Advances and challenges of AI-driven wireless indoor positioning

Wireless indoor positioning was deeply integrated into numerous scenarios including transportation navigation, industrial manufacturing, and public safety, serving as a crucial pillar for ubiquitous sensing in the 6G era. However, positioning accuracy was severely degraded by non-line-of-sight propagation and multipath characteristics in indoor environments, while its robustness was further undermined by environmental noise and interference. As artificial intelligence was deeply applied in wireless systems and 6G’s integrated sensing and communication capabilities continue to advance, new opportunities were identified to mitigate the aforementioned challenges. A task-oriented technical framework for AI indoor positioning was established, by which a profound progression "single-point location estimation" to "holistic spatial cognition" and further to "data-driven reverse optimization" was revealed across three task categories, namely improving positioning accuracy, enhancing environmental perception and generating positioning data. Subsequently, a comprehensive set of evaluation metrics tailored specifically for AI positioning systems was proposed, which highlighted the distinctive characteristics and multidimensional variations of AI-driven wireless positioning. Finally, critical challenges and future trends in AI-driven wireless indoor positioning technology were discussed, offering fresh insights for next-generation positioning technology advancement.