Human Prediction of Robot’s Intention in Object Handling Tasks

Abstract

Trained human workers can predict the intentions of other workers from observed movement patterns when working collaboratively. The intentions prediction is crucial to identify their future actions. In human-machine teams, predictable movement patterns can enhance the interaction and improve team performance. In this article, we investigated the effects of different robot trajectory characteristics on the early prediction performance in human-machine teaming and on perceived robot’s human-likeness. The results showed that humans can predict the robot’s intention quicker and more accurately when the observed robot’s trajectory was generated with relatively lower energy expenditure. We found that the amount of jerk and acceleration in the robot’s joint-space affected perceived robot’s human-likeness.