The influence of human-robot interaction order during fast lifting tasks for different levels of weight compensation

When human operators employ robotic lifting aids, haptic feedback about the lifted object is important. In an experimental study we manipulated two factors that influence haptic feedback to the operator: the percentage of compensated weight, and the way the lifted object is held: robocentric (the human hand lifting the robot that holds the object) or anthropocentric (the robot lifting the human who holds the object). We hypothesize that directly holding the object (anthropocentric approach) will improve the realized trajectories when rapidly lifting partly compensated weights. Subjects (n=10) performed a fast semi-repetitive lifting task, lifting a 4kg object to a designated target in either an anthropocentric or robocentric lifting scenario, at different levels (50%-75%-95%) of weight compensation. The anthropocentric approach yielded significantly smaller mean over- or under- shoot compared to robocentric lifting, especially for the first trials. The difference increased for higher levels of compensation. We conclude that for fast lifting, the anthropocentric approach better helps subjects to estimate the required forces to move the weight to the target, especially for unexpected movements at high levels of compensation.