Imitation learning-based spacecraft rendezvous and docking method with Expert Demonstration

TL;DR

The paper tackles autonomous $6$-DOF spacecraft rendezvous and docking by learning control policies directly from expert demonstrations, reducing reliance on accurate dynamics. It introduces IL-SRD, a Transformer-based imitation learning framework augmented with a variational autoencoder, an anchored decoder target to enforce physical consistency, and a temporal aggregation mechanism to stabilize long-horizon predictions. The approach predicts control sequences rather than single actions and is trained with a composite loss that includes reconstruction and KL terms, enabling smooth, model-free deployment. Empirical results show IL-SRD outperforms traditional controllers and several DRL baselines, matching MPC in energy efficiency while maintaining robustness under unknown disturbances, demonstrating practical potential for on-orbit operations. The work highlights a scalable path toward reliable, model-free autonomous docking, with future work aimed at improving terminal precision and exploring a direct state-to-action formulation.

Abstract

Existing spacecraft rendezvous and docking control methods largely rely on predefined dynamic models and often exhibit limited robustness in realistic on-orbit environments. To address this issue, this paper proposes an Imitation Learning-based spacecraft rendezvous and docking control framework (IL-SRD) that directly learns control policies from expert demonstrations, thereby reducing dependence on accurate modeling. We propose an anchored decoder target mechanism, which conditions the decoder queries on state-related anchors to explicitly constrain the control generation process. This mechanism enforces physically consistent control evolution and effectively suppresses implausible action deviations in sequential prediction, enabling reliable six-degree-of-freedom (6-DOF) rendezvous and docking control. To further enhance stability, a temporal aggregation mechanism is incorporated to mitigate error accumulation caused by the sequential prediction nature of Transformer-based models, where small inaccuracies at each time step can propagate and amplify over long horizons. Extensive simulation results demonstrate that the proposed IL-SRD framework achieves accurate and energy-efficient model-free rendezvous and docking control. Robustness evaluations further confirm its capability to maintain competitive performance under significant unknown disturbances. The source code is available at https://github.com/Dongzhou-1996/IL-SRD.

Figures (6)

• Figure 1: Network structure of IL-SRD
• Figure 2: Overview of the on-orbit spacecraft rendezvous and docking control
• Figure 3: Distribution of expert demonstrations
• Figure 4: Demonstration of temporal aggregation
• Figure 5: Spacecraft rendezvous and docking error diagrams for different control methods, (a)--(c) are the relative distance, (d)--(e) are the relative velocity, (g)--(j) are the relative attitude represented by quaternions, (k)--(m) are the relative angular velocities