Fusing Both RSSI and Distance Signal for UWB Indoor Localization via Cross Attention Mechanism

The ranging measurements and received signal strength indication (RSSI) are two signals commonly used in UWB localization and exhibit good complementarity. To address the challenges of efficiently combining RSSI signals and ranging measurements with different data distributions for UWB indoor localization, we propose a cross-attention-based UWB localization network (CAUL-Net), which effectively extracts and learns non-line-of-sight (NLOS) environmental information from RSSI signals, and seamlessly integrates two modalities, significantly enhancing the accuracy of UWB localization in NLOS scenes. Specifically, two parallel CNNs are used to extract features from each signal, followed by a cross-attention-based multi-modal fusion module that integrates the extracted features from both signals. The fused features are then processed by an LSTM module for temporal feature extraction and position prediction. Additionally, we demonstrate the advantages of the proposed method through three real-world experiments.