SENSOR FUSION AND TEMPORAL INTEGRATION FOR TOUCH INTERFACE INDOOR POSITIONING

RAMADHAN, HANI (2016) SENSOR FUSION AND TEMPORAL INTEGRATION FOR TOUCH INTERFACE INDOOR POSITIONING. Masters thesis, Institut Teknologi Sepuluh Nopember.

[img]
Preview
Text
5113201907-Master_Theses.pdf

Download (3MB) | Preview

Abstract

Dalam kunjungan wisata atau budaya, panduan terhadap objek menarik sangat berguna untuk menambah pengetahuan dan pengalaman pengunjung di lokasi tersebut. Dewasa ini, dengan bantuan teknologi modern, aplikasi bergerak mampu menjadi pemandu wisata mandiri otomatis dengan sistem sadar konteks. Kebanyakan, unsur konteks yang digunakan dalam aplikasi-aplikasi ini adalah posisi dua dimensi (2D). Meskipun begitu, ada beberapa kemungkinan lain agar tiap unsur konteks dari perangkat pintar ini dapat diteliti lebih lanjut. Berkat sensor dari ponsel pintar, konteks-konteks tersebut, yang terdiri dari konteks dalam 3 dimensi (3D) dari posisi dan orientasi (dalam sumbu X, Y, dan Z), dapat ditangkap oleh ponsel pintar. Dimensi-dimensi ini akan diteliti untuk mendapatkan kemungkinan keberhasilan digunakannya ponsel pintar yang digenggam sebagai pointer terhadap objek menarik. Hal ini dilakukan karena posisi 2D tidak bisa menangani konteks ketinggian. Sehingga, pengalaman pengguna dapat ditingkatkan karena mereka tidak terhalang secara visual dan audio. Tetapi, sensor-sensor ini memiliki galat pengukuran yang tinggi. Sehingga, suatu penggabungan sensor diterapkan untuk menangani galat tersebut. Penelitian ini menerapkan metode untuk memperkirakan orientasi sudut dan posisi dengan berbagai filter, yakni Complementary Filter dan Kalman Filter. Complementary Filter melibatkan gyroscope, magnetometer, dan accelerometer dari sensor inersial ponsel pintar. Sedangkan, Kalman Filter melibatkan accelerometer dan hasil Wi-Fi fingerprinting yang didapatkan dari pengamatan lingkungan. Evaluasi perkiraan-perkiraan hasil penggabungan observasi sensor oleh filter-filter tersebut menggunakan ilustrasi grafis dan evaluasi statistika untuk mengukur kualitas reduksi galat dari tiap filter. Hasil dari performa filter menunjukkan bahwa kualitas perkiraan orientasi oleh Complementary Filter cukup baik untuk menghasilkan sudut yang sesuai. Namun, perkiraan posisi oleh Kalman Filter menunjukkan hasil yang kurang baik akibat integrasi ganda terhadap derau dan pengaruh besar Wi-Fi fingerprinting. Hasil Wi-Fi fingerprinting menunjukkan perkiraan posisi yang tidak akurat. Hal ini menunjukkan bahwa perkiraan posisi tidak dapat digunakan dalam penelitian ini. Sedangkan, dalam percobaan menunjuk objek di laboratorium, perkiraan orientasi sudut memberikan hasl yang cukup baik dengan ponsel pintar. Secara ringkas, perkiraan posisi dan orientasi 3D dengan Complementary Filter dan Kalman Filter dalam ponsel untuk pointer tidak dapat digunakan menurut penelitian ini. Meskipun begitu, masih perlu diteliti mengenai penerapan filter lainnya untuk perkiraan posisi dan observasi lain untuk membantu perkiraan yang baik. Walaupun penggunaan filter dan observasi lain dapat mengorbankan sumber daya dari ponsel pintar. ======================================================================================================== During cultural or tourism visits, a guide of the interesting objects is useful to enhance the knowledge and the experience of the visitors. Nowadays, because of the modern technologies, mobile applications are capable to be a personal autonomous guide in the case of context-aware system. Mostly, the context element used in these applications is the position in two dimension (2D). However, there are more possibilities using the context elements from smartphone that can be explored. Thanks to smartphone sensors, the contexts which can be captured by smartphone are composed in 3 dimensions (3D) of both position and orientation (in X, Y, and Z axes). Those dimensions are used to explore the feasibility of smartphone which can held by hand as pointer to interesting objects, which can’t be handled by 2D position only. Thus, the user experience can be enhanced, as they don’t get vision-blocked or audio-blocked. However, those sensors have erroneous measurements. Hence, a sensor fusion is applied to overcome this drawback. The sensor fusion can be implemented not only using the internal smartphone sensors, but also the external environment. In this case of indoor environment, the Wi-Fi fingerprinting approach, which widely used as indoor positioning algorithm, can be considered as external observation. Even though so, the quality of the fusion should be studied to assure that it is feasible to use smartphone a pointing device in indoor environment. This study proposed a method to estimate orientation and position using different filters, namely Complementary Filter and Kalman Filter respectively. The complementary filter involves the gyroscope, magnetometer, and accelerometer from the smartphone inertial navigation sensors, while the Kalman Filter involves accelerometer and the Wi-Fi fingerprinting result which come from environmental measurement. To evaluate these estimations, the graphical representation and statistical evaluation are used to measure the filters’ quality in reducing the errors. The results of the filters’ performance showed that orientation estimation was adequate to give acceptable angle. But, unfortunately, position estimation had resulted in poor performance because of the double integration toward noise and the heavy influence from Wi-Fi fingerprinting. The Wi-Fi fingerprinting resulted inaccurate positioning. This concluded that the position estimation cannot be used at all in this study. In laboratory object pointing field experiment, the orientation estimation gave passable estimation to locate an object by a fixed smartphone position. To sum up, the 3D position and orientation estimation using Complementary Filter and Kalman Filter might not be feasible according to this study. However, regarding to 3D position estimation, possibly there are other methods than Kalman Filter which might be used as state estimator. And also, there are various external measurements which might help to achieve better estimation. Although, the drawbacks between the more sophisticated methods and the computation power and capability of smartphone should be considered for a good user experience.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Context-aware systems, Indoor positioning, Wi-Fi fingerprinting, sensor fusion
Subjects: T Technology > T Technology (General)
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105 Data Transmission Systems
Divisions: Faculty of Information Technology > Informatics Engineering > (S2) Master Theses
Depositing User: Mr. Fandika aqsa
Date Deposited: 09 Jan 2017 02:09
Last Modified: 27 Dec 2018 06:19
URI: http://repository.its.ac.id/id/eprint/1384

Actions (login required)

View Item View Item