Probabilistic Modeling of Reaction Force/Torque through Fourier Transform and Entropy Analysis

In this paper, we propose a method to improve the recognition performance of a probabilistic model through entropy analysis after transforming time-varying reaction force/torque signals into frequency components. To perform tasks that require physical interaction, it is important for robots to recognize reaction force/torque during the interactions between robots and environments. However, the reaction force/torque measured by F/T sensor contains a lot of noise signals due to the sensitivity of the sensor. Therefore, the recognition performance depends on noise signals included in training and/or test dataset. To solve this problem, the reaction force/torque signals are transformed from time domain into frequency domain by fast Fourier transform. Then, some task-relevant frequency components are selected based on their entropy analysis, after which they are used to learn a hidden Markov model. To evaluate our proposed method, several robot manipulation tasks are performed using an open dataset including reaction force/torque signals: approaching, transferring, and positioning.