Transferring from the Simulator to a Live Robotic Environment: The Effectiveness of Part-Task and Whole-Task Training

Abstract

Abstract : Part-task training methods are widely used in training when the full target task is considered too complex or impractical for training as a whole task. However, part-task training has had mixed success in transfer to the whole task. Limited research has focused on part tasks that do not interact but are performed concurrently in the whole task. The present research was designed to help address this gap in the context of an Army-relevant task and training situation: teleoperating a robotic device to detect and identify vehicles. The experiment involved training in a simulated environment and transfer to performance with a live robotic system. Training methods included part- and whole-task training in the simulation environment and part-task training in the live environment. Results indicated a significant benefit in vehicle mobility skills for the condition receiving part-task training both in the simulated and live environment. Participants in this training condition also reached transfer performance criteria in fewer scenarios, and the benefit to mobility persisted in the final freeform scenario. Additionally, part-task training in the live environment, prior to the transfer scenarios, was more efficient than other conditions, resulting in less time spent training on the live robotic system.