Neural Network-Accelerated Trajectory Optimization for Launch Vehicle Landing

This paper presents a novel trajectory optimization method for the 6-degrees-of-freedom powered landing problem in aerospace guidance and control. The method combines machine learning and convex optimization to achieve real-time performance. Specifically, we formulate the powered landing problem as an optimal control problem and transform it into a convex optimization problem. To enhance the state-of-the-art sequential convex programming (SCP) algorithm, we use a deep neural network as an initial trajectory generator to provide a satisfactory initial guess for the SCP algorithm. Simulation results show that the proposed method achieves precise guidance of the vehicle to the landing site. Monte Carlo tests demonstrate that it can save an average of 40.8% of the computation time compared to the SCP method. Therefore, the proposed scheme is suitable for real-time applications in the aerospace industry.