Human-robot interaction is one of the most important topics in nowadays robotics research. Up to now, interfaces such as control panels have been used to communicate with robots. However, modern robots like the DLR Lightweight Robot III were especially designed for interactive and cooperative applications. Its joint torque sensing enables to physically interact with humans. Other novel sensors like tracking sensors or the Microsoft Kinect allow to observe human movements or their behavior, respectively.
Basically, humans are able to predict future actions of other persons and use this information to decide on own actions. The aim of this work is to analyze human anticipation and transfer this knowledge to robot control. Movements of human beings shall be analyzed and learning algorithms be found to implement reaction strategies on robotic systems.
Hema S. Koppula and Ashutosh Saxena: Anticipating Human Activities using Object Affordance for Reactive Robotic Response