MoRe-T2 (mobility research trajectory tracker): validation and application

Abstract

It is important to assess the suitability of mobility aids before prescribing them to patients. This assessment is often subjectively completed by a therapist and it often includes a variety of basic practical tests. An objective assessment of a patient's capability, which captures not only speed of task completion and success, but also accuracy and risk of manoeuvres, would be both a fairer and safer approach. Yet until now such an assessment would have been cost-prohibitive, especially in low resource settings. We pave the way towards this end goal, by describing, validating and demonstrating a low-cost computer vision based system called MoRe-T2 (mobility research trajectory tracker). The open-source MoRe-T2 system uses low-cost off-the-shelf webcams to track the pose of fiducial markers, which are simply printed onto regular office paper. In this article, we build upon previous work and benchmark the accuracy of MoRe-T2 against an industry standard motion capture system. In particular, we show that MoRe-T2 achieves accuracy comparable to CODA motion tracking system. We go on to demonstrate a use case of MoRe-T2 in assessing wheelchair manoeuvrability over a relatively large area. The results show that MoRe-T2 is scalable at a much lower cost than typical industry-standard motion trackers. Therefore, MoRe-T2 can be used to develop more objective and reliable assessments of mobility aids, especially in low-resource settings.

Comparing shared control approaches for alternative interfaces: A wheelchair simulator experiment

“…PSC reduced collisions the most for the sip-puff switch, which was the most difficult interface to user [19]. Linear blending however, reduced the distance travelled and task duration the most for all interfaces.…”

“…Since the sip-puff switch is a two-switch interface and thus more difficult to use than the head-array, which is a threeswitch interface, we let σ sip-puff switch >σ head-array [19].…”

“…Our assumption is another way of saying we completely trusted the user's input so that σ → 0. We made this assumption for simplicity since the joystick has a high resolution of control and thus can be used to indicate intended motion with high accuracy [19]:…”

Probabilistic vs linear blending approaches to shared control for wheelchair driving

“…Therefore, we constructed a local occupancy grid of the space around the wheelchair, which spanned an area of 4 m × 4 m with a fixed grid resolution of 0.05 m × 0.05 m. The local map used log-odd update to populate the occupancy grids [15]. We updated the position of occupied grids using odometry, which was calibrated with our low-cost tracking toolkit, MoRe-T2 [16]. Occupied grids had a forgetting factor of 5 seconds, after which they become unoccupied [17].…”

“…For the objective evaluation, we used the standard metrics: distance travelled; task completion time; clearance and agreement [16]. Clearance is defined as the average minimum distance of the wheelchair from all obstacles for the duration of the task.…”