- 86 -

ICORR ’99: International Conference on Rehabilitation Robotics, Stanford, CA

A WHEELCHAIR MOUNTED ASSISTIVE ROBOT

Michael Hillman, Karen Hagan, Sean Hagan, Jill Jepson, Roger OrpwoodBath Institute of Medical Engineering Ltd, UK

Abstract

A robotic manipulator has beenmounted to an electric wheelchair toassist people with disabilities.Particular emphasis has been given tothe constraints and requirements forwheelchair mounting.

Background

Many different approaches to assistiverobotics have been both suggested andimplemented. Whilst in some situations(for example a vocational setting) afixed site workstation is suitable [1], inother cases (for example someoneliving independently in their ownhome) a mobile device [2] is moreappropriate.

An earlier project at our Instituteimplemented a low cost mobile robotby mounting a manipulator on a simplenon-powered trolley base, which couldbe moved around the home by a carer.

In order to extend the flexibility of thissystem, the same manipulator is nowmounted onto an electric wheelchair asdescribed in the current paper.

Methods

Central to the Institute’s designphilosophy [3] is the involvement ofusers at all stages of a device’s

development. In the case of thewheelchair-mounted robot project wehave been in contact with about 30volunteers, covering 5 disabilitygroups. Of these a smaller number oflocal volunteers have been involved inmore detailed discussions. We havealso tried to involve disabledvolunteer’s carers wherever possible,because they too are users of thedevice.

In order to gauge volunteers' reactionsto a device before investing time andexpense in producing a workingprototype it is often valuable to build amodel or full scale non-working mockup. In the case of this project, this wasa valuable way of gaining an insightinto how users might react to having alarge robotic device mounted to theirwheelchair.

A fully working prototype is necessaryto evaluate the functionality of adevice. However, the prototype is notan end in itself but is only the firststage in making finished devicesavailable to those who need them.

Specification

Many surveys [4] have reporteddifferent tasks which a disabled usermight use an assistive robot for. Otherpapers [5] have described the use ofrobots in real life situations. It is not

- 87 -

ICORR ’99: International Conference on Rehabilitation Robotics, Stanford, CA

appropriate to repeat these statistics.However it is useful to divide the tasksbriefly into groupings.• Eating and drinking• Personal hygiene• Work• Leisure• Mobility

Many of these task areas are commonto all assistive robot systems. Howeversome tasks are more appropriate for afixed site workstation, perhaps used fora vocational application, while others,are more specific to a wheelchair-mounted robot. These tasks includegeneral reaching operations as well asmore specific tasks related to mobilitysuch as opening doors and windowsand operating switches (e.g. lightswitches, lift call buttons).

Discussions with users identified someof the specific requirements andconstraints for a wheelchair-mountedmanipulator:

RequirementsIt must be able to:• reach to floor level;• reach to head height.

ConstraintsIt must not:• compromise manoeuvrability;• obstruct the wheelchair user’s

vision;• create a negative visual impact;• affect the steering or control of the

wheelchair;

• affect seat adjustment (or anysimilar facilities of the chair);

• affect transfers into or out of thewheelchair;

• cause an unacceptable drain on thewheelchair batteries.

Design Description

Vertical actuator & wheelchairmountingThe vertical actuator and how to mountit to a wheelchair are the most criticaldesign aspects of the project. Some ofthe initial concepts have already beenreported [6]. Use of a non-workingmock up allowed evaluation of theseconcepts.

Figure 1. Mock-up manipulator

- 88 -

ICORR ’99: International Conference on Rehabilitation Robotics, Stanford, CA

Both wheelchair users and others whosaw the mock-up thought that thesingle stage actuator was too obtrusive.In order to overcome this, an extendingmechanism was used which, in itsparked (lower) position, does notextend noticeably above head height.

The mechanism is based around twoparallel vertical tracks, linked by apulley. As the moving section of theactuator moves upwards relative to thefixed section, the upper arm mountingpoint moves upwards relative to themoving section. Two constant tensionsprings counterbalance the weight ofthe arm so that a small motor of only6W may raise the whole arm.

The mock-up mounted the manipulatoron a hinged mounting point towards therear of the wheelchair, allowing themanipulator to be swung forwardswhen required. It was found that theuse of a hinged mounting required toomuch clearance to the side of thewheelchair, often not possible in asmall room. The manipulator istherefore now mounted in a fixedposition above the rear wheels. Whilenot giving quite as much forward reachas had been originally specified thisseems a good compromise solution.

Mounting the manipulator at the side,close to the shoulder of the user,decreases the visual impact of thedevice and does not obstruct thewheelchair approaching a table or desk.Since the weight is over the fixed,rather than castoring, wheels the device

does not greatly effect the steering. Thefollowing photograph (Figure 2) showsthe prototype (without cosmetic covers)mounted on a "Scandinavian Mobility"electric wheelchair.

Figure 2. Manipulator mounted towheelchair.

Upper armThe basic design of the upper arm iscopied from the earlier trolley-mountedmanipulator. The main rotary joints(identified as shoulder, elbow and wristyaw) all move in a horizontal plane.Vertical movement comes from thevertical actuator described above. Atthe wrist there are roll and pitchmovements. The basic designcomprises an aluminium structure,

- 89 -

ICORR ’99: International Conference on Rehabilitation Robotics, Stanford, CA

within which the motors are mounted,covered by a vacuum-formed cosmeticmoulding.

The opportunity was taken to improvethe design, particularly in the area ofaccess for maintenance. The motors arenow mounted within modules, whichmay be easily removed formaintenance. The cosmetic covers arealso redesigned for easier removal andimproved aesthetics.

GripperThe earlier trolley-mounted robot useda prosthetic hand end effector. Thisnever proved totally effective as a robotgripper. A purpose made gripper hasbeen designed specifically for thecurrent device. It has the followingfeatures:• Two parallel moving jaws;• Slim profile to allow good visibility

of the item being gripped;• Compliant elements in the drive

train to allow variable forcegripping;

• Non backdrivable gearing andcompliance to maintain grip forcewhen power is removed from thedrive motor.

Figure 3. Gripper (without covers)

ElectronicsThe electronics design is based aroundan I2C serial link running through thelength of the manipulator. There arealso 5v (for digital electronics) and 24v(for motor power) power suppliesrunning through the manipulator. Asingle board PC compatible processor(GCAT from DSP Design, London,UK) mounted at the base of themanipulator sends command signals tomotor control boards mounted withinthe manipulator. On the control boards(size only 50mm x 50mm) the serialsignal is converted to a parallel signalfor the proprietary HCTL1100 motorcontrol chips. Motor control uses pulsewidth modulation.

Figure 4. Electronics block diagram

User InterfaceThere are two main approaches to userinterface design for an assistive robot.• Task command: This works well in

the structured environment of aworkstation. It may be lessappropriate in the undefinedenvironment within which a

- 90 -

ICORR ’99: International Conference on Rehabilitation Robotics, Stanford, CA

wheelchair-mounted robot will berequired to operate.

• Direct control: This allows the userto control the manipulator in anundefined environment. It does,however, make a greater demand onthe user and may be time consumingand tedious.

The main approach used for thewheelchair-mounted robot is directcontrol, although there are alsofunctions to allow the manipulator tobe moved easily to certain pre-setorientations.

Users of electric wheelchairs aregenerally able to use a two-degree offreedom input, either a conventionaljoystick or a head or chin operatedjoystick. It was decided that this wouldbe the most appropriate input for awheelchair-mounted robot (although aswitch-operated system will also beavailable as an option). The use of atwo-degree of freedom joystickprovides an intuitive form of control ofa manipulator in real time. In the longterm we envisage the user being able touse the same joystick to control bothwheelchair and manipulator.

Control of a six-degree of freedomdevice with a two-degree of freedominput requires mode switching. Thescheme used for the wheelchair-mounted robot uses the joystickmovements to navigate around a map(Figure 5), displayed on a small LCDscreen, or to switch to an alternativemode.

Figure 5. User interface display.

Conclusions

At the time of writing (Jan 99), thesystem is at the stage of final assemblyand debugging of software. Briefevaluations are due to start in April 99.A mobile base has been designed, ontowhich the manipulator can be mounted.This can be wheeled up close to a user’swheelchair and will enable evaluationsto be carried out from the user’s ownwheelchair.

Further developments are plannedincluding the facility to integrate thesystem with a range of wheelchairs.This will enable longer termevaluations to take place towards theend of the year.

Acknowledgements

The authors are grateful to the SouthernTrust for their generous support of thiswork. A panel of 29 electric wheelchairusers has given vital user input to theproject. The authors also acknowledgethe contribution to the project from thetechnical staff at the Institute,particularly Martin Rouse (Mechanical

- 91 -

ICORR ’99: International Conference on Rehabilitation Robotics, Stanford, CA

Workshop) and Simon Gale(Electronics Laboratory).

References

1. Hammel J, Van der Loos HFM"Factors in the prescription & costeffectiveness of robot systems for high-level quadriplegics", Proc RESNA1991, 14, 16-18, 1991.

2. Kwee HH, Duimel JJ, Smits JJ,Tuinhof de Moed AA, van WoerdenJA, v.d. Kolk LW, Rosier JC, "TheMANUS wheelchair-borne manipulatorsystem review and first results", Proc.2nd Workshop on Medical &Healthcare Robotics, Newcastle uponTyne, UK, 385-403, 1989.

3. Orpwood R, "Design methodologyfor aids for the disabled", Journal ofMedical Engineering & Technology,14, 1, 2-10, 1990.

4. Prior S, "An electric wheelchairmounted robotic arm – A survey ofpotential users", Journal of MedicalEngineering & Technology, 14, 4, 143-154, 1990.

5. Hillman M, Jepson J, "Evaluation ofa trolley mounted robot – A casestudy", Proc. ICORR'97, 95-98, 1997.

6. Hagan K, Hagan S, Hillman M,Jepson J, "Design of a wheelchairmounted robot", Proc. ICORR'97, 27-30, 1997.

Author Address & contactinformationDr Michael HillmanBath Institute of Medical EngineeringWolfson CentreRoyal United HospitalBath BA1 3NG. UKTel (+44) 1225 824103Fax (+44) 1225 824111M.R.Hillman@bath.ac.ukhttp://www.bath.ac.uk/~mpsmrh/