Showing papers by "Dongsoo Han published in 2019"

City Radio Map Construction for Wi-Fi-Based Citywide Indoor Positioning

Abstract: A Wi-Fi fingerprint consists of received signal strength values at a particular location along with the location information. A Wi-Fi radio map is constructed by maintaining a database of Wi-Fi fingerprints at various points in a building. A city radio map is a collection of radio maps for most buildings in a city. A highly precise citywide indoor positioning service is possible, if an accurate city radio map is available. However, tremendous time and effort are required to construct a city radio map through manual calibration. This paper proposes a method for constructing a city radio map through the crowdsourcing of Wi-Fi fingerprints from numerous smartphones. The proposed method classifies the buildings in a city into three categories: buildings in residential areas, commercial areas, and public areas. Then, it develops location-labeling techniques appropriate for the collected fingerprints from buildings in each category. Experiments conducted in the cities of Daejeon and Seoul revealed that the proposed method can construct a precise city radio map with minimal cost in a short period of time. Once city radio maps are constructed for most cities around the world, the global indoor positioning system will be completed.

Automatic Radio Map Construction Exploiting Mobile Payments

Abstract: Radio map construction automation by location-labeling of crowdsourced fingerprints is drawing a great attention these days. It allows radio maps of most of buildings in cities to be constructed at a very low cost. This paper proposes an adaptive semi-supervised location-labeling method for the crowdsourced fingerprints. The method is distinguished from the existing semi-supervised learning methods in that it uses address-labeled fingerprints collected during offline mobile payments for its location references. Despite inexactly specified location references, the method finds an optimal placement of location-unlabeled fingerprint sequences by varying the locations of address-labeled fingerprints. When the proposed method was evaluated at three large-scale landmark buildings in Seoul, the effectiveness of using location references collected during mobile payments for the proposed adaptive semi-supervised location-labeling method was apparent. Highly precise radio maps could be constructed for the buildings without any manual calibration efforts. The method can be used to automatically construct radio maps for most downtown buildings.

Neural Network for Predicting Error of AP Location Estimation Method Using Crowdsourced Wi-Fi Fingerprints

Abstract: RSS values observed from a smartphone are related with distances to each AP. Therefore, AP locations can be estimated when enough number of location-labeled Wi-Fi fingerprints are obtained. Since manually collecting Wi-Fi fingerprints costs human labor, crowdsourcing approach is preferred. Crowdsourced Wi-Fi fingerprints usually need an additional step to tag a location label. The low accuracy of indirectly acquired location labels affects the result of AP location estimation. Therefore, some AP locations need to be discarded if the error of estimated AP location is high. To measure the error, it is necessary to survey the ground truth of AP location. Since surveying true AP locations also costs human labor, an error prediction method is helpful. We propose the neural network that predicts the error of an estimated AP location. The performance of the proposed method was tested on KAIST N1 building, Cheongju airport, and Lotte World mall.

Detecting Arrivals and Departure of Subway Train Using Linear Accelerometer (poster)

Abstract: Subway is one of the public transport which carries people at the exact time. In the metropolis all over the world, it transfers countless people in the rush hour. Since there is no traffic jam in the subway, accuracy is one of the most crucial characters of the subway. However, subway trains are often delayed by some reasons like an accident. The delay can make many people confused and inconvenient. In this context, the need for dynamic timetable model which corrects the error of timetable in real time has emerged. Since the method to detect train moving is necessary to modify timetable, various solutions are proposed in the indoor positioning way, such as Wi-Fi fingerprint, magnetometer and so on. A method using Wi-Fi fingerprint is a primary way in indoor positioning, but it is tough to build a radio map for every single station. Shin et al. suggested a solution using a magnetometer, with simple judging criteria named 'decision peak.' This research proved that a magnetometer is a practical solution. Nonetheless, it can not detect the train moving in some cases. We propose a new method to detect the subway moving using a linear accelerometer, based on the essence of the problem.

Rover Who Make Indoor Radio Map (poster)

Abstract: The fingerprinting-based indoor positioning technique requires a database, i.e., a radio map, containing indoor scenes through the training phase. Since that offline phase is labor intensive, numerous studies are underway to minimize the effort. In this paper, we present a concept that applies to a collaborative crowdsourcing method. We designed a system that relies on people and robots moving in the indoor space to construct a radio map. Experiments at two testbeds provide proof of the concept and are the result of the last step of the proposed system.