Automatic Radio Map Adaptation for Robust Indoor Localization With Dynamic Adversarial Learning

Recently, deep-learning-based wireless localization has become one of the most promising technologies for intelligent location-based services. However, classical schemes have extracted the appropriate features to construct a static radio map without environmental adaptability, resulting in severe accuracy degradation. To address this issue, we propose a novel approach of robust indoor localization with dynamic adversarial learning, known as DadLoc, which realizes automatic radio map adaptation for accuracy improvement. DadLoc can incorporate multilevel robust factors underlying different fingerprint databases to develop a dynamic adversarial adaptation network, which can learn the evolving feature representation with the complicated environmental dynamics. Furthermore, we adopt the training strategy of prediction uncertainty suppression with source-target dynamic adversarial adaptation, which can enhance the location discriminability of the transferable feature representation. With extensive experimental results, the satisfactory accuracy over other comparative schemes demonstrates that the proposed DadLoc can achieve an average accuracy of $1.78\,\mathrm{m}$ with the robustness of indoor environmental dynamics.