Crowdsensing based Bluetooth Radio Map Reconstruction for Indoor Localization

At present, the widespread popularity of wireless mobile terminals has led to the emergence of a new perception mode in the field of the Internet of Things, namely "crowdsensing". A large number of mobile terminal users, as the basic unit of the crowdsensing network, can collaborate to complete complex social perception tasks. This article uses crowdsensing network to construct the indoor radio map. However, users participating in crowd sensing have hotspot tendencies and site constraints, which means that user trajectories cannot cover all detection areas, resulting in some areas having only a small amount of wireless signal information or even no information. Therefore, this article aims to propose a radio map reconstruction method that can utilize the small and incomplete wireless signal information reported by crowdsensing users to complete the inference and reconstruction of the entire radio map. In order to achieve high-quality radio map reconstruction, we utilize the sparsity prior of wireless signal information structure, introduce the Fourier sparsity loss factor to constrain the structural complexity of reconstructed signal samples, and combine deep feature extraction model with self representation learning framework, to propose a radio map reconstruction method based on a small number of samples. In order to verify the effectiveness of the proposed method, we constructed an indoor Bluetooth positioning system in the school's college building, using ordinary mobile devices for signal acquisition and map construction. By comparing with traditional methods, we verified the improvement effect in reconstruction performance of the proposed method.