Trajectory prediction of space robot for capturing non-cooperative target

Abstract

The trajectory tracking and prediction of space non-cooperative target has been the key technique in on-orbit capturing tasks of space intelligent robot. The difficulty of long-term prediction not only because visual measuring equipment may be obstructed by robotic arm but the noise statistics unknown in complex space environment. This paper proposes a robust hybrid Time/Frequency domain algorithm based on vision system. An Extended Finite Impulse Response (EFIR) filter estimates the states related to rotation and the Discrete Fourier Transform (DFT) filter estimates the dynamics parameters related to translation. Subsequently, we apply the estimated states and parameters to dynamic equations of free-floating object and then achieve the long-term prediction of the motion. An experiment with ground robot is presented to verify the correctness and effectiveness of the proposed method. The results show that the motion of space non-cooperative target can be predicted rapidly and accurately.