Quadratic interpolation based on robot trajectory denoising for human motion imitation

Telerobotic has been introduced into complex environment to imitate human dexterous motion, such as remote surgeries on medical sites. However, presence of noise often leads to non-smooth target trajectory, causing robot to vibrate severely. Although most high-end robotic arms have facilitated torque control with torque sensors, it is costly and hard to realize widely in different scenarios. For the purpose of inexpensive and easy realized method, this paper proposes a model free based approach, which is divided as follows: The first part is “signal denoising” by applying empirical mode decomposition to target signals and smoothing out some of components with Savitzky–Golay filter. The second part is “quadratic interpolation” by applying down sampling operation on target signals to minimize robot vibration, adding more sample points to fit original signals, and then interpolating with quadratic polynomial functions. Through this scheme of signal processing, we generate simpler target signals that resemble original signals with piecewise constant acceleration values. The results are also provided to demonstrate the improvement of several input parameter values during experiments. In this paper, an approach is presented to match telerobotic target signals for human motion imitation to robotic arms kinematic motion.