BACKGROUND

The wheelchair wheelie is an important foundation for many advanced wheelchair skills.¹ Kauzlarich et al² modeled the wheelie balance problem as an inverted pendulum and predicted that raising the center of gravity (CG) would make balance easier. Koshi et al³ have documented that increased rolling resistance (RR) reduces the perceived difficulty of maintaining wheelie balance. Davis and Heller^4,5 provided empirical evidence to corroborate the effects of RR and CG height on the ease of robotic wheelies. We have been attempting to develop a robotic wheelie system for two reasons – to improve training methods through better understanding of the skill and as a means of improving powered-wheelchair performance. The objective of the current study was to assess a robotic-wheelie prototype by testing the hypotheses that the duration of balance can be increased by increasing the rolling resistance and raising the CG.

METHODOLOGY

The prototype that we assessed was developed by Boone et al.⁶ It employed a 4-wheeled chassis to autonomously balance an inverted pendulum by means of digitized angular feedback and a proportional-integral-derivative (PID) controller. This was a single-input/single-output control system, where the input was the angle of the pendulum from vertical and the output was the voltage applied to the drive wheels. The device was programmed to employ a reactive balance strategy.⁷

We used two RR conditions, tile for low RR and carpet for high RR. To manipulate the height of the CG, we placed a perforated tennis ball on the pendulum that could be slid up and down. We recorded the balance times in 4 conditions: 1) low RR/low CG, 2) low RR/high CG, 3) high RR/low CG, and 4) high RR/high CG. Two people were required for testing. The first person balanced the inverted pendulum manually. The second person switched the device on when the first person released the pendulum. All trials were videotaped. Balance times were recorded from the videotapes.

RESULTS

Table 1 shows the results. The conditions in which the RR and CG height were greater corresponded to greater durations of robotic wheelie balance. The lowest duration of balance was in the low RR/Low CG condition and the greatest duration was in the High RR/High CG condition. The variance of the data, reflected by the standard deviations, was moderate to large.

Table 1. Duration of robotic wheelie balance time (in sec) in 4 conditions

Parameter	Condition
	Low RR/Low CG	Low RR/High CG	High RR/Low CG	High RR/High CG
Mean	4.6	6.5	5.2	11.4
Standard deviation	1.6	3.9	2.4	7.4
Abbreviations: RR = rolling resistance, CG = height of center of gravity

DISCUSSION

Increasing the RR or the CG height each increased the duration of balance. The balance time in the high RR/high CG condition was 2.48 times longer than in the low RR/low CG condition. Refinement of our robotic wheelie device has the potential for even better understanding of the wheelie skill. For instance, the balance strategy for this project was solely reactive, whereas we plan to combine reactive and proactive strategies in future iterations.

Nevertheless, our findings already have implications for training. We have been using a high-RR setting for learning to perform the wheelie skill and learners report finding this technique to be helpful.⁸ We are also considering means by which the CG could be raised (e.g. with an attachment to the wheelchair, above the wheelchair user’s head, to which weight could be added).

In addition to looking at the wheelie as performed by the users of manual wheelchairs, we have begun considering how this information could be of use for powered wheelchairs. Although there has been some work in this area,⁹ the existing devices do not mimic well the way in which manual wheelchair users achieve, maintain and utilize the wheelie position in community settings.