Assistive teleoperation for manipulation tasks

摘要: How should a human user and robot collaborate during teleoperation? The understands the full semantics of task: they know, for example, what search in cupboard, or that it be more careful when moving near glass water than box tissues. Since might not have this knowledge, allowing to operate fully autonomously may risky; its model is incomplete policy wrong. On other hand, teleoperating through every motion slow tiresome, especially on difficult tasks. Between these two extremes lies spectrum, from almost no assistance at all (very timid) autonomy aggressive). So appropriate level assistance? And how do factors like task difficulty correctness affect decision?