Abstract
This paper introduces a computer-assisted teleoperation system, where the control over the teleoperator is shared between a human operator and computer assistance in order to improve the overall task performance. Two units, an action recognition and an assistance unit are introduced to provide contextspecific assistance. The action recognition unit can evaluate haptic data, handle high sampling rates, and deal with human behavior changes caused by the actived haptic assistance. Repairing of a broken hard drive is selected as scenario and three different taskspecific assistance functions are designed. The overall computerassisted teleoperation system is evaluated in two steps: first, the performance of the action recognition unit is evaluated and then, the performance of the integrated computer-assisted teleoperation system is compared with an unassisted system by means of a user study with 15 participants. Overall action recognition rates of about 65% are achieved. Multivariate paired comparisons show that the computer-assisted teleoperation system significantly reduces the human effort and damage possibility compared with a teleoperation system without assistance.