A DNN-based WiFi-RSSI Indoor Localization Method in IoT

Indoor automatic localization technology is very important for the Internet of Things. With the development of wireless technology and the diversification of location service requirements, especially in complex indoor scenarios, users are increasingly demanding location-based services. Traditional Global Positioning System (GPS) location technology is difficult to solve some positioning problems in indoor environments, and WiFi is now available in most indoor environments. Therefore, using WiFi for positioning does not require additional deployment of hardware devices, which is a very cost-effective method. However, WiFi-based indoor positioning requires a large amount of data, so we can use artificial intelligence methods to analyze the data and obtain a positioning model. The traditional indoor positioning methods based on WiFi signals have some problems such as long positioning time and poor accuracy. In order to solve the above problems, this paper proposes an indoor localization method based on Deep Neural Networks (DNN) for WiFi fingerprint. In particular, a DNN-based WiFi-RSSI positioning method is proposed for indoor automatic localization. Besides, in the process of DNN training, a joint training method based on unsupervised learning and supervised learning is adopted and the special loss function is defined. Extensive experiments are carried out in both the UJIIndoorLoc public database and a real scenario, and a thorough comparison with several existing approaches indicates that the proposed scheme improves the localization accuracy on average.