The Research Of Mobiole Terminal Location Algorithms Based On Measurement Report

Measurement report (MR) is a new kind of data source for the field of cellular network optimization. It’s a data segment combined by the user equipment (UE) report and Node B report. It includes rich information of the wireless signals, such as the received signal level (RXLEV) from the neighbor cells, signal quality from the serving cell, and the uplink chan-nel characteristics. MR can be intercepted from the ABIS interface for GSM or Iub interface for UMTS, and with the practical signal information, coverage and interferences of the cells can be gotten and optimized furthermore, and the precondition of implementing the func-tions is to locate where the MR generated. There have been many cellular network based lo-cation methods, such as least square (LS) algorithms, data fusion, fingerprints, and proximity algorithms. The LS algorithms, including Taylor series method, extended Kalman filter, Chan, etc., smooth the measurements noise and utilize utmost available information by transforming nonlinear measurement equations to linear ones. However, there are many limitations of utilize the measurements. The parameters, such as AOA, TOA/TDOA, are al-ways unachieved in the network. NLOS effects also dominate the estimation error, and though some researchers suggested alleviation methods, there isn’t an available method adapted for complex scenes. Fingerprints algorithms which utilizes transcendental database are considered as available positioning in NLOS environments. But collecting dense prior data in a big city which owns millions of inhabitants and its maintenance is so big a load. Proximity algorithms, such as CI-AVG, CI-RXLEV, and CI-ANGLE, etc., which are always low performance in traditional views of researchers haven’t been concerned before. But it’s a usual way of operation and application. Most of the parameters could be extracted from the MR, and the simplest algorithm cell identifier (CI) has been utilized by many operators. It satisfied the requirement of E-911 phase I, and the Chinese operators still implement it for the basic location.To get practical MRs, we carried out a field measurement in city Huangshi with the help of Huangshi Branch of China Mobile Communications Corporation (CMCC) on a sunny day. Huangshi is the second biggest city of Hubei and on the northeast of the province. It has over 2.5 millions inhabitants and the area is about 180 square kilometers. Most of the landform is hills. It’s a sunny day. We drove the car along the road and recorded data at the carriageway, footway, square, lawn, grove or lakefront. The wireless network was GSM, and power con- trols of the cells were opened.Based on the filed measurements, this paper compares the performances of proximity algorithms CI, CI-AVG, CI-RXLEV and CI-ANGLE firstly. The experiment shows that the simplest algorithm CI achieves the highest accuracy, and the position error of CI-RXLEV which utilizes the most parameters is inaccurate due to the neighbor record matching to cell parameters database and complex terrains. So we advance two suggestions and the position-ing accuracy is enhanced at 47%.Based on the results of CI-RXEV enhancements (CRE), extended Kalman filter (EKF) is also analysized. The measurement equation sets is contructed by the linear operations of Okumura-Hata, and the 0-order minor is approximated. The experiment shows that EKF could achieve more accurate positions due to the NLOS environments and slow convergence of Taylor series. It ask for more available information to reduce error. Then we compare the time series operation methods moving average and vector Kalman filter. Due to the big noise power of measurements, Kalman filter also doesn’t take advantages in this comparison. Based on the results of time series operation, this paper considers to utilize the parameters of CRE algorithm for estimating its positioning error. The statistics of the parameters are cal-culated and fused as a estimator. The experiment results show that the estimator could fit the big-error ones, but couldn’t distinguish the small-errors kindly.This paper also suggests a notion of virtual antenna. The virtual antenna is defined for combing parameters of the cells which belong to the same Node B for accuracy enhancement. It’s based on not only the objective MR but also the MRs generated nearby the object. The received signal level of virtual antenna is defined for more available weight used in CI-RXLEV-E estimation, and the azimuth of virtual antenna is defined for indicating UE direction more accurately. Then a step-by-step rotation algorithm is proposed for location with the azimuth. Experiment results based on the field measurement show that the utiliza-tion of virtual antenna is capable of achieving higher performance than the fingerprints method with greatly reduced manpower and material cost.