A kinesthetic Turing test for robotic systems

As the physical interaction of robots with humans becomes more common, there is a growing need to design robots that are tactilely perceived as humanlike. One approach of implementing a human-like mechanical impedance is to simulate models of the human neuromuscular system. However, the necessary level of complexity of these models is unknown. The goal of this study is to determine the least complex neuromuscular model which can achieve human-like mechanical impedance. We have developed an objective test of a human subject´s ability to tactilely distinguish a robot and human, the kinesthetic Turing test. The results from six subjects show that a robot with linear viscoelastic properties could be distinguished from a human wrist joint. A robot simulating a nonlinear Hill muscle model was also distinguishable, but subjects´ ability to discriminate was significantly less. These results are important for the design of robots emulating human-like qualities.