An exploration of the utility of augmented haptic feedback for learning a curve-tracing task

University of Toronto, PhD Thesis. Published online July 28th, 2018.

URL: http://hdl.handle.net/1807/89670

Camille Kimberly Williams

ABSTRACT:
Technology for the field of robot-assisted rehabilitation (and haptic training, in general) is rapidly advancing. However, there is some concern among the rehabilitation community that the outcomes, in terms of motor recovery, have fallen short of expectations. One reason for this may be that–despite the understanding that motor learning is a key component of motor recovery–motor learning scientists are not always involved in the design, development and deployment of these systems. In this dissertation, I explored optimization of haptic feedback from a motor learning perspective. In the first study, I compared learning outcomes after participants practised a curve-tracing task with assistive, error-augmenting or no haptic feedback. In study two, I investigated whether learning outcomes were affected by the bandwidth (i.e., error tolerance) at which each of the two forms of haptic feedback were provided. For the third study, I explored whether and how self-controlled or experimenter-imposed assistive haptic feedback schedules influenced motor learning. In all experiments, motor learning was assessed using a composite measure of performance efficiency (the speed accuracy cost function) on retention tests conducted immediately (10 min) and 1 day after skill acquisition. Results showed that error-augmenting haptic feedback was the most effective for learning (study 1) but its superiority was not observed with a wider bandwidth (study 2). When learners could self-control their assistive haptic feedback schedule, a lower frequency of feedback across practice blocks appeared to enhance learning (study 3). This final study also showed that whether learners had self-control over their feedback frequency was associated with their views about the utility of assistive haptic feedback for performing and learning the task. Additionally, these expressed views were significantly associated with both self-chosen feedback frequency and motor learning outcomes. This research demonstrates that error-augmenting haptic feedback is indeed beneficial for tasks which rely on error-based learning mechanisms but that this effect is bandwidth-dependent. Further, it highlights the importance of and need for continuing research to explore how learners’ views about the training environment impact motor learning. These principles are useful for researchers and practitioners using haptic training in both basic and applied domains.