Previous Abstract Return to Session C5 Next Abstract

Session C5: Navigation Using Environmental Features

RSS-Based Localization Neural Network Algorithm in Cellular Networks
Ali A. Abdallah and Samer S. Saab, Lebanese American University; Zak (Zaher) M. Kassas, University of California, Riverside
Location: Windjammer

Due to coverage limitations of the global navigation satellite system (GNSS), substantial work has been oriented towards finding efficient alternative solutions. Signals of opportunities (SOPs) has been proposed to be one of the best alternatives in the absence of GNSS [2] or as a complementary system [1]. Cellular signals has been proven to be the most effective SOPs due to their abundance, geometric diversity, high carrier frequency, large bandwidth, and high received power [3]. Cellular code division multiple access (CDMA) and long-term evolution (LTE) signals are exploited to create navigational frameworks and receivers that are capable of navigating with an acceptable accuracy of meters using the forward link structure and frame structure for CDMA and LTE, respectively [4], [5].
Over the past two decades, cellular signals have had a significant presence in the positioning and navigation community. The employment of received signal’s strength (RSS) is one technique that has been considered for that purpose. However, due to various effects, the contents of fingerprints may vary over time even when measured at the same location. The RSS fluctuations are due to typical multipath, and variation of the set of cell towers [8]. Methods that are based on statistical techniques deal with multipath variations (see, e.g., [6], [7]). Other methods take into consideration the two sources of RSS fluctuations leading to 40% performance improvement in comparison to traditional methods [8].
Moreover, several algorithms have been proposed for the sake of improving RSS-based navigational systems in cellular networks. One of these algorithms is a data fusion that employs RSS and time advance (TA). The integration of the latter measurements of cellular signals leads to a better performance in comparison to using each measurement technique individually [9]. However, most of the algorithms imply non-stochastic techniques without directly accounting for neither noisy signals nor path loss variations (specifically indoors). The optimal stochastic Newton-Raphson algorithm has been proposed in [11] to account for the previously mentioned challenge concerning noisy signals. Also, the study to estimate the optimal path loss in RSS navigation systems reflected a significant improvement, despite the costly stages that needs to be accounted for during measurements [12]. In addition, applications that employ k-nearest neighbors (KNN) estimation method or other dependent distance algorithms show that implementing a mathematical position error model provides improvement in positioning estimation [10].
This paper proposes a localization algorithm that maps the RSS-based fingerprints to the location of the receiver without using the knowledge of BTS locations and without using any RSS mathematical model. This paper focuses on providing a robust standalone RSS-based solution to the localization problem. The proposed localization method is based on a machine learning approach, where the estimation accuracy depends on the involvement of the learning process. In what follows, we elaborate on the robustness issue. From a theoretical point of view, whenever we deploy sufficiently large number of BTSs, N, spread around an urban area, then each receiver would possess a unique fingerprint composed of N RSS values. Consequently, based on the values of RSS, one can identify the receiver position. However, due to multipath phenomena the large variations of the RSS values could result in many singularities. This singularity problem can be partially resolved by taking the time average of RSS values over some time interval while considering selected locations with diverse fingerprints.
The algorithm employs an integrated Weighted K-Nearest Neighbor algorithm (WKNN) [13] and a multi-layer Neural Network (NN) in order to estimate the location of the receiver. The proposed scheme is based on learning the environment in an offline fashion before implementation. We divide the localization domains into several clusters and strategically select few reference points (RP) within each cluster where each RP is expected to have a unique fingerprint. We record the position of the RP and the average of the RSS values taken over several time samples. The weights and biases of the NN in each cluster are trained using the backpropagation approach by going over several locations and at different times within each cluster while feeding the NN with the locations and their associated RSS values. In the implementation phase, we first use WKNN to identify the specific cluster and to also estimate the location of the receiver. Based on the WKNN identified cluster, the NN is loaded with the corresponding training weights that correspond to the identified cluster. Subsequently, the NN estimates the receiver position based on the corresponding RSS values.
The main intension of this work is intended to present a proof of concept of the proposed approach. In particular, we study the robustness and performance of the proposed localization scheme versus the offline training involvedness based on different experimental demonstrations.
References:
[1] M. Enright., and N. Christopher, “A signals of opportunity based cooperative navigation network,” in Proceedings of the IEEE National Aerospace & Electronics Conference, April 2009, pp. 213-218.
[2] C. Yan, and H. Fan, “Asynchronous differential TDOA for non-GPS navigation using signals of opportunity,” Acoustics, Speech and Signal Processing, March, 2008.
[3] Z. Kassas, J. Khalife, K. Shamaei, and J. Morales, “I Hear, Therefore I Know Where I Am: Compensating for GNSS Limitations with Cellular Signals,” IEEE Signal Processing Magazine 34, no. 5, 2017, pp. 111-124.
[4] J. Khalife, K. Shamaei, and Z. Kassas, “A software-defined receiver architecture for cellular CDMA based navigation,” in Proceedings of IEEE/ION Position, Location, and Navigation Symposium, April 2016, pp. 816–826.
[5] F. Knutti, M. Sabathy, M. Driusso, H. Mathis, and C. Marshall, “Positioning using LTE signals,” in Proceedings of Navigation Conference in Europe, April 2015, pp. 1–8.
[6] S. C. Ergen, H. S. Tetikol, M. Kontik, R. Sevlian, R. Rajagopal, and P. Varaiya, “Rssi-fingerprinting-based mobile phone localization with route constraints,” IEEE Transactions on Vehicular Technology, 2014, pp. 423–428.
[7] K. K. Mohamed Khalaf-Allah, “Mobile location in gsm networks using database correlation with bayesian estimation,” in 11th IEEE Symposium on Computers and Communications, Los Alamitos, CA, USA, 2006, pp. 289–293.
[8] Chitraranjan, D. Charith, A. Denton, and A. Perera, “A Complete Observation Model for Tracking Vehicles from Mobile Phone Signal Strengths and Its Potential in Travel-Time Estimation,” In Proceedings of 84th IEEE Vehicular Technology Conference, 2016, pp. 1-7.
[9] Zhang, Miao, S. Knedlik, P. Ubolkosold, and O. Loffeld, “A data fusion approach for improved positioning in GSM networks,” in Position, Location, And Navigation Symposium, 2006 IEEE/ION, pp. 218-222.
[10] Moghtadaiee, Vahideh, A. Dempster, and B. Li. “Accuracy indicator for fingerprinting localization systems,” in Proceedings of IEEE/ION Position, Location, and Navigation Symposium, 2012, pp. 1204-1208.
[11] K. Saab, and S. Saab, “Application of an optimal stochastic Newton-Raphson technique to triangulation-based localization systems,” in Proceedings of IEEE/ION Position, Location, and Navigation Symposium, 2016, pp. 981-986.
[12] Mazuelas, Santiago, F. Lago, D. González, A. Bahillo, J. Blas, P.Fernandez, R. Lorenzo, and E. Abril. “Dynamic estimation of optimum path loss model in a RSS positioning system,” in Proceedings of IEEE/ION Position, Location, and Navigation Symposium, 2008, pp. 679-684.
[13] Liu, Zihan, X. Luo, and T. He, “Indoor Positioning System Based on the Improved W-KNN Algorithm,” in 2nd Advanced Information Technology, Electronic and Automation Control Conference, March, 2017.



Previous Abstract Return to Session C5 Next Abstract