On-line estimation of Jacobian matrix based on particle filter

Proposes a new method of estimating Jacobian matrixes on-line for image-based robot visual servo systems. A vector is firstly formed from the elements of a Jacobian matrix, and the problem is converted into one of state-estimation. Particle filtering suitable for non-liner non-Gaussian systems is utilized to solve the Jacobian estimation problem. The proposed method and the one based on Kalman filtering are tested to track a moving target on a two-degree-of-freedom system with non-Gaussian noise. The results showed the effectiveness and the robustness of the proposed method. System calibrations can be avoided and no specification on system noises is needed.