Data-driven predictive control for multi-arm spacecraft after target capture

This paper aims at realizing the post-operation maneuvering of multi-arm spacecraft after capturing non-cooperative targets. To solve the problem that the combination dynamic parameters are unknown, we propose a Manifold-constraint-based Subspace Identification Integral Predictive Control (MSI-IPC) method, which integrates a Manifold-constraint-based Subspace Identification (MSI) module with an Integral Predictive Control (IPC) scheme. By collecting historical input-state data of the multi-arm spacecraft, MSI extracts the inherent structural properties of the state equation parameters as structured constraints, thereby constructing a constraint manifold that enforces adherence to system invariants and enhances state prediction accuracy. IPC utilizes the control input increments as optimization variables within the predictive control problem, effectively suppressing control input chattering. The resulting MSI-IPC method is applicable to multi-arm spacecraft in arbitrary open-loop configurations without requiring knowledge of any dynamic parameters. It overcomes the limitations of conventional indirect subspace predictive control in high-dimensional, noisy systems, where input chattering often prevents convergence. A formal Input-to-State Stability (ISS) proof for the closed-loop system is provided via Lyapunov analysis. Numerical simulations under measurement noise demonstrate that MSI-IPC significantly surpasses conventional Indirect Subspace Predictive Control (ISPC) in both state prediction accuracy and control input smoothness, validating the method’s effectiveness and robustness.