Evaluation of an isometric joystick with control enhancing algorithms for improved driving of electric powered wheelchairs

Introduction. Ten isometric joysticks (IJ) for driving electric powered wheelchairs (EPW) were constructed by the Human Engineering Research Laboratorles (HERL). These prototypes used strain gage technology, multi-layer PC boards, surface-mount components, machined metal enclosures, standard operator controls and met current safety standards. The IJ's microcontroller firmware was revised to execute variable gain algorithms (VGA). A Target Acquisition study based on Fitts' law was conducted to compare two isometric control algorithms with the driving accuracy provided by a conventional, movement-sensing joystick (MSJ). The first algorithm (FSJ) emulated a square template with a fixed gain cut-off strategy. The second algorithm was designed to duplicate the profile of a standard MSJ. Forces associated with moving the MSJ stick were characterized and replicated on the VGA. Gain and arctangent lookup tables scaled the VGA's output vector and maintained the input/output aspect ratio even when the operator exceeded maximum input force---emulating the MSJ's circular template. Methodology. Thirteen subjects with disabilities, daily users of EPWs drove a test EPW (Quickie P300) to nine circular targets (black vinyl, 155 cm diameter) in a laboratory 8 x 7 meters. From one starting position, subjects drove to each target three times with each joystick; 81 trials total. Targets and joysticks were randomized. Four optical sensors on the test chair undercarriage detected target acquisition and interfaced to a computer to record Reaction, Traverse, Dwell, Elapse times in ms and outcome (Hit or Miss). Joystick force or movement signals were recorded as time-series data at 83 Hz. Results. With targets combined, using trials with "Hits", alpha 0.05, a within subjects ANOVA detected an effect between joysticks. Tukey post-hoc analysis found no significant difference between the Traverse times of the VGA and MSJ joysticks. It did reveal a significant difference between the FSJ and the MSJ; the FSJ had significantly longer Traverse times. Discussion. Results suggest the VGA algorithm may be an improvement over the FSJ. Boxplot graphics imply user performance with the VGA improves with practice. Median Traverse time scores of the VGA and MSJ appear virtually identical by the third presentation but could not be verified by a Kruskal-Wallis analysis. Future studies with multiple trial sessions could determine if the VGA performance would exceed the MSJ with more practice.