilitation Fund; 1986:51-53.

[35] Hall K, Glass K, Hammel J, et al. Evaluation of a tabletop robotic aid for quadriplegics. Proceedings of the 10th Annual Conference on Rehabilitation Technology. San Jose, Calif: RESNA Press; 1987:776-777.

[36] Kwee HH, Duimel JJ, Smits JJ, et al. The Manus wheelchair-borne manipulator. Proceedings of the First International Workshop on Robotic Applications in Medical and Health Care. Ottawa, Ontario, Canada: National Research Council; 1988:6.1-6.10.

[37] Carr J, Shepherd R. A Motor Relearning Program for Stroke. London, England: William Heinemann Medical Books Ltd; 1987.

[38] Kristy Ka. Use of robotic arm as a rehabilitation exercise partner: a therapeutic enhancement. In: Proceedings of the International Conference of the Association for the Advancement of Rehabilitation Technology (ICAART 88). Washington, DC: RESNA Press; 1988: 452-453.

[39] Cook AM, Hoseit P, Liu KM, et al: Using a robot arm system to facilitate learning in very young disabled children. IEEE Trans Biomed Eng. 1988;35:132-137.

[40] Karlan GR, Nof SY, Widmer NS, et al. Preliminary clinical evaluations of a prototype interactive robotic device (IRD-1). In: Proceedings of the International Conference of the Association for the Advancement of Rehabilitation Technology (ICAART 88). Washington, DC: RESNA Press; 1988:454-455.

[41] Gosine RG, Harwin WS, Jackson RD. Applications of a programmable robotic workstation using the CURL programming environment. In: Proceedings of the First International Workshop on Domestic Robots and Second International Workshop on Medical and Health Care Robotics. London, England: Department of Trade and Industry; 1989:43-50.

[42h Topping MJ, Hegarty JR. The potential of low-cost computerised robot arms as aids to independence for people with physical disability. In: Proceedings of the First International Workshop on Domestic Robots and Second International Workshop on Medical and Health Care Robotics. London, England: Department of Trade and Industry; 1989:303-307.

[43] Howell RD. Designing a robotically-aided science laboratory for physically disabled students. In: Proceedings of the Cambridge Workshop on Rehabilitation Robotics. Cambridge, England: University Engineering Dept TR78; 1989:45-49.

[44] Hammel J, Hall K, van der Loos HFM, et al. Clinical evaluation of a desktop robotic aid for severely physically disabled individuals. In: Proceedings of the International Conference of the Association for the Advancement of Rehabilitation Technology (ICAART 88). Washington, DC: RESNA Press; 1988:448-449.

[45] Valettas N, Goldenberg AA, Milner M. Clinical testing of the UT/HMMC robotic aid. In: Proceedings of the First International Workshop on Domestic Robots and Second International Workshop on Medical and Health Care Robotics. London, England: Department of Trade and Industry; 1989:51-58.

[46] van der Loos HFM, Michalowski SJ, Hammel J, et al. Assessing the impact of robotic devices for the severely physically disabled. In: Proceedings of the First International Workshop on Robotic Applications in Medical and Health Care. Ottawa, Ontario, Canada: National Research Council; 1988:1.1-1.6.

[47] Johnson GI, Clegg CW, Ravden SJ. Towards a practical method of user interface valuation. Applied Ergonomics. 1989;20:255-260.

[48] Poulton EC. Bias in Quantifying Judgments. Hove, England: Lawrence Erlbaum Associates Ltd; 1989.

[49] Baecker RM, Buxton W. The haptic channel. In: Baecker RM, Buxton W, eds. Readings in Human Computer Interaction: A Multidisciplinary Approach. Los Altos, Calif: Morgan Kaufmann Publishers Inc; 1987; chap 8.

[50] Bicker R, Burn K, Maunder L. The man-machine interface in remote manipulation. In: Mason CA, ed. Teleoperation and Control Proceedings of the International Symposium (IFS Publications). Berlin, Federal Republic of Germany: Springer-Verlag; 1988:151-157.

[51] Soede M. On the Mental Load in Arm Prosthesis Control. Leiden, the Netherlands; Nederlands Institut voor Praeventieve Gezondheidszorg-TNO; 1980.

[52] Kwee HH. Spartacus and Manus: telethesis developments in France and the Netherlands. In: Foulds R, ed. Interactive Robotic Aids--One Option for Independent Living: An International Perspective. New York, NY: World Rehabilitation Fund; 1986:7-17.

[53] Ottinger LV. Questions potential robot users commonly ask. Industrial Engineering. August 1982:28-33.

[54] Meieran HB, Blatt SE. Ergonomic and safety considerations for teleoperated robotic devices controlled by the handicapped and the elderly. In: Mason CA, ed. Teleoperation and Control: Proceedings of the International Symposium (IFS Publications). Berlin, Federal Republic of Germany: Springer-Verlag; 1988: 309-316.

P van Vliet, MSc, is Lecturer, School of Physiotherapy, Coventry Polytechnic, Priority St, Coventry, CV1 5FB, United Kingdom.

A Wing, Phd, is Senior Scientist, Medical Research Council Applied Psychology Unit, 15 Chaucer Rd, Cambridge, CB2 2EF, United Kingdom. Address all correspondence to Dr Wing.