Calibration-Free Indoor Positioning Using Crowdsourced Data and Multidimensional Scaling

Indoor positioning plays an important role in various location-based services (LBSs). In conventional systems, the process of constructing radio maps for positioning usually involves labor-intensive manual calibrations, which seriously limits the system's scalability and adaptiveness. In this paper, we propose an efficient calibration-free method by leveraging on crowdsourced WiFi signal data that are captured passively through a WiFi sensing testbed. Since the ground truths of the crowdsourced data are unavailable, the radio maps cannot be directly constructed. In the proposed method, we adopt the multidimensional scaling (MDS) technique to compute the positions of the unlabeled data thereby generating radio maps. In order to enable MDS, we estimate the pairwise distances among the unlabeled data by using an improved trilateration method and a law of cosine (LoC)-based geometrical algorithm without online pairwise measurements. Experimental results show that the accuracy of the proposed method is higher than trilateration-based method and reasonably lower than that of calibration-based method. Meanwhile, the run time of the proposed method is shorter than previous optimization-based methods. The short run time allows the radio maps to be dynamically updated against the environmental variations.