ESA GNC Conference Papers Repository
A Novel Attitude Estimation Filter for the PLATO Space Mission based on Moving Horizon Estimation
PLATO (PLAnetary Transits and Oscillations of stars) is a proposal submitted in response to the ESA Cosmic Vision 2015 - 2025 announcement of opportunity. Its objective is to detect and characterize exoplanets by means of their transit signature in front of a large sample of bright stars, and measure the seismic oscillations of their parent stars. In order to achieve the science goals, very stringent requirements on the pointing error must be met. This precise and stable pointing requires an accurate estimation of the spacecraft attitude. This estimation is based on different onboard attitude sensors, which are subject to measurement noise and bias. Some sensors, e.g. sun sensors and magnetometers, provide only vector observations and the estimation of the spacecraft attitude is further complicated by the nonlinear spacecraft dynamics and quaternion kinematics. This paper presents the moving horizon estimation of the spacecraft state, and shows that it outperforms the extended and unscented Kalman filters, which are the current workhorse filters in the space industry. A moving horizon estimator determines the current state by solving a constrained numerical optimization problem considering a finite sequence of current and past measurement data, an available spacecraft dynamics model and quaternion kinematics constraints. A symplectic integrator is developed to preserve the quaternion norm over the estimation horizon. The onboard sensors run at different sampling rates. This can be dealt with in a natural way in the moving horizon estimator. The objective function to be minimized is typically a trade-off between minimizing measurement noise and process noise. In order to solve this constrained optimization problem in real-time, an efficient numerical solution method based on the iterative Gauss-Newton scheme has been implemented and specific measures are taken to speed up the calculations such as hot starting from previous solutions and exploiting the sparsity and band structure of matrices to be inverted. A technology demonstration for this attitude estimation algorithm is planned on one of the QB50 cubesats, to be launched in 2013. This will be used to verify on-orbit performance before implementation in the PLATO flight software. Pre-launch performance validation of the technique has been carried out in a simulation environment for both missions. Efficient C++ code for implementation in the onboard attitude control software is developed and current research is aimed towards real-time estimation of the different misalignment parameters.