Data-driven identifier–actor–critic learning for cooperative spacecraft attitude tracking with orientation constraints

• Published in: Automatica (Oxford) Vol. 173; p. 112035
• Main Authors: Xia, Kewei, Wang, Jianan, Zou, Yao, Gao, Hongbo, Ding, Zhengtao
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.03.2025
• Subjects: Control barrier function; Distributed control; Orientation constraint; Reinforcement learning; Spacecraft attitude; Distributed control; Orientation constraint; Control barrier function; Spacecraft attitude; Reinforcement learning
• ISSN: 0005-1098
• DOI: 10.1016/j.automatica.2024.112035

This paper investigates the cooperative attitude tracking issue of a cluster of spacecraft subject to orientation constraints. In particular, all the involved spacecraft cooperatively adjust their attitudes to track a time-varying reference via local information exchange while constraining them inside a mandatory orientation zone as well as outside forbidden orientation zones. A dynamic identifier is first exploited to compensate for the dynamics uncertainty. Next, by integrating the sliding mode with the dynamic identifier, a distributed actor–critic reinforcement learning (RL) control algorithm is designed. Moreover, a data-driven online learning algorithm is proposed for the update of the learning weights, which effectively relieves the typical persistent excitation (PE) to the finite excitation (FE). To overcome the orientation constraint dilemmas, a robust control barrier function (CBF) based quadratic programming optimization is designed. It is shown that the attitude tracking errors are ultimately driven to a small tunable neighborhood of origin without violating the underlying orientation constraints. Finally, simulation results validate and highlight the proposed theoretical results.