Robust Estimation of Motion States for Free-Floating Tumbling Target Capture

Abstract

In this paper, we propose a novel Extended Kalman Filter (EKF) to aid the capture of a free-floating tumbling satellite (Target) with a manipulator-equipped spacecraft (Servicer) in the close-range reach phase. For such a control problem, the interfacing of a fast-sampled robot controller with a slow-sampled Guidance, Navigation and Control (GNC)-bus on the spacecraft creates a down-sampling in terms of the measured state. This causes a performance loss for the robot controller. Additionally, only slow-sampled and noisy exteroceptive sensors, like a camera, which provides relative poses may be available for feedback. In order to circumvent this problem, the main objective of the proposed method is to provide fast relative state reconstruction between Target and Servicer. To this end, the proposed EKF estimates the inertial motion states of the Target and the Servicer-base at high rate using slow-sampled and noisy exteroceptive measurements, which include relative poses from cameras and absolute orientation from star/sun trackers. The state information is combined with the Inertial Measurement Unit (IMU), forward kinematics (using joint encoders) and a priori known transformations to reconstruct a fast-sampled estimate of the inertial states and hence, the relative states for control. As a robust validation, the results of 100 Monte-Carlo simulation runs are presented. Furthermore, the validity of the proposed EKF is demonstrated by driving it in closed-loop with a combined controller on a Guidance, navigation and control development environment software platform. Through these results, it is shown that the proposed EKF is robust towards camera occlusions and noise. Additionally, the proposed method also estimates the tumbling velocity of the Target for feed-forward in the control method.