Robust Booster Landing Guidance/Control

Sammanfattning: The space industry and the technological developments regarding space exploration hasn’t been this popular since the first moon landing. The privatization of space exploration and the vertical landing rockets made rocket science mainstream again. While being able to reuse rockets is efficient both in terms of profitability and popularity, these developments are still in their early stages. Vertical landing has challenges that, if neglected, can cause disastrous consequences. The existing studies on the matter usually don’t account for aerodynamics forces and corresponding controls, which results in higher fuel consumption thus lessening the economical benefits of vertical landing. Similar problems have been tackled in studies not regarding booster landings but regarding planetary landings. And while multiple solutions have been proposed for these problems regarding planetary landings, the fact that the reinforcement learning concepts work well and provide robustness made them a valid candidate for applying to booster landings. In this study, we focus on developing a vertical booster descent guidance and control law that’s robust by applying reinforcement learning concept. Since reinforcement learning method that is chosen requires solving Optimal Control Problems (OCP), we also designed and developed an OCP solver software. The robustness of resulting hybrid guidance and control policy will be examined against various different uncertainties including but not limited to wind, delay and aerodynamic uncertainty.