RAID - a vocational robotic workstation
John L Dallaway and Robin D Jackson
Department of Engineering
University of Cambridge, UK
1 Introduction
The RAID (Robot for Assisting the Integration of the Disabled) project exists under the European Community TIDE (Technology for the socio-economic Integration of Disabled and Elderly people) initiative. RAID is aligned with Action Line 2.1 of the TIDE initiative, entitled 'Environmental Control and Robotics'. Its aim is to develop and demonstrate a prototype robotic workstation and is primarily targeted at office environments. Initial trials will concentrate on Computer Aided Design.
RAID draws upon existing technologies which have been identified through an extensive State of the Art Survey and intends to customise these where appropriate to form a reliable system within the time constraints of the TIDE pilot phase. Further research in critical areas will provide the stimulus for system refinements in future TIDE phases.
The following organisations are associated with RAID:
- UMI Group, UK
- OxIM, UK
- CEA/DTA/UR, France
- Lund University, Sweden
- Armstrong Projects, UK
- Cambridge University, UK
- HADAR, Sweden
- Equal Design, France
- Papworth Group, UK
The diversity of experience reflected in the consortium ensures a balanced perspective in all aspects of the work.
2 Background
The potential for robotic devices as aids for disabled people is widely recognised and has been previously demonstrated in the user trials of several existing prototypes [1, 2, 3, 4, 5]. However, relatively few prototypes have reached the commercialisation stage and these aids have not established their anticipated market. For example, the workstation developed by Boeing Computer Services [6] and subsequently marketed by PRAB Command used an RTX robot arm to manipulate reference books, change diskettes and retrieve hard copy output. The rehabilitation robotics community must learn from the development and marketing strategies of such products if it is to successfully turn the prototypes of today into the products of tomorrow.
It is the experience of the RAID consortium that projects which are led by the requirements of users have a substantially higher probability of success compared with those which seek to push technology into the marketplace. However, it is surprisingly easy to lose sight of user-led objectives during the specification of robotic aids. A major difficulty lies in the exposure of the general public to other forms of robotic device such as those of vehicle assembly lines and science fiction movies. Potential users must be re-educated in their understanding of robot capabilities if they are to participate effectively in the task definition stage of product design.
The user interface is another area which must be addressed if products are to be successful. It is interesting to note that the software company Microsoft has recently established a usability laboratory where computer users are monitored as they familiarise themselves with new software [7]. The mistakes and frustrations of the users are evaluated and fed back into the interface design process. The productivity increases apparent in the move to Graphical User Interfaces should be mirrored in the way robotic devices are directed.
3 Project structure
3.1 Overview
RAID uses the structure of a conventional development project. A survey into the current state of rehabilitation robotics and associated technologies was used to reveal the foundation of research on which the development would be based. The project specification was divided into the following parts, listed in chronological order:
- The user requirements specification
- The functional specification
- The technical specification
- The test specification
A conceptual layout for the workstation, based on the project specification, was then fed back to potential users for their analysis and comment. The prototype RAID system has been designed and is now under construction. On completion, the prototype will undergo user trials which will provide essential feedback in preparation for the process of commercialisation and the definition of future research and development.
3.2 State of the art survey
The papers and published data of rehabilitation robotics projects from 1975 to the present day were used to establish the state of the art relating to each of the following sub-systems:
- Robot arms
- Control systems
- Input devices
- User interfaces
- Sensors
- End effectors
Projects were additionally classified according to the fundamental nature of the work. The following levels were identified:
- Application enhancement (eg workstation layout)
- Software sub-system enhancement (eg task level control)
- Mechanical sub-system enhancement (eg pneumatic actuators)
- Peripheral enhancement (eg end effector functionality)
The technical evaluation of previous projects in this manner proved useful in preparing the technical specification by indicating which functional sub-systems were technically available within the limited time-scale of the RAID pilot phase.
The clinical trials relating to a number of projects were also evaluated. However, it was found that the trials investigated had not been founded on a rigorous functional analysis. Many results were linked to a specific robot aid and could not be applied to the case of a generic system. Consequently, no conclusions could be drawn for the purpose of establishing the functional requirements for RAID.
Other relevant technologies were also investigated as part of the survey. Page turning devices and computer access products are of critical importance to the RAID project and the technologies used in their implementation came under scrutiny. The available page turners are insufficiently versatile to accommodate the various paper documents found in an office environment and a robotic page turning solution was therefore adopted. A variety of customised keyboards, guards, speech recognition units, scanning systems and pointing devices are available to facilitate computer access. Many of these products are highly specialised and would be inappropriate for the RAID user group. In view of the recent trend towards graphical user interfaces (GUIs) for office software, an access strategy suitable for such software was considered essential. The application consistency enforced by GUIs ensures that the computer access problem need be solved only once for each user. It is intended that initial RAID users will employ a visual keyboard operated using their own powered wheelchair joystick via mouse emulation hardware.
3.3 User requirements specification
The user requirements specification describes the RAID workstation from the user's perspective and formed the basis for the development of the functional specification. The following definition of the RAID user group was used as a starting point:
Wheelchair users who have insufficient functions to be able to operate a computer workstation unaided, but who have at least two degrees of movement available, typically enough to operate a joystick, trackball or chin switch.
The prototype RAID workstation will be more specifically designed for a reference group of seven individuals who have similar user interfacing requirements. The user interface may therefore be standardised for the initial system trials. User requirements were classified according to the following areas:
- User interface
- Physical layout
- Peripheral equipment
- Additional software
The physical layout takes into account both the needs of the user and the integration of RAID into the office environment. The workstation layout must facilitate access using a variety of wheelchairs and accommodate the ergonomic requirements of any computer user. The design of shelving must allow for the manual storage and retrieval of documents by other office personnel. Aesthetic factors must also be considered since the workstation should not draw attention to itself in an office environment [8]. Individual offices may have specific peripheral or software requirements, such as a network or fax interface and the potential need for these sub-systems must not be excluded in the design phase.
The user requirements for RAID were determined using the following techniques:
- Structured interviews and questionnaires
- Response to sketches of potential functional options
- Indirect advice from qualified occupational therapists
In order to achieve a user-led design it is important that the user requirements specification does not limit itself to those areas which are known to be technically achievable. Many of the facilities desired by the RAID reference group will not be implemented in the short term due to technical constraints but care must be taken to ensure that the validity of the completed design is not compromised by these omissions. For this reason, regular contact was maintained with the reference group throughout the design phase.
3.4 Functional specification
The functional specification defines the user requirements of the RAID workstation in terms of the specific functions required to meet those requirements. It also addresses the safety issues involved in the implementation of the functions. In view of the limited scope of the TIDE pilot phase, it was necessary to constrain the functional specification to those areas which are technically achievable using existing technologies. Future phases of the TIDE programme may provide for the realisation of additional sub-systems involving a significant element of research.
A functional analysis of the RAID workstation was achieved by considering the system in terms of information input, output and storage. This information was additionally categorised according to the media on which it is transferred or recorded. A generic representation of this approach is shown in figure 1.

Figure 1. Information flow within a vocational workstation
It can be seen that information enters and leaves the workstation physically (on paper or diskette) and electronically (via a network or phone line). Paper and diskette media enter the robot domain, while electronic data enters the computer domain. The scanner, printer and diskette drive provide the interface between the two domains. Consumables, such as printer paper and blank diskettes, must also enter the robot domain while waste paper must be removed from the workstation.
3.5 Technical specification
The technical specification precisely describes the way in which each function of the RAID workstation will be implemented. It provides sufficient information to allow a conceptual design of the prototype system. The following sub-systems are described within the technical specification:
- Computer
- Software
- Printer
- Scanner
- Input and output trays
- Reader board
- Hard media storage devices
- Robot and controller
- End effectors
- Wheelchair interface
- Safety system
- Telephone
Precise products were specified only where necessary for the clarification of the technical requirement. This provides the flexibility to obtain the best product available at the procurement stage. In summary, a 486 PC employing optical disk storage has been specified for the computer domain. This PC will link to a transputer-based motor control board which will replace the standard control hardware of an RTX arm to enhance the robot's control characteristics. The control architecture of the robot sub-system is shown in figure 2. A number of interchangeable end effectors will provide manipulative flexibility in the robot domain. All application software will run under the Microsoft Windows GUI. A Windows version of the MASTER 2 robot control language will be used in the pilot phase. Windows allows several applications to co-exist on the screen. This feature has made the use of a single computer, for both robot control and vocation related software, a viable proposition. The WiViK visual keyboard [9] will be used to provide access to Windows from the user's wheelchair-mounted joystick operating as a mouse emulator via an infra-red link.

Figure 2. Control architecture of the RAID workstation
3.6 Test specification
A specification for verifying the adequate operation of each RAID sub-system was written as part of the design process. The specification itemises the functions provided and describes a procedure by which each function may be compared with the design with respect to operation, tolerance, safety and reliability as appropriate. The test specification forms a key element in the quality assurance plan of the RAID project.

Figure 3. Conceptual design of the RAID workstation
3.7 Conceptual workstation layout
The conceptual design of the RAID workstation (figure 3) was a valuable part of the overall design process. The drawings were produced on a CAD system while the Workspace robot simulation package [10] was used to create a number of animated task sequences based on the conceptual design. Workspace proved to be a useful tool in the visualisation of RAID tasks. It is currently the only PC-based robot simulation package known to the authors and is supplied with several robot models including one for the RTX. However, hard copy output is restricted to screen dumps at the resolution of the display adapter, as shown in figure 4. The RAID simulation was used to assess the geometry of the workstation with respect to the robot arm prior to construction. Additional feedback from the simulation was provided by the reference group. Note that the CAD and simulation drawings differ in detail since the simulation was created at a later stage in the design process. The robot body was omitted from the simulation for clarity.

Figure 4. Screen dump of the RAID workstation from Workspace
The workstation uses an RTX robot arm with enhanced vertical travel provided by a modified column. The arm is mounted on a linear track which provides motion across the width of the workstation. The book shelves, document rack, diskette rack and PC are all mounted within a rigid frame behind the robot. It is envisaged that peripheral devices such as the printer and the scanner will be situated in front of the robot below the book shelves while the user's desk and reader board will be to one side and angled with respect to the manipulation system. For reasons of safety, the user will not sit within the arm's working envelope.
3.8 User trials
A programme of user trials has been planned to follow the construction of the prototype workstation. The benchmark assessment of RAID will involve use of the workstation in the context of Computer Aided Design (CAD). While RAID is suitable for many computer related vocations, CAD presents the greatest challenge to the system due to the quantity and diversity of reference material which must be accessed during the design cycle.
A version of the AutoCAD software package employing a Windows interface will be used throughout the initial evaluation of RAID. Productivity measurements will compare the duration of various tasks with those of an able bodied reference group. An estimation of the available independence time for disabled users will also be made.
4 Workstation capabilities
4.1 Introduction
The capabilities of the RAID workstation are closely related to the tasks defined within the function specification but have been modified in response to limitations revealed within the technical specification. Implementation of several tasks were possible in either the robot or computer domains and the balance between task duration, reliability, implementation costs and space requirements was considered before opting for a particular solution. For example, users require a telephone but the reference group stated that they prefer to use a handset (requiring robot support) rather than a loud-speaking telephone for reasons of privacy. The functional specification dictated that the telephone should be answered within 5 seconds. This is not achievable by robot when the arm is currently involved in another task. Consequently, a non-robotic solution was selected, maintaining privacy by another means.
The RAID workstation capabilities may be conveniently reviewed with reference to the functional specification by considering the tasks performed in the robot and computer domains and at the interface between them separately. The potential for object transfer within the workstation is indicated in table 1.

Table 1. Object transfer potential
4.2 Tasks in the robot domain
The majority of tasks within the robot domain involve the transfer of physical media from one location to another. The media types accommodated by the RAID prototype are listed below:
- Single sheets of paper
- Paper documents
- Manuals
- Books
- Diskettes
- CD-ROMs
These items may be transferred between appropriate locations within the workstation using pre-programmed commands. The locations and via points will initially be taught using an on-screen teach pendant. Modified storage devices will be employed where necessary to provide robot access and a system of interchangeable end effectors will facilitate grasping of the various media types.
The key point of interaction between the robot domain and the user is at the reader board. The robot arm will deposit reference materials on the board for study and perform a page turning function for various document types. Pages retrieved from the printer may also need to be scrutinised before they are filed or leave the workstation. A reader board has been designed specifically for interaction with the end effectors of the RAID workstation, it employs a pneumatic finger which holds books open while they are read and retracts as the robot turns the pages using a specialised end effector.
4.3 Tasks in the computer domain
The tasks provided within the computer domain are precisely those available to able-bodied individuals within a modern electronic office environment. Information in the computer may enter and leave the workstation by the following routes:
- Fax
- Modem
- Local Area Network (including electronic mail)
- Typing (using visual keyboard or equivalent)
- Telephone (verbally relayed from display)
- Reading (directly from display)
Information may be transferred in these forms at any time using standard application software. The trend towards having a computer on every office desk will enhance the potential of physically disabled people in many vocations involving considerable interaction with other office staff.
4.4 Tasks at the domain interface
While computers provide almost unlimited flexibility in the processing of information, the majority of this information still enters and leaves a worker's desk on physical media. Interfacing devices are therefore necessary to transfer information between the robot and computer domains. It is relatively easy to create hard copy output from computer data using a printer, but less easy to transfer information from paper into the computer.
The RAID workstation will make use of a Document Image Processing (DIP) system which allows printed matter to be read into the computer as an electronic image using a flat-bed scanner. The images are then stored in a database together with indexing information allowing the image of a particular page to be rapidly identified and displayed. A major drawback of DIP lies in the size of the stored images. A 16-colour A4 image at a resolution of 300dpi will require approximately 4MB of storage. Standard compression algorithms allow such images to be reduced in size with little or no degradation, but high capacity storage devices are still necessary to hold a useful quantity of information. The RAID prototype will use 3.5" optical diskettes to hold the images.
The robot arm will be used for the following purposes during the transfer of information between the robot and computer domains:
- Insertion and removal of diskettes and CD-ROMs from their respective drives
- Retrieval of printer output
- Feeding the scanner
It is envisaged that only small quantities of information will be scanned by the user. Large jobs may be contracted to computer bureaux. DIP can only be viewed as a partial solution to the document storage requirements of disabled workers since the temporary nature of many large documents renders scanning uneconomic. The requirement for manual manipulation will therefore remain.
4.5 Independence issues
The feasibility of a workstation such as RAID has frequently been measured in terms of the duration of independent operation afforded. The RAID consortium recognises that the goal of completely independent operation over a full working day is beyond the scope of the pilot phase project. However, the level of manual intervention required is low and related to the activities of other personnel. For example, the refilling of printer paper trays and the opening of mail may be achieved by post delivery personnel when the mail is delivered. The durations of these duties are short enough to have a negligible effect on overall productivity and the function of the workstation is not compromised.
5 Summary
The specification and design of the prototype RAID workstation has been described. RAID uses the structure of a conventional development project and is led by the requirements of potential users. A rigorous specification was defined for the workstation design, based on user requirements, functional and technical issues. A functional analysis of the user requirements was achieved by considering the workstation in terms of information input, output and storage.
The conceptual design of the RAID workstation has been evaluated using the Workspace robot simulation package. This has enabled potential users to visualise the robot sub-system and provide feedback prior to construction. The robot arm will be used to transfer physical media around the workstation using a set of pre-programmed command sequences.
The RAID workstation will facilitate the employment of severely physically disabled individuals in computer oriented vocations. Initial trials will be specifically targeted towards Computer Aided Design due to the quantity and diversity of reference material which must be accessed for this purpose.
6 References
- Hammel JM & Van der Loos HFM (1990) A Vocational Assessment Model for use of Robotics Technology. RESNA 90 - Proceedings. 327-328
- Hillman M, Pullin G, Gammie A, Orpwood R & Stammers C (1991) Evolutionary Design: The Role of Clinical Trials in the Development of a Robotic Workstation. ICORR 91 - Conference Papers. 93-108
- Détriché J-M & Lesigne B (1992) MASTER Technical Evaluation in Rehabilitation Centers. RESNA 92 - Proceedings. 608-610
- Gosine RG, Harwin WS & Jackson RD (1990) An Interactive Robot Workstation for Applications in Rehabilitation. IROS 90 - Proceedings. 977-983
- Birch GE & Fengler M (1992) Experience with the Neil Squire Foundation/Regenesis Robotic Assistive Appliance. RESNA 92 - Proceedings. 598-601
- Fu C (1986) An Independent Vocational Workstation for a Quadriplegic. RESNA 86 - Proceedings. 182-184
- Smith JU (ed) (1992) The Way We Work. Microsoft Magazine. 2. 4. 24-27
- Birch GE & Cameron W (1990) User Acceptability in Robotic Assistive Devices. ICORR 90 - Conference Proceedings. 83-94
- Shein F, Hamann G, Brownlow J, Treviranus J, Milner M & Parnes P (1991) WiViK: A Visual Keyboard for Windows 3.0. RESNA 91 - Proceedings. 160-162
- Workspace User Manual, Version 2.0 (1991) Robot Simulations, Newcastle-upon-Tyne, UK
This paper may be referenced as follows:
Dallaway JL, Jackson RD (1992) RAID - a vocational robotic workstation. Proceedings of the third International Conference on Rehabilitation Robotics.