Empowering All-in-Loop Health Management of Spacecraft Power System in the Mega-Constellation Era via Human-AI Collaboration

TL;DR

The paper tackles SPS health management in the satellite mega-constellation era by introducing SpaceHMchat, a human-AI collaboration framework guided by the AUC principle to enable all-in-loop health management across work condition recognition, anomaly detection, fault localization, and maintenance decision-making. It validates the approach with a hardware-realistic SPS fault-injection platform and the XJTU-SPS dataset, demonstrating high performance across 23 metrics and showing strong capability for end-to-end HM in a scalable, interpretable manner. Key contributions include the AUC-guided HAIC design, an open AIL HM dataset for SPS, and open-source code and simulation models that facilitate replication and further research. The work highlights significant potential for HAIC-enabled SPS HM in the SMC era, while acknowledging limitations and outlining future directions toward fully autonomous agents under the SUC paradigm and expanded aerospace knowledge management.

Abstract

It is foreseeable that the number of spacecraft will increase exponentially, ushering in an era dominated by satellite mega-constellations (SMC). This necessitates a focus on energy in space: spacecraft power systems (SPS), especially their health management (HM), given their role in power supply and high failure rates. Providing health management for dozens of SPS and for thousands of SPS represents two fundamentally different paradigms. Therefore, to adapt the health management in the SMC era, this work proposes a principle of aligning underlying capabilities (AUC principle) and develops SpaceHMchat, an open-source Human-AI collaboration (HAIC) framework for all-in-loop health management (AIL HM). SpaceHMchat serves across the entire loop of work condition recognition, anomaly detection, fault localization, and maintenance decision making, achieving goals such as conversational task completion, adaptive human-in-the-loop learning, personnel structure optimization, knowledge sharing, efficiency enhancement, as well as transparent reasoning and improved interpretability. Meanwhile, to validate this exploration, a hardware-realistic fault injection experimental platform is established, and its simulation model is built and open-sourced, both fully replicating the real SPS. The corresponding experimental results demonstrate that SpaceHMchat achieves excellent performance across 23 quantitative metrics, such as 100% conclusion accuracy in logical reasoning of work condition recognition, over 99% success rate in anomaly detection tool invocation, over 90% precision in fault localization, and knowledge base search time under 3 minutes in maintenance decision-making. Another contribution of this work is the release of the first-ever AIL HM dataset of SPS. This dataset contains four sub-datasets, involving 4 types of AIL HM sub-tasks, 17 types of faults, and over 700,000 timestamps.

Figures (10)

• Figure 1: SpaceHMchat: a Human-AI collaboration framework for AIL HM, guided by the AUC principle. (a) The future belongs to the SMC era, in which the explosive growth in spacecraft numbers introduces challenges, such as manpower cost explosion, textual information deluge, and operational complexity. (b) AIL HM encompasses work condition recognition, anomaly detection, fault localization, and maintenance decision-making, which are respectively performed by space mission operators, telemetry data analysts, senior subsystem experts, and on-orbit response teams. (c) This work proposes the AUC principle as the design principle of SpaceHMchat. The core idea is to first identify the intrinsic nature of a given subtask, such as recognizing anomaly detection as a tool dependent task, then determine the underlying human capabilities required to accomplish it, such as tool operation and professional skills, and finally align these human capabilities with the corresponding capabilities of LLMs, such as function calling and tool integration, on the basis of which a HAIC framework is developed. (d) The developed HAIC framework, SpaceHMchat, enables human-in-loop learning and interaction with a low usage barrier, while delivering high quality and efficiency AIL HM, allowing senior experts to focus on more challenging tasks and the iterative improvement of SpaceHMchat, while routine tasks are rapidly and automatically executed by the copilot under assistant commands. (e) To validate the effectiveness and necessity of SpaceHMchat, both simulation models and a hardware-realistic fault injection experimental platform are established, resulting in the AIL HM dataset covering four subtasks. And all resources are open-sourced.
• Figure 2: The interaction results of work condition recognition task. Our proposed SpaceHMchat's dialogue results are shown on the left, while the right side shows the response of a base model without any techniques mentioned in Figure \ref{['Fig_Flow diagram of our method']}. An example of AIL HM dialogue with complete information is shown in \ref{['Appendix: Complete Chat Examples']}.
• Figure 3: Results of SpaceHMchat on Anomaly Detection Task. Our proposed SpaceHMchat's dialogue results are shown on the left, while the right side shows the response of a base model without any techniques mentioned in Figure \ref{['Fig_Flow diagram of our method']}. An example of AIL HM dialogue with complete information is shown in \ref{['Appendix: Complete Chat Examples']}.
• Figure 4: Results of SpaceHMchat on Fault Localization Task. Our proposed SpaceHMchat's dialogue results are shown on the left, while the right side shows the response of a base model without any techniques mentioned in Figure \ref{['Fig_Flow diagram of our method']}. An example of AIL HM dialogue with complete information is shown in \ref{['Appendix: Complete Chat Examples']}.
• Figure 5: Results of SpaceHMchat on Maintenance Decision-making Task. Our proposed SpaceHMchat's dialogue results are shown on the left, while the right side shows the response of a base model without any techniques mentioned in Figure \ref{['Fig_Flow diagram of our method']}. An example of AIL HM dialogue with complete information is shown in \ref{['Appendix: Complete Chat Examples']}.