基于CSI的工业互联网深度学习定位
WiFi-based indoor localization is a vital part of applications that rely on the positions of terminals.In recent years, the learning-based methods has shown their strong latent capacity of high reliability and accuracy. However, they are more or less environment dependent since the received data inevitably contain environment features that perhaps interfere the learning process. A localization model trained by the data collected in a specific and static environment does not work well if it is applied to predict the position in a new or varying situation. Slight changes of physical environment may cause the learned model mismatch, leading to large positioning errors.Thus, the paper focus on this challenge and address the environment dependent problem of indoor localization. We construct an adversarial network that can dissociate the target features and environment features, where the position is obtained from target features only. The proposed method is validated by experimental WiFi data collected in different rooms or in the same room with changes. The localization performance is independent to the environment.