Kalman Filtering for Relative Spacecraft Attitude and Position Estimation

This thesis is framed within the context of the ESA Space Rider Observer Cube (SROC) mission. In this in-orbit servicing mission, a CubeSat is deployed from the Space Rider’s cargo bay to carry out a visual inspection of the vehicle, followed by a rendezvous and docking maneuver to return to the cargo bay. This work covers the development of an Extended Kalman Filter (EKF) to be implemented in the Navigation system of SROC for the phases of relative navigation (i.e under 1 km). The filter shall estimate the relative position, velocity, attitude and angular velocity between the Chaser and Target spacecraft. The accuracy of the position estimation is driven by the mission requirements, and is quantified as single-axis error of 0.4m in the range [40; 20] m, 0.2m in the range [20; 10] m, 0.1m in the range [10; 5] m and 0.02m in the range [5; 1] m. For the other estimates, an acceptability criterion in terms of absolute error is considered based on the estimated quantity (velocities and angles), e.g. the misalignment shall be less than 1 deg. The design of the filter is carried out by means of simulation, inside the framework of a Complete Flight Simulator developed on Matlab® \& Simulink® software. To achieve the Navigation task, first a camera model was developed to simulate the measurements done by SROC’s camera, expressed by a 12 × 1 vector containing a pair of 2D coordinates on the image plane for each of the 6 key points present on Space Rider’s body. The new state vector estimate is given by the Kalman filter that operates using this measurement, the state transition matrix and the measurement matrix. The simulation is carried out for a nominal case scenario and a worst case scenario in terms of measurement noise in order to assess the robustness of the solution. The results on relative position estimation comply with the requirements, for both study cases. The estimations of attitude and velocities are also satisfactory, while showing worst results for the noisier case study. Overall, the implementation of the EKF for relative position and attitude estimation is considered successful.