Collision Detection of Flexible Manipulators Based on Unscented Kalman Filter and Numerical Evaluation of its Unscented Transformation

Abstract

This paper investigates Unscented Kalman Filter (UKF) based collision detection for flexible manipulators subject to distributed random disturbance and numerical analysis of its unscented transformation. The mathematical model of the flexible manipulator is described by the nonlinear stochastic system subject to random disturbance modeled by white Gaussian noise. In order to detect the occurrence of collision when the flexible manipulator is in operation, the UKF is employed as a state estimator for the nonlinear system corresponding to the flexible manipulator. The collision can be detected using the intensity of the innovation process of the UKF. The performance of the collision detection depends on the calculation of the sigma points using the unscented transformation. In the numerical simulations, we evaluate the unscented transformation of the flexible manipulator using the histogram. Furthermore, the performance of collision detection function is demonstrated.