Learning-based control of a spacecraft with sloshing propellant

One of the major needs of present and future spacecraft is to fulfil highly demanding pointing requirements when performing attitude manoeuvres without losing stringent control over their flexible parts. For long-duration missions, the propellant mass is a significant portion of the overall mass budget of the satellite. The interaction between the fluid and the tank walls can lead to instability problems and even to mission failure if not properly accounted for in the design phase and control synthesis. The combined liquid-structure dynamic coupling is usually extremely difficult to model for a space system. An equivalent mechanical system is then desirable to carry out a computationally-efficient simulation of the liquid behaviour inside the tank. In this paper, the 3-D model of a spacecraft equipped with flexible appendages and tanks containing liquid propellant is presented. A learning-based control strategy using on-orbit available data is designed to improve the attitude tracking precision for repetitive on-orbit manoeuvres, compensating for cyclic disturbances such as liquid fuel sloshing effects. A co-simulation procedure between MSC Adams and Simulink is then carried out to test the performance of the controller. The effectiveness of the proposed control strategy is analysed and discussed and conclusions are presented.