Instability Detection during Physical Human Robot Interaction Based on Wavelet Decomposition

Physical human robot interaction is an important research topic in robotics. Admittance control has been widely used for collaborative tasks because of its compatibility with low-cost robots. However, stability issue arises especially when the interacting environment (including humans) becomes stiff. A common way to guarantee stability is changing control parameters, especially the damping parameters, but it will cause heavy react during interaction. Therefore an effective way to detect the instability and change parameters only when oscillation occurs is necessary. In this paper, we analyzed the frequency of normal operation and oscillation during interaction process, and proposed a real time instability detection method based on wavelet decomposition. Experiments on human interaction with an admittance-controlled robot were performed and the effectiveness of the proposed instability detection method was confirmed.