Progressive haptic guidance for a dynamic task in a virtual training environment

This thesis presents the motivation for and implementation of a novel progressive haptic guidance scheme designed to improve the efficiency of a virtual training environment used for skill acquisition. A detailed expertise-based analysis of the dynamic human motor task identifies the key skills required for success and motivates the progressive haptic guidance scheme. The thesis compares the effectiveness of the scheme to similar visual guidance, written guidance and no-guidance. The experimental training protocol presents a target-hitting training task in a virtual environment that utilizes an LCD display for visual feedback and a force feedback joystick for haptic interactions. This protocol lasts eleven sessions over a two-month period, thereby ensuring the performance saturation of participants. During each session, the number of target hits obtained becomes the objective measure of performance. Two additional measures, trajectory error and input frequency, are defined and implemented to calculate the performance of participants in two key skills. The guidance scheme then employs these last two measures as gain inputs to the guidance controller, which in turn progressively diminishes the forces that display guidance as virtual walls. The haptic controller design initially restricts a participant's motion to a preferred task path, but increased performance results in decreased guidance from one trial to the next. In addition to these measures, the protocol also presents the computerized version of the NASA Task Load Index (TLX) to all participants at each session, thereby providing cognitive workload measurements throughout the entire training period. The results demonstrate that this progressive haptic guidance scheme, one that integrates key skills and measures of performance, significantly outperforms three other guidance modes early on in the training and only when guidance is active. The data failed to show whether the haptic guidance scheme has significantly higher performance when the guidance is inactive. This scheme also generates less frustration and mental workload than visual guidance. Possible applications for these findings include virtual training environments designed for surgery and rehabilitation