DAN: A Domain-Level Attention Network for Cross-Environment UWB Indoor Localization

Ultra-Wideband (UWB) localization technology offers centimeter-level accuracy in indoor environments, but suffers from significant performance degradation when deployed in new environments due to Non-Line-of-Sight (NLOS) conditions and multipath effects. Existing domain generalization approaches primarily focus on extracting domain-invariant features while neglecting valuable domain-specific information. This paper proposes a domain generalization localization method based on a domain-level attention network (DAN) that enables highprecision localization in previously unseen environments without requiring additional online training. DAN dynamically adjusts features based on domain prototypes and incorporates an adaptive gating structure to regulate attention intensity, effectively integrating both domain-invariant and domain-specific information. Extensive experiments illustrate that the proposed method outperforms other deep learning approaches in localization accuracy in unseen environments.