EUVIMEDEUVIMED entdecken
Crossref

ADAPTIVE SPACECRAFT RELATIVE ON-OFF CONTROL VIA R EINFORCEMENT LEARNING WITH STABILITY GUARANTEES

S. V. KHOROSHYLOV, V. V. SOROCHINSKII

Kosmìčna nauka ì tehnologìâ · 09.06.2026 · 32 (3) · 003-016

DOI: 10.15407/knit2026.03.003 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Kosmìčna nauka ì tehnologìâ, online erschienen am 09.06.2026. Verfasst von S. V. KHOROSHYLOV und 1 weiteren Autor:innen. Die Arbeit nennt 65 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Spacecraft relative motion control under on-off TH constraints presents challenges for achieving required performance in terms of accuracy, effi ciency, and robustness. Th is study introduces a stability-guided deep reinforcement learning framework that directly generates TH-fi ring policies without relying on thrust modulators. Th e proposed agent integrates a sliding-mode control structure with an explicit discrete-time Lyapunov-decrease penalty into the proximal policy optimization algorithm, ensuring stability while enabling effi cient learning. Numerical simulations demonstrate that the method satisfi es established performance requirements, including settling time, steady-state accuracy in both position and rate, and propellant effi ciency. Th e sliding-mode structure enhances the interpretability of the learned policies and provides a principled basis for selecting initial agent parameters, thereby improving sample effi ciency during training. Furthermore, the agent exhibits strong robustness to parameter uncertainties and maintains reliable performance across varying operating conditions. Its adaptability is further underscored by the capacity for progressive improvement through online learning. Th ese fi ndings highlight the potential of stability-guided deep reinforcement learning as a viable and eff ective approach for spacecraft relative control under discrete actuation constraints, off ering both theoretical rigor and practical applicability for future space missions.

Zitieren

APA

KHOROSHYLOV, S. V., & SOROCHINSKII, V. V. (2026). ADAPTIVE SPACECRAFT RELATIVE ON-OFF CONTROL VIA R EINFORCEMENT LEARNING WITH STABILITY GUARANTEES. Kosmìčna nauka ì tehnologìâ, 32(3), 003-016. https://doi.org/10.15407/knit2026.03.003

Vancouver

KHOROSHYLOV SV, SOROCHINSKII VV. ADAPTIVE SPACECRAFT RELATIVE ON-OFF CONTROL VIA R EINFORCEMENT LEARNING WITH STABILITY GUARANTEES. Kosmìčna nauka ì tehnologìâ. 2026;32(3):003-016. doi:10.15407/knit2026.03.003

BibTeX
@article{KHOROSHYLOV2026,
  author = {KHOROSHYLOV, S. V. and SOROCHINSKII, V. V.},
  title = {ADAPTIVE SPACECRAFT RELATIVE ON-OFF CONTROL VIA R EINFORCEMENT LEARNING WITH STABILITY GUARANTEES},
  journal = {Kosmìčna nauka ì tehnologìâ},
  year = {2026},
  volume = {32},
  number = {3},
  pages = {003--016},
  doi = {10.15407/knit2026.03.003}
}
Kostenlos starten →