A Survey on Data Augmentation for WiFi Fingerprinting Indoor Positioning

WiFi fingerprinting has been a prominent solution for indoor positioning, yet its dependence on labor-intensive data collection and susceptibility to environmental dynamics are on-going major challenges. Thus, this article presents a comprehensive survey and analysis of the data augmentation techniques designed to enhance WiFi fingerprinting datasets, focusing on the efficiency in data construction and the robustness in positioning accuracy. We reviewed more than 70 studies and proposed a novel taxonomy that categorizes existing methods into six groups: traditional (e.g., interpolation and perturbation), propagation models, machine learning, deep learning, hybrid approaches, and other emerging techniques. Our quantitative analysis correlates key metrics, such as input data size, synthetic data volume, and augmentation ratios, with positioning performance. We found that traditional methods achieved notable performance enhancements with minimal computational overhead. Surprisingly, deep learning models became less efficient when generating more data, particularly when the synthetic data exceeded a threefold ratio over the input samples. Our findings provide actionable guidance for selecting data augmentation strategies and bridge the gap between theoretical advancements and practical deployment for WiFi fingerprinting dataset enhancement.