Landing a rocket on Earth is a key factor in enabling quicker and more cost-effective access to space. However, it poses significant challenges due to the highly uncertain environment. A robust, reliable, and real-time capable Guidance, Navigation, and Control (GNC) system is essential to guide the vehicle to the landing site while meeting terminal constraints and minimizing fuel consumption.. A 6-Degrees Of Freedom (DOF) flight simulator is developed, including accurate vehicle and environmental models such as variable Mass, Center of Mass and Inertia (MCI), winds, and detailed aerodynamics. Furthermore, not only thrust magnitude and engine deflections are used as controls, but also two sets of aerodynamic fins, as they are present in real rockets. Finally, initial condition dispersions and uncertainties in dynamics, navigation, and controls are included, to assess the robustness of the GNC strategy.
This study applies Meta Reinforcement Learning (meta-RL) to the terminal rocket landing problem. Through repeated interactions between an agent and its environment over multiple episodes, a Neural Network (NN) develops a policy that maps observed states to control actions. Unlike traditional methods, meta-RL leverages both current and past observations to output the optimal action, enhancing the robustness of the policy. Recurrent Neural Networks (RNNs) like Long-Short Term Memory (LSTM), are commonly used in meta-RL for their ability to handle sequential data. However, this thesis investigates also the use of attention-based Gated Transformer XL (GTrXL) networks, which are promising to improve the solution accuracy.
This research confirms this hypothesis, demonstrating that the GTrXL-based policy significantly outperforms the LSTM-based one. This policy is also less sensitive to internal hyperparameters, provided that the NN has a sufficient number of weights and biases to capture all the relevant features. The results indicate that incorporating complex vehicle and environmental models, along with dispersed initial conditions, aerodynamic and wind uncertainties, navigation and control errors, and actuator deflection rate constraints into the training process, yields a robust GTrXL-based policy. This guidance policy is able to meet terminal constraints on position, horizontal velocity, vertical angle, and angular rates in 1000/1000 simulations. The only exception is the vertical velocity which is exceeded on average by only 2-3 m/s. Including a thrust rate constraint mitigates this issue, reducing the average violation to 1 m/s. Finally, developing a terminal patch to increase the thrust magnitude in the last few meters before touchdown completely eliminates the vertical velocity issue, producing a 100% success rate, with all the Monte Carlo runs meeting all the terminal constraints. Moreover, the NN produces a solution in about 6 ms, showing great potential for its real-time use. The results are consistently superior to those of a conventional Guidance and Control (G&C) strategy where a Linear Quadratic Regulator (LQR) controller tracks an optimal trajectory, consuming only 6% more fuel.
Recommendations for future works include improving the reward function to better handle the vertical velocity constraint, and penalizing control effort to have smoother Thrust Vector Control (TVC) and fins control profiles. It is also suggested to expand the simulation scenario, including the unpowered aerodynamic descent.

With the onset of the age of space travel, asteroid missions have been steadily gaining interest. The pristine nature of asteroids due to their preserved state since the formation of the Solar System is an opportunity to unravel many mysteries about the Solar System. Also with the ever-growing need for resources, asteroids prove to be a plentiful source. With the discovery of asteroids in the close vicinity of our planet and a probable threat to the preservation of life, a need for defence missions has also arisen. The ever-growing need for better computational speed and accuracy has led to the development of new representations for attitude and position of the spacecraft in the past. The usual methods of representations of position and attitude (pose) are the Cartesian coordinates and quaternions. A recent development is the simultaneous representation of the pose of the spacecraft using dual quaternions which are eight-dimensional vectors. In this thesis, an attempt to incorporate these state-of-the-art guidance technologies to asteroid missions has been made. The novelty in this thesis is, using dual quaternions for SC pose and attitude representation to autonomously guide the spacecraft for mapping an asteroid and perform a touch and go descent using sampling-based motion predictive optimisation and successive convexification respectively. After the development and verification of these algorithms, different scenarios have been designed to find out the robustness of the dynamic successive convexification method. The scenarios have been run by both the Cartesian quaternion and dual quaternion based algorithms. They are found to behave similarly in their results, besides the latter being computationally more expensive as proved by different theses so far. The algorithm performs better than the one developed by Szmuk et al. (2017), but faces difficulties with badly scaled problems as is the nature of missions to asteroids. The software used for the thesis is ECOS, which faces numerical problems in these mission scenarios even when the problem is feasible and has an optimal result. The availability of a solution to the optimal control problem depends on the scaling of the penalty weights used in the cost function to penalise virtual controls and trust region, which are used to prevent infeasibility and bound the problem, respectively. It also depends on finding an appropriate final time for the descent along with the number of nodes for discretisation. This research proves, that successive convexification indeed provides a speedy solution for an autonomous precision descent but needs further work to make it robust and stable. The outcome of this thesis is to carry out further research to understand the complex relations, between the scale of the problem, simulation parameters for the optimal control problem and final time to make the algorithm robust and safe for an autonomous mission. Another important future research prospect is to incorporate the sampling based model predictive optimisation, for the SC to autonomously map the target body and help in selecting a landmark for a descent.