From barren to beautiful: A pattern-matching localization scheme integrating heterogeneous network data

Among all the wireless localization techniques, the Wi-Fi pattern-matching scheme is one of the most widely used approaches, which estimates the user's location by comparing his/her device's received signal strength (RSS) against a pre-trained radio map on the fly. The pattern-matching solutions have an inherent drawback: the expensive calibration operation of war-driving. In order to reduce the calibration operation cost of war-driving, many solutions based on the community approach have been proposed [1]-[3], which collect the Wi-Fi training data from users' contributions. However, most works only consider the Wi-Fi patterns. For the heterogeneous training data, such as a radio map combined signals from both Wi-Fi and cellular networks, a method shows how to collect and exploit the data through an ongoing way is still missing. In this paper, our objective is to grow a heterogeneous radio map from barren to beautiful over a large area, such as a regional area or a national area. Based on the concept of community approach, we propose a geography-based method to combine the cellular and the Wi-Fi radio maps. We believe that our framework can provide a valuable solution for pattern-matching localization which shows how to effectively build a radio map and quickly estimate the user's location with acceptable distance errors.