Adaptive signal-correction-based identification for friction perception of the vibration-driven limbless robot

Friction determines the locomotion performance of the limbless robot. However, most sensors cannot measure the friction precisely because the invasive assembly will change the contact feature between the robot and the ground. This paper, in another way, proposes a novel method that only uses the acceleration signal to realize the accurate identification of friction and structure parameters. Note that the friction force and the structural coupling are functions of velocity and displacement. The proposed method first introduces two constant coefficients to construct a linear algebraic transform, which maps the measured acceleration to the uncertain velocity and displacement in the orthonormal polynomial space. The introduced two coefficients are then identified together with the friction and structure parameters via particle swarm optimization. Because the identification is integrated with coefficient correction of uncertain velocity and displacement, this architecture is called the adaptive signal-correction-based identification. Numerical and experimental examples suggest that if the polynomial order makes the signal correlation larger than 0.98 and the time-window width is approximately one oscillation period, the proposed method possesses a high identification accuracy. Further robustness discussions on the parameter uncertainty and noise disturbance demonstrate that the identification architecture also owns good reliability for engineering applications.