Guidance, Navigation, and Control for Very Small Asteroid Mitigation Missions

Airbus Defence and Space GmbH in Friedrichshafen is the project coordinator of the NEOShield-2 project whose purpose is to develop the technology needed for near-Earth object (NEO) impact prevention. This research is located within the scope of a kinetic impactor mission which typically consists of two spacecraft: impactor and observer. This master’s thesis focuses on the latter and starts from a preexisting translational Image Processing (IP) navigation to develop a guidance and control algorithm for the translational motion and a complete guidance, navigation and control (GNC) algorithm for the attitude one. The study focuses on bodies with diameter less or equal than 300 m and the asteroid 2001 QC34 is taken as a baseline. First, the main perturbations influencing the dynamics of the spacecraft are assessed, showing that the best strategy is to do near-inertial hovering. Due to navigation purposes, the objective is set to minimise the amount of manoeuvres needed to keep the spacecraft in the desired location. Classical methods, based on reflecting or inverting the velocity vector, all based on the concept of control box, are simulated showing limitations for long hovering periods. Two new strategies also based on dead-band control theory are shown, one operating at the nominal performance (1 manoeuvre per day) and the second reducing by half these ones (0.5 manoeuvres per day). However, both methods suffer very much from initial conditions and on orbit operations, this being the reason why a second approach is formulated. A method based on Lyapunov control function is pro- posed showing that it reduces the commanded manoeuvres a ∼ 25% with respect to the classical ones (0.6-0.8 manoeuvres per day). It was also found that increasing the control box improves the propellant consumption. For the Lyapunov case, an increment in size of ∼ 30% has savings of ∼ 8% of propellant. The study continues with the implementation of the attitude GNC which uses only star tracker measurements for attitude estimation. The absolute performance error (APE) was found to be 0.15 deg, or distributed per axis as [0.005, 0.01, 0.01] deg. This pointing performance is in line with the establish in other space missions such as Rosetta. The main sources of error were found to be the bias of the star tracker and the gyroscopic term coming from angular momentum of the reaction wheels. The latter source can be reallocated per axis changing the initial conditions of these ones. The APE was only lost when translational manoeuvres were performed. This was due to the torque induced into the spacecraft because of thrusters errors. These manoeuvres can cause that the star tracker reach its physical limitations. Finally, the momentum dumping of the reaction wheels has been studied. Due to the distribution of the propulsion system, a thruster management function needs to be implemented for thrust allocation. For this purpose, linear programming (LP) techniques are used. In particular, using a new approach, and combining machine learning techniques (neural networks) with a pulse-width pulse-frequency (PWPF) modulator. The study shows that although momentum dumping manoeuvres are within the sensor physical limitations, the fact of having both rotational and translational motion coupled and the errors induced by the PWPF modulator, cause that the spacecraft needs to also correct after each desaturation manoeuvre its translational motion. For coarse pointing the method might work, but for momentum dumping where the frequency of the manoeuvres is low, classical methods are preferred as no relevant savings in terms of computational speed can be achieved by the inclusion of DNNs.