Wheelchair Simulation M.Grant, 1 C. Harrison,2 and B. Conway3

1 Introduction Recent times have seen an upsurge in interest in the area of “inclusive design” within which access to the built environment has enjoyed a prominent position. There are a number of factors providing the impetus for this, not least a growing awareness of the quality issues incumbent in inclusivity but also there is evidence of a response to the threat of the impending legislation within the Disability Discrimination Act (DDA) (DDA, 1995). These factors have lead to the production and availability of a range of tools targeting design issues within these sectors. Among these projects are developments at Strathclyde University that sought to combine advanced graphics with an allied haptic interface in order to construct a wheelchair motion platform capable of simulating wheelchair navigation in virtual buildings. This is arguably one of the more sophisticated approaches now on offer yet it still fails to address all the problems that a designer might face regarding access and interaction within our built environment.

2 Background A number of other research teams have adopted similar methodologies targeting wheelchair training, cognitive development, hardware interfacing and architectural design criteria. In the light of the widespread interest and reportage of these 1

ABACUS, Dept Architecture, University of Strathclyde, 131 Rottenrow, Glasgow G4 ONG, UK 2 Telepresence Research Group, Dept of DMEM, University of Strathclyde, Glasgow, G1 1XJ, UK 3 Bioengineering Unit, University of Strathclyde, Glasgow, G4 0NW, UK

102

Grant, Harrison and Conway

projects this paper seeks to examine some aspects of this technology, looking at the strength of the concept and identifying future directions and opportunities. Even cursory research on the Internet reveals a number of institutions engaged on research and development that falls under the remit of wheelchair simulation. An initial overview is reported in the following table which indicates the wide range of applications currently being addressed. • • • • • • • • • • • •

Manual Wheelchair Simulation; Powered Wheelchair Simulation; Assistive Technology Research; Wheelchair Control Interface Design; Built Environment Access Awareness; Architectural Route Planning; Architectural Design Evaluation; Building Code Compliance; Wheelchair Operation and Training; Cognitive Development; Virtual Environment Design; Haptics and Sensory Feedback.

3 Review of Current Practice 3.1 Powered Operation Wheelchair usage tends to be categorised under two main headings, one relating to powered operation and the other to manual modes of propulsion. The common interface to the control of a powered wheelchair is through a joystick, a system not dissimilar from that used with PC based games and other interactive graphics programs. This then makes it relatively straight forward to emulate wheelchair navigation using consumer grade hardware and readily available software. The deployment of virtual environments in this mode traditionally takes the familiar form of the game engine format with terrain following and collision detection facilities but offering little if any haptic feedback. This level of technology provides for the basics in wheelchair operation and training, giving access to the basic coordination required to safely navigate within the built environment. Further enhancements to this methodology with the addition of audio and other perceptual cues allow the use of the system to be directed towards the realm of cognitive development and therapeutic applications (Tefft 1999).

3.2 Manual Propulsion The simulation of manual propulsion is more complex as the interface is through the rotation of the driving wheels and is more directly linked to the users own physical effort. This is explored in the systems developed at Strathclyde (Harrison,

Wheelchair Simulation 103

2003), Umeå University (Holmlund K, 2002). and Mitsubushi Precision, Japan (Mitsubushi 1997) These projects seek to offer a more implicit simulation as opposed to the empirical implementation of what can be termed the standard format. This is not to imply that the more simple of these design tools are any less worthy than those of greater sophistication, indeed it is easy to argue that those that are less expensive and more easily deployed have the greater utility. The purpose of exploring this hierarchy is to allow a comparison of capabilities and to propose a specification that can integrate and embody the strengths of all approaches to the problem area.

4 Implementation Issues The design of a wheelchair simulator can be considered as being derived from three main functional components. The visual simulation is concerned with the generation and display of the virtual environment, the physical simulation dictates how the wheelchair and its occupant interact with the virtual world and the control system governs the interface between the user and their chair.

4.1 The Visual Simulation The realism of even a simple model is generally acknowledged to be adequate in generating a sufficiently complex rendition of the built environment which can in turn provide all the necessary visual cues required for surface recognition and spatial navigation. The quality of the visual environment is then a function of the graphical power of the display generator which must be capable of processing a reasonably complex geometrical data set while delivering a frame rate that provides for a smooth visual experience. The greatest leverage on visual realism is provided by the display system. In order to increase the net benefit of a graphical display the goal has always been directed towards the achievement of a sense of immersion. This state is achieved when the perceptual barrier between the viewer and the scene is diminished to the extent that they become “immersed” within the virtual environment. 4.1.1 Head Mounted Displays Head mounted displays have enjoyed a brief period of popularity but in practice the difficulties associated with wearing a headset, especially where the user is perhaps already muscularly impaired, and having to deal with the associated problems of heavy trailing cables (Browning, 1996) has meant that their use has become depreciated. Many other systems rely solely on the traditional computer monitor, still others rely on projection (Cremers 2001), and some are based around more technologically advanced systems such as the CAVE (Browning, 1996) or the Reality Room (Grant, 2003). In the case of the later it can be demonstrated that the display system demonstrates a mix of positive and negative aspects relating to

104

Grant, Harrison and Conway

the technology employed. The wide angle of view is perceived as a benefit as this allowed the user’s direction of view to be decoupled from the direction of motion of the wheelchair. This enables a user to look around within the environment rather than be constrained to the narrow view frustum common to conventional graphics displays. The downside is that this form of display provides an “out of the window” view which separates a user from objects which would otherwise be within arms length. This had been thought to be a drawback as wheelchair users tend to make reference to the extremities of their chair when negotiating obstacles. In trials with the Strathclyde system this did not seem to be a disadvantage, (Grant 2004) however it should be noted that all participants in this particular experiment were experienced wheelchair users and a novice may encounter greater difficulty in judging clearances without visual feedback. A possible technological solution for all “out of the window” projection based systems is to provide for a “3 rd person view” where the camera is de-couple from the user’s eye point and can be autonomously directed at the point of interest. This does come at some increase in the complexity of the development of the graphical software system.

4.2 The Physical Simulation In order to reproduce the experience of navigating within the built environment the simulator must map intended motion into the virtual world while reflecting the influence of collisions, gradients, surface properties and the interaction of the user with his environment. All aspects of the physical simulation are based on the interaction of geometry within the virtual world and as such any refinement or extension of these capabilities would usually be just a matter of providing additional software functionality. Collision detection is a fundamental requirement within a simulator. The least complex form of implementing this feature tends to treat the user as a single point object which can then be tested for contact with surfaces within the environment. This scenario is sufficient for basic navigation but when the goal is to determine minimum clearances in an architectural context then not only should the entire volume of the chair be considered but also the dimensional detail of the footplate and other participating protuberances. (Han, 2002) (Stredney 1995). 4.2.1 Collision Detection While basic collision detection keeps the chair anchored to the ground plane and prevents the user from driving through walls and other boundaries, other environmental features such as kerbs represent a singular challenge to both real world wheelchair users and the simulation software. In a simulation this feature can be represented by making the upstand of the kerb a very short, but steep, incline as opposed to being truly vertical. This allows the software to treat kerbs in the same manner as all other inclines, requiring substantial input to climb the obstruction, but not faithfully mimicking real world practice. Other objects within the environment, such as furniture and any natural obstructions, must also be present in the simulation but the precise recognition of collisions with these entities predicates a high resolution algorithm for accurate detection.

Wheelchair Simulation 105

4.2.2 Mechanical Feedback Powered wheelchairs generally have sufficient torque to make light work of gradients in the real world, a manual wheelchair requires significantly more user input. If the goal is to faithfully simulate real world conditions then this must be taken into account. In the same way the effect of varying surface characteristics on rolling resistance is a fundamental physical limitation on manual wheelchair operation. (Maver 2001) Both these characteristics assume some degree of mechanical feedback to the user via their operation of the chair however this only becomes available after a significant investment in engineering directed towards a motion platform embodying haptic actuators. Most systems are primarily directed towards the interaction of the wheelchair with the environment as opposed to modelling the user’s personal interventions. This form of interaction is exemplified by the manner in which a user might negotiate swing doors. In this event a wheelchair occupant might tend to use their knees, or the chair itself, to wedge the door open while manoeuvring for a favourable position from which to exert additional leverage. This was beyond the scope of current implementations. Similarly if this mode of personal interaction is constrained then the user cannot reach into the environment to manipulate objects, operate switches or open cabinets or drawers. This could be addressed through the use of body worn sensors, such as a data-glove, which would further enhance the functionality of the system and improve on its use as a tool for examining the usability of room layouts designed for wheelchair users. The use of such sensors is made more difficult by the requirement that the operator’s hands should be free to control the wheelchair. In this scenario the trailing cables commonly associated with these devices are a potential hazard. Tether free optical sensors are becoming available which could provide a solution especially as the stray magnetic fields generated by powered wheelchairs have proven to interfere with the electronic varieties. Some developers have proposed the provision of a tilting platform to simulate the attitude of the chair on varied slopes. This is a feature that comes at a price of considerable additional complexity. In an immersive environment the image horizon tilts in conjunction with the normal vector of the ground plane at the chairs point of contact and in practice the Strathclyde research team discovered that this was sufficient to generate a user’s perception of the slope. (Grant 2004).

4.3 The Control System 4.3.1 Standard Interface Control of a wheelchair simulation falls into either of two classifications as previously described. The first is the most easy to implement and can consist of a standard computer peripheral joystick used in a manner directly analogous to a powered wheelchair’s control interface. This provides adequate directional control but does not account for individual wheelchair dynamics or feedback from any interaction with the virtual world. This system obviously benefits from ease of implementation and can be supplemented or supplanted by any of the specific devices provided for the disabled to interact with the computer. (Roast, 2002). This

106

Grant, Harrison and Conway

can be extended in that a wide range of interfaces can be offered for testing, for example button boxes, halo, sip and puff actuators or muscle controlled devices, many of which can be readily adapted to operate with a standard PC. 4.3.2 Motion Platform The second classification relates to those systems that use a motion platform to interface between the user’s wheelchair and the virtual world, (Harrison 2000), (Mitsubishi 1997), (Holmlund 2002). In this instance the task is to translate the wheel rotations from the incremental movement of rotary sensors into translation and rotation of the wheelchair via a motion algorithm. This set of function accurately models the gross behaviour of the wheelchair and has the advantage of replicating all the characteristics of the control device, if powered, or the feedback characteristics of the manually operated chair. The sophistication of the algorithm that translates wheel rotation into progress can be crucial in determining the realism of the simulation. For example, the position and function of the wheelchair castors exerts a subtle influence on progress. This castoring action introduces two further complications to the model. Firstly, the castors tend to transfer torque between the driving wheels. This tends to stabilise the heading of the chair, an effect that increases with speed. Secondly, the orientation of the castors is a function of the previous direction of motion; any subsequent movement on a new heading must first re-align the castors with the new direction. This can result in unexpected deviations from the desired course, especially when first encountered by new wheelchair users. When the simulation failed to take account of the first feature it will be difficult to maintain a constant heading but a simple algorithm can be introduced to mimic the effect of torque transfer and this succeeds in damping out the oscillations. The second feature is more subtle and perhaps of greater concern to powered wheelchair users. This is particularly noticeable when manoeuvring into a tight space as there is often insufficient space to align the castors before making the final movement. To enhance the perceived impression of reality an interface should ideally mimic real world practice. The most important factor is that it must recognise and respond to real world constraints. If the means of locomotion is some form of vehicle then it must “drive” as it would in reality but if the metaphor is that of some means of human locomotion then the opportunity for kinesthetic feedback is much greater. The disadvantage of a passive interface is that there is no energy input into the system, with a human locomotion interface effort is rewarded by progress. This results in more realistic navigation and an enhanced appreciation of the spatial layout. It has also been demonstrated that locomotion helps to calibrate vision and can improve any decision making processes that rely on the judgement of distances or spatial qualities. A manual wheelchair offers an ideal interface device in that the human input into the system is constrained and channelled through the action of the upper limbs in conjunction with the driving wheels. As this is the only pathway through which input is translated into action it is relatively easy to design instrumentation with which to bridge the real and virtual worlds. Since there is no conceptual abstraction as to the method of locomotion then the potential of providing accuracy with realism is at it highest.

Wheelchair Simulation 107

5 The Desire The stated aims of the Strathclyde project were to provide a virtual reality facility that could be used to generate, via an interaction between architects, designers and wheelchair users, guidelines which address the issue of wheelchair access to, and within, the built environment. The project aimed to design and build a wheelchair motion platform through which wheelchair users could explore virtual representations of buildings. It was envisaged that such a facility would form a powerful and cost effective means of evaluating wheelchair access provision early in the design of new buildings and in the redevelopment of existing buildings (Forest and Gombas 1995). Accordingly the following preliminary objectives were required: • • •

the ability to accurately monitor intended wheelchair motion and have the capability to provide physical and optical feedback to the wheelchair user on the presence of virtual obstacles or changes in floor coverings or slope; an interface between the platform and a virtual reality facility in order to provide an immersive virtual environment within which navigation is linked to the intended wheelchair motion; the ability to generate virtual representations of a range of building types in order to test and calibrate the performance of the platform and perform an evaluation of the system by wheelchair users.

This concept embodied in this last point has become known as “Virtual Reality” within which the ability to derive any form of meaningful interaction with a computer model can be said to be dependant on three characteristics: • • •

the quality of the model, which must be dimensionally consistent, visually compelling, and should accurately simulate the physical constraints of the real world; the ability to visualise the graphical output free from the subjugation of the traditional computer monitor; the ability to interact with the virtual world in a manner that is free from the constraints and abstraction of an artificial control metaphor.

The development and construction of a wheel chair motion platform, based on the above principles, required the incorporation of several key features which are outlined below: • • • •

the platform must accurately detect the rotation of the driving wheels and use this to provide realistic navigation within the VR world; non-visual environmental feedback (haptics) should be provided that match the altered sense of effort needed to propel a wheelchair over varying surfaces and slope conditions; collisions with virtual objects should combine visual and non-visual simulation directly analogous to that encountered in reality; communications between platform and VR host should be fast and provide near real-time interaction;

108

Grant, Harrison and Conway

• •

the platform should allow wheelchair users to use their own wheelchairs while navigating in the virtual environment. The platform must have the flexibility to accommodate chairs from a range of manufactures; the navigation route through a VR model together with any collision points and other data should be logged for off-line evaluation purposes.

Systems that utilise many aspects of this philosophy has been developed by a number of project teams, yet there is no one project that seeks to address then all. The conjunction of these capabilities would undoubtedly make for a powerful tool with the flexibility and capability to address all areas of interest.

6 Summary This review has shown that there are many areas of research currently being addressed covering a wide spectrum of applications. The field of driver training is one that is popular among researchers and if the goal is just to extend the user’s capabilities in basic operations such as turning, stopping and obstacle avoidance then the demands on the technology are slight. Simulation not only offers the ability to train novice users in a safe environment but also gives those charged with equipping them an early insight into capabilities of the user. To extend this basic functionality in order to encourage cognitive development again promises great rewards for a modest technological investment. Additional functionality within the graphical environment can reinforce the given stimulus by providing enhanced interactivity and incentives for exploration and discovery. Many studies have linked the acquirement of self mobility and independence to increased social interaction and perceptual development. The assessment of accessibility within the built environment brings not only a greater requirement for a true appreciation of time, visual perspective and scale but also for a more realistic motion model that can accurately replicate the dynamics of a wheelchair. Merely adhering to prescriptive design guidelines does not always result in the optimal, or even usable, environment. The ultimate goal of a simulator is to communicate the experience of wheelchair operation as this is arguably the only indirect means of making qualitative judgements on issues of ease of navigation and accessibility. This is perhaps most useful when trying to place a value on what, in the terminology of the DDA is “reasonable” when it comes to debating interventions in buildings. Not withstanding future technical developments of the system the main research challenge will be to convince architects and building managers that there are real benefits to be gained through the use of VR as a tool to improved disabled access.

Wheelchair Simulation 109

7 References DDA, “Disability Discrimination Act” (1995) (http://www.disability.gov.uk/dda/index.html) Grant M, Harrison C, Conway B (2004) An haptic interface for wheelchair navigation in virtual worlds. Presence, MIT Press (In Press) Grant M, Harrison C, Conway B (2003) VR in the service of people with special needs. Include 2003: The practice of inclusive design, Royal College of Art, London, 25-28 Maver T, Harrison C, Grant M (2001) Virtual environments for special needs. CAAD Futures 2001, Eindhoven. Edited by; Bauke de Vries, Jos van Leewen, Henri Achter, Klewer Academic Press Harrison C (2000) Development of a Wheelchair Virtual Reality Platform for Use in Evaluating Wheelchair Access, 3rd International Conference on Disability, VR and Associated Technologies, Sardinia, Edited by P.Sharkey ICDVRAT Cremers G, et al. (2001) Simulation of wheelchair use. International Journal on Human Friendly Wellfare Robotic Systems, Korea. Special issue on rehabilitation robotics 4: (1) Han C, et al. (2002) Compliance analysis for disabled access. Advances in Digital Government Technology, Human Factors, and Policy. W J McIver, Jr and AK Elmagarmid (eds) Boston Kluwer Roast C (2002) Virtuality for assistive technology experience. Proceedings of CWUAAT2002, Edited by: S Keates, P Langdon, PJ Clarkson and P Robinson, Cambridge pp 141-144 Tefft T et al. (1999) Cognitive predictors of young children's readiness for powered mobility. Developmental Medicine & Child Neurology, 41: 665-670 Sheldon S (1998) Commercially viable force feedback controller for individuals with neuromotor disabilities. (http://www.osc.edu/research/Biomed/past_projects/wheelchair/index.shtml) Holmlund K (2002) Implementing force feedback on a wheelchair. (http://www.vrlab.umu.se/forskning/rullstol.shtml) Mitsubushi (1997) Learning to be free. (http://www.mitsubishi.or.jp/e/monitor/monitor_old/97version/monitor7/chairsim.html) Browning R (1996) Input interfacing to the CAVE by persons with disabilities. CSUN’s Annual International Conference, Carolina State University Forrest D, Gombas G, (1995) Wheelchair accessible housing: Its role in cost containment in spinal cord injury. Archives of Physical Medicine and Rehabilitation, 76 (5): 450-452 Stredney D, et al. (1995) The determination of environmental accessibility and ADA compliance through virtual wheelchair simulation. Presence: Teleoperators and Virtual Environments, 4 (3): 297-305