Mars powered descent phase guidance law based on reinforcement learning for collision avoidance
Mars powered descent phase guidance law based on reinforcement learning for collision avoidance
This paper proposes a reinforcement learning-based guidance law for Mars pow- ered descent phase, which is an effective online calculation method that handles the nonlinearity caused by the mass variation and avoids collisions. The reinforcement learning method is designed to solve the constrained nonlinear optimization problem by using a critic neural network. Specifically, to cope with the position constraint (i.e. glide-slope constraint) and the thrust force limit constraint, a modified cost function is proposed, and the associated Hamilton-Jacobi-Bellman equation is solved online without using an actor neural network, which significantly reduces the computational burden. The convergence of the critic neural network is proven. Simulation results show the effectiveness of the proposed method.
collision avoidance, constraints, deep space exploration, powered descent phase
10378-10392
Zhang, Yao
a4f30318-ab42-4b38-a60d-f7199ff3a02a
Zeng, Tianyi
5d0025da-2f29-4130-bddc-9d29e3c0c2b6
Guo, Yanning
9fd3825e-0c6b-45e3-a7a6-ddbd0b0ac70a
Ma, Guangfu
dd3ce7d7-b1e5-445a-a6f2-5107024dd80f
25 November 2023
Zhang, Yao
a4f30318-ab42-4b38-a60d-f7199ff3a02a
Zeng, Tianyi
5d0025da-2f29-4130-bddc-9d29e3c0c2b6
Guo, Yanning
9fd3825e-0c6b-45e3-a7a6-ddbd0b0ac70a
Ma, Guangfu
dd3ce7d7-b1e5-445a-a6f2-5107024dd80f
Zhang, Yao, Zeng, Tianyi, Guo, Yanning and Ma, Guangfu
(2023)
Mars powered descent phase guidance law based on reinforcement learning for collision avoidance.
International Journal of Robust and Nonlinear Control, 33 (17), .
(doi:10.1002/rnc.6651).
Abstract
This paper proposes a reinforcement learning-based guidance law for Mars pow- ered descent phase, which is an effective online calculation method that handles the nonlinearity caused by the mass variation and avoids collisions. The reinforcement learning method is designed to solve the constrained nonlinear optimization problem by using a critic neural network. Specifically, to cope with the position constraint (i.e. glide-slope constraint) and the thrust force limit constraint, a modified cost function is proposed, and the associated Hamilton-Jacobi-Bellman equation is solved online without using an actor neural network, which significantly reduces the computational burden. The convergence of the critic neural network is proven. Simulation results show the effectiveness of the proposed method.
Text
wileyNJD-AMA
- Accepted Manuscript
More information
Accepted/In Press date: 15 February 2023
e-pub ahead of print date: 26 February 2023
Published date: 25 November 2023
Additional Information:
Funding Information:
National Natural Science Foundation of China, Grant/Award Numbers: 61876050; 61973100; 62203219 Funding information
Funding Information:
This work is supported by National Natural Science Foundation of China under grants 62203219, 61973100, 61876050.
Publisher Copyright:
© 2023 John Wiley & Sons Ltd.
Keywords:
collision avoidance, constraints, deep space exploration, powered descent phase
Identifiers
Local EPrints ID: 475974
URI: http://eprints.soton.ac.uk/id/eprint/475974
ISSN: 1049-8923
PURE UUID: 77e7a1f5-4ff4-4c45-b9ca-360dfeabac01
Catalogue record
Date deposited: 03 Apr 2023 16:39
Last modified: 06 Jun 2024 04:16
Export record
Altmetrics
Contributors
Author:
Yao Zhang
Author:
Tianyi Zeng
Author:
Yanning Guo
Author:
Guangfu Ma
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics