TR2023-098

DEEP REINFORCEMENT LEARNING FOR STATION KEEPING ON NEAR RECTILINEAR HALO ORBITS

    •  Suda, T., Shimane, Y., Elango, P., Weiss, A., "DEEP REINFORCEMENT LEARNING FOR STATION KEEPING ON NEAR RECTILINEAR HALO ORBITS", AIAA/AAS Astrodynamics Specialist Conference, August 2023.
      BibTeX TR2023-098 PDF
      • @inproceedings{Suda2023aug,
      • author = {Suda, Takumi and Shimane, Yuri and Elango, Purnanand and Weiss, Avishai},
      • title = {DEEP REINFORCEMENT LEARNING FOR STATION KEEPING ON NEAR RECTILINEAR HALO ORBITS},
      • booktitle = {AIAA/AAS Astrodynamics Specialist Conference},
      • year = 2023,
      • month = aug,
      • url = {https://www.merl.com/publications/TR2023-098}
      • }
  • MERL Contact:
  • Research Areas:

    Control, Dynamical Systems, Machine Learning

Abstract:

In this work, we develop and evaluate a soft actor-critic (SAC) deep reinforcement learning (DRL) policy for station keeping of a spacecraft on a near-rectilinear halo orbit (NRHO) in the full-ephemeris dynamics. Monte Carlo simulations show that the DRL-based NRHO station-keeping policy maintains an approximately linear increase in delta-v at the apolune of each revolution, with a low spread in the delta-v gradient across the samples