Master of Science By Research

Permanent URI for this collectionhttp://192.248.9.226/handle/123/18682

Browse

Browsing Master of Science By Research by Subject "DATABASE CORRELATION METHOD"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    item: Thesis-Full-text
    Cellular positioning by location fingerprinting with the aid of propagation models
    Wijesinghe, WHMP; Dias, D
    The Fingerprinting method or the Database Correlation Method (DCM) is a network based positioning technique which has shown superior accuracy. DCM is based on a pre-measured database of location dependent variables such as Received Signal Strength (RSS). The major challenge of the technique is the effort involved in forming the database, which prevents it being deployed in large, dynamic networks. The work presented in this thesis investigates the possibility of using network planning tool predictions instead of field measurements to create the fingerprint database for DCM. While the accuracy of this approach is lower than the DCM method with field measurements, further tuning of the predictions in order to improve the performance is proposed. The tuning method is defined as cell-wise calibration, which calibrates the predictions by using a lesser number of field measurements in a cell-by-cell basis. In addition, a novel fingerprint filtering approach and a fingerprint matching technique (a cost function) are proposed. The trial results show that, the performance of DCM using the proposed database is inferior to that using a measured database. However, the application of calibration process for predictions improves the performance up to an acceptable level. The calibration method, designed for the bad urban scenario is based on curve fitting whereas that for urban, suburban and rural environments is based on neural networks. In addition, the novel fingerprint filtering approach is robust for the bad urban environment while the novel cost function shows higher performance with the proposed database. The best positioning accuracy for the. bad urban environment is 200m in 80% of the estimates and that for the urban environment is 125m (80%). Remarkable performance improvement can be observed in the rural environment giving a positioning error less than 385m in 80% of the estimates. The performance in suburban environment is inferior to that-in both urban and rural, with an error less than 550m in 80% of the time. The proposed solution for positioning is best suited for the deployment in large dynamic networks as a network-based method to provide basic information services, such as nearest ATM machine, petrol. station or hospital, traffic information and location based advertising.
  • Thumbnail Image
    item: Thesis-Abstract
    Database correlation for GSM location in outdoor and indoor environments
    Lakmali, BDS; Dias, D
    Accurately estimating the location of a Mobile Station is a key requirement to effectively provide a wide range of Location Based Services over mobile network. Since the mobile phone has become a common device in today's society, location based services are very popular among cellular subscribers. Hence developing cellular positioning techniques has been a key research problem and numerous localization solutions have been proposed. These include technologies such as Cell ID, angle and time of arrival methods and fingerprinting methods.// This thesis presents fingerprinting based positioning techniques suitable for different outdoor and indoor environments. Thus multiple positioning techniques are proposed, implemented and evaluated for different environments. Three outdoor trials in areas falls under urban, suburban and rural areas and two indoor trials in two nuiti storey buildings were used for evaluation. The ultimate solution proposed in this thesis is not a single positioning technique; rather it presents several positioning techniques that achieve optimum performance in each test environment.// This thesis proposes a novel fingerprint collection process for outdoor positioning and introduces a more accurate correlation algorithm. This thesis reports 67% positioning error as 12 n, 299 r and 221 l for urban, suburban and rural areas respectively. Experimental results show that the proposed positioning methods achieve accuracy far better than Cell-ID and trilateration approaches for the tested network environments especially for rural area. The 67 % positioning error for rural area is IC45 t and 1386 n with basic Cell-ID and trilateration techniques while proposed fingerprinting based technique reports 67% positioning error as 221 m.// With indoor positioning this thesis reports 50% positioning error as 20.5m and 8.7m for the selected two buildings. Also it was possible to accurately differentiate between floors in these multi storey buildings. Results achieved for Building 2 is reasonable when compared with the results reported in a similar study by Ve jc Otsason et ii (2005).