Return to Session C5 Next Abstract

Session C5: Navigation Using Environmental Features

A Deep Learning Approach to Mitigate LTE Multipath Localization Error
Kaleel Mahmood, Joe J. Khalife, and Zak (Zaher) M. Kassas, University of California, Riverside
Location: Windjammer

Signals of Opportunity (SOPs), such as cellular [1], Wi-Fi [2], and digital television [3], have been shown to be promising navigation sources in environments where global navigation satellite systems (GNSS) signals are intermittent or unavailable, e.g., deep urban canyons, indoors, and in the presence of unintentional interference and intentional jamming. In these environments, a specialized receiver could produce navigation observables by tracking the code and carrier phase of SOPs [4-7]. Cellular long-term evolution (LTE) signals have received increased attention recently, due to their inherent desirable attributes: geometric diversity and abundance of transmitting towers, high received power, and high transmission bandwidth. These attributes enable LTE signals to produce a robust and accurate standalone navigation solution [11].
While LTE systems are nominally designed to support time-difference-of-arrival (TDOA) positioning, most carriers disable this service in favor of transmitting communication data instead [8]. Recent research has shown that one may exploit other LTE synchronization signals for time-of-arrival (TOA) estimation, such as the primary synchronization signal (PSS) and the secondary synchronization signal (SSS) [9,10]. Experimental results demonstrated meter-level navigation accuracy with LTE signals for unmanned aerial vehicles (UAVs) and ground vehicles, with errors exceeding 10 meters in urban environments due to multipath [11].
In urban and indoor environments, the most significant source of error in the TOA estimate is multipath. In these environments, received non-line-of-sight (NLOS) signals due to nearby reflections distort the shape of the correlation function, which introduces biases in the correlators. The resulting errors are particularly significant when the TOA is estimated from the PSS or the SSS, since these signals have a relatively low bandwidth (930 KHz) [12]. Higher bandwidth signals are more robust to multipath-induced error. Therefore, one may use another LTE signal that possesses a much higher bandwidth: the cell-specific reference signal (CRS), which has a bandwidth up to 20 MHz [7]. The CRS is a pilot signal that is transmitted for channel estimation purposes. The TOA may subsequently be estimated from the channel impulse response (CIR) which is in turn estimated from the CRS. Although the effect of multipath on the CRS is less severe than its effect on the PSS and SSS, multipath will still introduce significant errors in the TOA estimate, especially in deep urban canyons and indoors.
Previous work to mitigate multipath include custom antenna design [13], receiver-based correlation techniques [7], and statistical [8] and machine learning based processing of the data [14]. In [13], a special antenna was designed to mitigate multipath that arises with aircraft landing. In [7], an orthogonal frequency division multiplexing (OFDM)–based delay-locked loop (DLL) that utilizes the estimation of signal parameters by rotational invariance techniques (ESPRIT) algorithm was proposed. In [8], a joint maximum likelihood (JML) time-delay estimator to improve the ranging performance in the presence of short-delay multipaths was proposed. In [14], a mitigation technique using support vector regression (SVR), trained on past repeated multipath measurements was proposed.
In this work, a novel machine learning approach that uses a special convolutional neural network is developed to mitigate the LTE multipath localization error. The proposed approach differs from previous machine learning approaches to mitigate multipath in two ways. First, previous approaches used training data based on navigation observable measurements. In contrast, the proposed machine learning approach is done directly at the correlation level. Here, instead of focusing on the navigation observable measurements, the network is directly trained with the SSS correlation function and the CIR produced by the receiver.
The second difference lies in the unique neural network structure that is developed to solve the multipath mitigation problem. Convolutional neural networks [15], [16] are traditionally composed of a series of convolution layers, followed by a feed forward neural network. The convolution layers act as feature extractors and the feed forward neural network acts as a classifier. The proposed architecture includes an autoencoder neural network [17] after the convolution layers. Autoencoders have been shown to be a powerful denoising tool [18]. This network architecture uses a stacked autoencoder to further denoise the features extracted from the convolution layers instead of using a traditional feed forward neural network for classification. This paper will present simulation and experimental results comparing (1) the performance and (2) the computational complexity of the proposed novel convolutional neural network architecture to existing approaches in the literature.
References
[1] R. Faragher, C. Sarno, and M. Newman, “Opportunistic radio SLAM for indoor navigation using smartphone sensors,” in Proceedings of IEEE/ION Position Location and Navigation Symposium, April 2012, pp. 120–128.
[2] C. Yang, T. Nguyen, and E. Blasch, “Mobile positioning via fusion of mixed signals of opportunity,” IEEE Aerospace and Electronic Systems Magazine, vol. 29, no. 4, pp. 34–46, April 2014.
[3] Z. Kassas, J. Khalife, K. Shamaei, and J. Morales, “I hear, therefore I know where I am: Compensating for GNSS deficiencies with cellular signals.” IEEE Signal Processing Magazine, pp. 111–124, September 2017.
[4] P. Thevenon, S. Damien, O. Julien, C. Macabiau, M. Bousquet, L. Ries, and S. Corazza, “Positioning using mobile TV based on the DVB-SH standard,” NAVIGATION, Journal of the Institute of Navigation, vol. 58, no. 2, pp. 71–90, 2011.
[5] 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.
[6] M. Driusso, C. Marshall, M. Sabathy, F. Knutti, H. Mathis, and F. Babich, “Vehicular position tracking using LTE signals,” IEEE Transactions on Vehicular Technology, vol. 66, no. 4, pp. 3376–3391, April 2017.
[7] K. Shamaei, J. Khalife, S. Bhattacharya, and Z. Kassas, “Computationally efficient receiver design for mitigating multipath for positioning with LTE signals,” in Proceedings of ION GNSS Conference, September 2017, accepted.
[8] J. Peral-Rosado, “Evaluation of the LTE positioning capabilities in realistic navigation channels”, Ph.D. Dissertation, Universitat Autonoma de Barcelona, 2014.
[9] J. del Peral-Rosado, J. Parro-Jimenez, J. Lopez-Salcedo, G. Seco-Granados, P. Crosta, F. Zanier, and M. Crisci, “Comparative results analysis on positioning with real LTE signals and low-cost hardware platforms,” in Proceedings of Satellite Navigation Technologies and European Workshop on GNSS Signals and Signal Processing, December 2014, pp. 1–8.
[10] K. Shamaei, J. Khalife and Z. Kassas, “Performance characterization of positioning in LTE systems,” in Proceedings of ION GNSS Conference, September 2016, pp. 2262-2270.
[11] Z. Kassas, J. Morales, K. Shamaei, and J. Khalife,“LTE steers UAV,” GPS World Magazine, vol. 28, no. 4, pp. 18–25, April 2017.
[12] K. Shamaei, J. Khalife, and Z. Kassas, “Comparative results for positioning with secondary synchronization signal versus cell specific reference signal in LTE systems,” in Proceedings of ION International Technical Meeting Conference, January 2017, pp. 1256–1268.
[13] D. Aloi and M. Sharawi,”High fidelity antenna model validation results of a GNSS multipath limiting antenna,” in IEEE Transactions on Aerospace and Electronic Systems, vol. 47, no. 1, pp. 3-14, January 2011.
[14] Q. Phan and S. Tan,“Mitigation of GPS periodic multipath using nonlinear regression,” in Proceedings of the European Signal Processing Conference, August 2011, pp. 1795-1799.
[15] Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner,”Gradient-based learning applied to document recognition,” in Proceedings of the IEEE, vol. 86, no. 11, November 1998, pp. 2278-2324.
[16] I. Goodfellow, Y. Bengio, and A. Courville,”Deep learning,” MIT press, 2016.
[17] G. Hinton and R. Salakhutdinov,”Reducing the dimensionality of data with neural networks,” Science, vol. 313, no. 5786, July 2006, pp. 504-507.
[18] P. Vincent, H. Larochelle, Y. Bengio, and P. Manzagol,”Extracting and composing robust features with denoising autoencoders," In Proceedings of the 25th international conference on Machine learning, pp. 1096-1103, ACM, 2008.



Return to Session C5 Next Abstract