一句话：提出 SHE 健康度与失效定义，并用 Transformer 自编码器预测、预防失效及构建数字孪生多智能体架构。
论文关注智慧人本生态系统（SHE，如智慧城市）的可靠性问题：SHE 由异构且独立定义的系统共存组成，即使各组成系统行为正确，整个 SHE 仍可能失效，而现有方法忽视了对 SHE 可靠性的保障。作者的目标是通过测试、预测并最终预防 SHE 失效来确保其可靠性，为此提出四项工作：(i) 以健康度和失效为核心精确定义 SHE 质量；(ii) 使用 Transformer 自编码器进行异常检测以预测 SHE 失效；(iii) 提出预防 SHE 失效并恢复正常状态的方法；(iv) 构建 SHE 数字孪生的多智能体架构。摘要未报告具体实验评估与结果。
Smart Human-centric Ecosystems (SHEs), like smart cities, emerge from the co-existence of heterogeneous systems with independently defined specifications. SHEs may fail despite the correct behavior of the systems that comprise the SHE, yet current approaches overlook the critical need of ensuring the reliability of SHEs. Our main goal is to ensure the reliability of SHEs by testing, predicting and ultimately preventing failures in SHEs. We propose (i) a clear and precise definition of the quality of SHEs in terms of healthiness and failures, (ii) a transformer-autoencoder anomaly detection approach to predict SHE failures, (iii) an approach to prevent SHE failures and restore the normal state of the SHE, and (iv), a multi-agent architecture for Digital Twins (DTs) of SHEs.