We built a simulator that allows users to experience being wheelchair bound.
We performed a pilot study to compare the effect of a manual versus an electric wheelchair on presence and distance estimation in a virtual environment.
In the study, users had to navigate through a few city blocks to meet a friend at a hotel.
They also had to provide distance estimates to objects along the way.
The manual wheelchair consists of a kitchen chair, a pair of bicycle wheels mounted on bicycle trainers, and a pair of Nintendo Wii Remotes (wiimotes).
The bicycle trainers allow the user to turn the bicycle wheels without actually traveling anywhere.
They also provide resistance so a realistic amount of force is needed to spin the wheels and they stop spinning quickly after the user lets go.
Wiimotes are mounted between the spokes, in the "plane" of the wheel.
The wiimote contains three acceloremeters, one oriented along each of the x, y, and z axes.
We use the x and y acceleration data to approximate the angle of the wiimote relative to the hub of the wheel.
This allows us to compute an approximate angular velocity as the user spins the wheel.
There are small inaccuracies due to centripetal acceleration and spikes when the user applies force to the wheel, but they do not cause significant problems.
The electric wheelchair consists of a non-swiveling office chair and a wiimote with the nunchuk attachment.
The nunchuk attachment has a small joystick similar to the joystick found on a real electric wheelchair.
The nunchuk is strapped to the armrest of the chair.
Pushing forward and backwards moves the wheelchair forward and backward.
Pushing left and right turn left and right.
Display and tracking
Users wore a V8 HMD which provides 640x480 in each eye and a 60 degree diagonal field of vision.
The environment was rendered in stereoscopic 3d using Ogre.
Head tracking was done using in Optitrak system and Vtrak, a software tracker written by Kyle Johnsen at the Univeristy of Floriad.
We first told the user that they were playing the role of a soldier who lost his legs in Iraq.
We then presented the user with an actual newspaper article about the soldier.
After the user read the article, we showed them a map of the city block and told them they were trying to meet a friend at the hotel.
We pointed out the bus stop where they would start and the hotel where they would finish.
We also told them that their friend would call and provide directions over a "bluetooth" headset.
The "bluetooth" headset was simply the headphones built into the V8 HMD.
Only the right speaker worked, so it provided sound only in one ear like a bluetooth headset.
This allowed us to guide the user and ask him or her the distance estimation questions without a break in presence.
This audio was pre-recorded as a sequence of prompts and questions.
We also played city background noises on a set of external speakers.
The chair provides passive haptic feedback that the user is sitting in a wheelchair.
For the manual wheelchair, the bicycle wheels turning with resistance provided feedback.
This study was only a pilot study.
The population is too small to have significant results.
Also, technical difficulties made it difficult to control conditions.
The manual wheelchair broke in one case, and a softawre update window interrupted another user's experience.
Participants were recruited from friends, classmates, friends of classmates, and professors.
They were randomly divided into the manual and electric conditions.
Five participants used the manual wheelchair and six participants used the electric wheelchair.
None of the manual participants had ever used a wheelchair before.
Two of the electric participants hade used a manual wheelchair before.
Video game playing frequency varied greatly, from never to daily.
Five of the participants had used an HMD before.
The users started their experience at a bus stop on a city street.
They were then told to look for a food cart and estimate the distance.
After that, they had to navigate to the food cart.
They then had to find a mailbox and estimate the distance.
From there, they had to cross an intersection, then travel down the block to another intersection.
At that intersection, they turned and continued on to the hotel.
At the front of the hotel, they were presented with a ramp and stairs.
The experience ended when they chose either the ramp or stairs to go up.
Since distance estimation is a difficult task for many people and often even harder in a HMD, we wanted to compare the distance estimates to objects before and after moving.
The actual distances for both objects were identical.
Unfortunately, many of the users did not provide distance estimations.
Two of the manual users reported the second distance as further, one as the same, and the other two did not respond.
One electric user reported the second distance as further, one as the same, and the other four did not respond.
We expected the manual condition to report farther distances because research shows increased effort relates to increased distance estimations.
We gave the participants a questionnaire based on the standard SUS presence questionanire.
With such small groups in each condition, it is difficult to obtain statistically significant results.
Surprisngly, the electric wheelchair group had better presence scores.
Five of the six participants in that group rated the statement "There were times during the experience when the city was the reality for me" a 6 or 7 compared to only one in the manual group.
Four of the six electric participants rated the statements "I had a sense of being there in the city" and "The city seems to be more like somewhere that I visited" a 6 or 7 compared to only 1 in the manual group.
We wanted to see if users were more likely to use the ramp in the manual condition.
Only two of the manual users used the stairs, but one purposefully did it as a joke.
Only one joystick user used the stairs, but two got lost and did not reach the hotel and a software update window interrupted the last user, making him unable to complete the task.
Source code is available here as a mercurial repository. Login with the username public, password public.