Domain Generalization in Ultra-Wideband Ranging Error Mitigation

Description

Ultra-wideband technology enables highly precise indoor localization. However, its accuracy degrades significantly in non-line-of-sight conditions, where obstacles block the direct signal path. Such obstructions lead to overestimation of the transmitter-receiver distance due to signal reflections and reduced propagation speed as the signal passes through materials. Machine learning models have been successfully applied to mitigate such ranging errors.

However, their generalization to previously unseen environments, which is crucial for real-world deployment, remains a challenge. While transfer learning has been shown to improve predictive performance, it relies on labeled data collected in the target domain. This time-consuming process becomes increasingly impractical when deploying a large number of indoor localization systems across diverse environments. To address this limitation, we explore the potential of semi-supervised learning and domain adversarial training to improve generalization and enable model adaptation to new environments.

We also study the effect of incorporating unlabeled data from the target domain, which is usually much easier to obtain. This provides insights into enhancing the cost-effectiveness of deploying indoor localization systems in a variety of real-world scenarios.