Channel Charting for UAV Navigation in RIS-Assisted ISAC Systems

Reliable uncrewed aerial vehicle (UAV) navigation in global navigation satellite system denied environments is essential for low-altitude economy applications such as urban air mobility and infrastructure monitoring. Traditional navigation methods often suffer from signal blockage, drift, or sensitivity to environmental changes. Integrated sensing and communication offers a promising alternative by leveraging communication signals for joint localization and connectivity. However, most existing methods rely heavily on dense ground-truth labels or strong geometric assumptions. This article proposes a semisupervised channel charting framework for UAV navigation, enhanced by reconfigurable intelligent surfaces to improve channel diversity and signal coverage. A uniform manifold approximation and projection guided Siamese neural network is employed to construct channel charts that preserve both local and global geometry in the channel state information (CSI), enabling accurate mapping to physical coordinates with minimal labeled data. We develop a full-stack simulation platform using Sionna’s ray tracing in realistic urban environments. Experimental results show that our method achieves a localization accuracy of 3.66 m at 100-meter UAV altitude using only 20% labeled CSI data. The approach outperforms existing fingerprinting and unsupervised charting methods and remains robust under both line-of-sight (LoS) and non-LoS conditions.