Real-time guidance for powered landing of reusable rockets via deep learning

Wang, Jinbo; Ma, Hongjun; Li, Huixu; Chen, Hongbo

doi:10.1007/s00521-022-08024-4

Cited by 9 publications

(5 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In a recent study [8], the authors proposed a real-time guidance policy named DCRNG (Deep Classification and Regression Network-based Guidance), which employs two DNNs to establish a nonlinear mapping between the ideal state and control pairings for a 3-degree-of-freedom motion model similar to the framework used in our study. (While they utilized three translational kinematic equations, we opted for two translational kinematic equations combined with one rotational dynamic equation.)…”

Section: Discussionmentioning

confidence: 99%

“…Optimal control problems can be solved to generate fuel-optimal trajectories. However, the computational complexity of solving Optimal Control Problems (OCPs) makes on-board (online) implementation impractical [7], [8]. To overcome this challenge, extensive research has been conducted in the field of the convexification of OCPs, which allows on-board computations.…”

Section: A Literature Reviewmentioning

confidence: 99%

“…Even in studies such as [13], where successive convexification of a 6-degree-of-freedom model for Mars landing was achieved, simplified model assumptions were made, such as using a constant inertia matrix during landing, which is unrealistic during fuel consumption. To address these computational limitations, [8] discussed the use of low-frequency guidance command generation with a high-fidelity 6-degree-offreedom model. The authors employed a successive convexification methodology at the beginning of the powered landing phase and used a reference tracking controller to follow the trajectory throughout the landing.…”

Section: A Literature Reviewmentioning

confidence: 99%

“…The utilization of DNN and recurrent neural networks (RNN) has been observed in the context of addressing reusable rocket landings and planetary landing problems. In [8], the authors proposed a real-time guidance policy called Deep Classification and Regression Network-based Guidance (DCRNG), which employs two DNNs to establish a nonlinear mapping between the ideal state and control pairings. This was achieved by fitting the optimal trajectory data generated by solving OCPs.…”

Section: A Literature Reviewmentioning

confidence: 99%

See 3 more Smart Citations

Optimal Reusable Rocket Landing Guidance: A Cutting-Edge Approach Integrating Scientific Machine Learning and Enhanced Neural Networks

Çelik,

Demirezen

2024

IEEE Access

View full text Add to dashboard Cite

This study presents an innovative approach that utilizes scientific machine learning and two types of enhanced neural networks for modeling a parametric guidance algorithm within the framework of ordinary differential equations to optimize the landing phase of reusable rockets. Our approach addresses various challenges, such as reducing prediction uncertainty, minimizing the need for extensive training data, improving convergence speed, decreasing computational complexity, and enhancing prediction accuracy for unseen data. We developed two distinct enhanced neural network architectures to achieve these objectives: Adaptive (AQResNet) and Rowdy Adaptive (RAQResNet) Quadratic Residual Neural Networks. These architectures exhibited outstanding performance in our simulations. Notably, the RAQResNet model achieved a validation loss approximately 300 times lower than the standard architecture with an equal number of trainable parameters and 50 times lower than the standard architecture with twice the number of trainable parameters. Furthermore, these models require significantly less computational power, enabling real-time computation on modern flight hardware. The inference times of our proposed models were measured in approximately microseconds on a single-board computer. Additionally, we conducted an extensive Monte Carlo analysis that considers a wide range of factors, extending beyond aerodynamic uncertainty, to assess the robustness of our models. The results demonstrate the impressive adaptability of our proposed guidance policy to new conditions and distributions outside the training domain. Overall, this study makes a substantial contribution to the field of reusable rocket landing guidance and establishes a foundation for future advancements.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: A Literature Reviewmentioning

confidence: 99%

Section: A Literature Reviewmentioning

confidence: 99%

Section: A Literature Reviewmentioning

confidence: 99%

See 2 more Smart Citations

Optimal Reusable Rocket Landing Guidance: A Cutting-Edge Approach Integrating Scientific Machine Learning and Enhanced Neural Networks

Çelik,

Demirezen

2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The application of neural networks to optimal control tasks has already received some attention in the aerospace field, especially for mapping time-optimal [12][13][14] or fuel-optimal [12,[15][16][17] control solutions for landing [12,[14][15][16][17] or orbital transfer problems [13]. Another area of application is the trajectory planning for hypersonic reentry [18][19][20].…”

Section: Abbreviationsmentioning

confidence: 99%