RF fingerprinting based GSM indoor localization

Abstract

OzetçeAlışveriş merkezi, havaalanı gibi binaiçi ortamlarda konumtabanlı servislere olan ihtiyaç giderek artmaktadır. Bu tür servisleri desteklemek için konum kestirim metodları gerekmektedir. Bu çalışmada, binaiçi ortamlarda mobil kullanıcının konumunu tespit etmek için alınan sinyal gücü (RSS) tabanlı RF parmakizi yöntemine dayalı K-Nearest Neighbor (K-NN) ve random decision forest (rastgele karar ormanı, RKO) iki algoritma sunulmuştur. Bu algoritmaların başarımlarını gerçek binaiçi ortamlarda elde etmek ve karşılaştırmak için Istanbul Ataşehir'deki Palladium alışveriş merkezinde GSM ag altyapısı kullanılarakölçümler yapılmıştır. Toplanılan GSM alınan sinyal gücü verileri kullanılarak her iki algoritma için CDF konum kestirim hata sonuçları elde edilmiştir. Yine, buölçüm verileri kullanılarak farklı iki telefon modeli olan Sony Ericsson ve Nokia telefonlarının konum kestirim hatasıüzerindeki etkileri araştırılmıştır. Elde edilen sonuçlara göre RKO algoritması K-NN algoritmasından az da olsa daha iyi başarım saglamıştır. Ayrıca, Sony Ericsson telefon modeli Nokia telefon modelinden daha iyi konum tespit başarımı göstermiştir. Abstract The demand for location-based services (LBS) in indoor environments such as shopping malls and airports has increasedrecently. In order to support such LBS applications accurate indoor localization systems are required. Therefore, in this paper, K-Nearest Neighbor (K-NN) and Random Decision Forest (RDF) algorithms for GSM RSS based RF fingerprinting method are presented in order find the location of mobile users in indoor environments. For studying the performance of these two algoritms in realistic indoor environments, a measurement campaign is conducted in Istanbul Ataşehir Palladium shopping mall using GSM cellular networks. The location estimation error performance of these two algoritms are obtained in the form of CDF results by using the collected GSM RSS data. Moreover, the effects of different mobile phone brands (Sony Ericsson and Nokia) on the location estimation error performance are investigated using the measurement data. According to the results, RDF method performs slightly better than K-NN method. Additionally, Sony Ericsson mobile phone provides better location estimation performance than that of Nokia mobile phone.

FloorLoc-SL: Floor Localization System with Fingerprint Self-Learning Mechanism

“…This objective of this module is to provide a current activity state of mobile phone user for floor counting module. Due to the fact that FloorLoc-SL is a localization system on mobile phone platform without any support from back-end server, the activity classification hit ← 0 (6) for each access point from top ap list do (7) if access point is in ap list of floor fingerprint then (8) hit ← hit + 1 (9) end if (10) end for (11) if ℎ MAX = hit do (12) current floor level ← UNKNOWN (13) else if ℎ MAX < hit do (14) ℎ (2) floor fingerprint ← fingerprint of floor in fingerprint database (3) else do (4) floor level ← floor (5) ap list ← null (6) floor fingerprint ← floor level, ap list (7) end if (8) for each access point from top ap list do (9) if access point is not in ap list of floor then (10) ap list.add(access point) (11) end if (12) end for Algorithm 2: Fingerprint algorithm.…”

“…After initializing this parameter, the algorithm waits until a difference of atmospheric pressure can reach to the thresholds for one-floor (6) previous pressure ← current pressure (7) activity state ← DOWN (8) else if Δ ≤ up then (9) previous pressure ← current pressure (10) activity state ← UP (11) else then (12) activity state ← UNCHANGED (13) end if (14) ← + window size (15) end for Algorithm 4: Activity classification using atmospheric pressure approach. …”

“…Many indoor localization systems have been proposed to locate a mobile device in a building [1][2][3][4][5][6][7][8][9][10][11][12]. The previously proposed systems apply fingerprint-based approach [1-4, 6, 7, 9-12].…”

Utilization of room-to-room transition time in Wi-Fi fingerprint-based indoor localization

“…To calculate distance metrics and similarity rates, fingerprinting methods are used in combination with classification algorithms such as k-Nearest Neighbors [7] [8] [9], artificial neural networks [8] [13] [20], Support Vector Machines [8] and Naïve Bayes Classifiers [8].…”

“…Among these techniques, fingerprint-based methods make use of identifiers (e.g. BSSID for Wi-Fi signals) of signal sources present in the environment (such as GSM [2] [3] [9], Wi-Fi [7] [8] [10] [12] [16], Bluetooth [13] [14] [15] and FM radio [11] [17] [18] [19]), and signal strength values. This information is recorded at different locations in the environment and stored in a dataset together with the corresponding locations.…”