Particle Filter Based Positioning with 3GPP-LTE in Indoor Environments

Services and applications based on accurate knowledge of the mobile terminal (MT) location play a fundamental role in current and future wireless communications systems. In addition, the United States Federal Communications Commission (FCC) has stated accuracy requirements on the location determination process of enhanced 911 (E-911) emergency callers. Global navigation satellites systems (GNSSs) based positioning provides a sufficient accuracy in rural and suburban environments, where a sufficient number of satellites are visible in line-of-sight conditions. However, in GNSS critical environments, such as dense urban, urban canyon or even indoors, view to sky is limited. In these environments, low signal power, bad satellite constellations, severe multipath and non-line-of-sight propagation causes erroneous and biased position estimates. Especially in these environments, cellular wireless communication systems provide good coverage and can be used for position determination of the MT. Mobile radio communications systems like GSM, UMTS or the currently deployed 3GPP-LTE primarily target on optimizing communication performance figures such as bandwidth efficiency or data throughput. Availability, signal strength or even signal bandwidths, however, make them interesting for positioning. However, the correlation properties of the synchronization signals of the 3GPP LTE system limit the positioning performances. Moreover, while non-line-of-sight propagation improves the communication performance, it degrades the navigation performance due to the additional distance the signal might travel. Hence, this paper shows an indoor positioning approach with the 3GPP-LTE mobile communication standard, which is currently deployed in many countries. Moreover, it shows the benefit of using the 3GPP-LTE mobile communication system for indoor positioning. Therefore, this paper describes a novel real-time mobile radio based positioning system using time-difference-of-arrival (TDOA) measurements. The paper considers an indoor scenario, where the transmitters are located outdoors and the MT is moving in an office building. The position estimation is done by a particle filter. Furthermore, to improve the positioning accuracy, this paper derives a time-variant error model for indoor positioning. Using this time-variant error model, the positioning error of the MT can be decreased significantly. The downlink of 3GPP-LTE is based on Orthogonal Frequency Division Multiplexing (OFDM), which allows a spectral efficient and flexible usage of the available frequency spectrum. The 3GPP-LTE standard specifies signal parts, dedicated to time and frequency synchronization. For our investigations we use these synchronization signals for TDOA based position estimation. The 3GPP-LTE signal structure, in particular the properties of the synchronization properties, will be discussed in detail in the final paper. For the 3GPP-LTE positioning we apply a particle filter in order to process TDOA measurements of the 3GPP-LTE base stations. The TDOA measurements are taken from DLR´s 3GPP-LTE positioning test-bed. The test-bed consists of up to four synchronous transmitters. From each transmitter site a predefined 3GPP-LTE OFDM frame can be transmitted periodically. The test-bed operates at 2.4 - 2.5 GHz, providing signal bandwidths of up to 20 MHz. At the receiver we convert the received signal down to base band and sample both the inphase and quadrature component. The sampled signal is stored on a hard disk, which allows both offline and real-time processing. These sampled data are processed by the TDOA estimation algorithm and consists of two steps. The first step estimates the TDOA roughly by correlating the narrow band synchronization signals with the received sequence and the second step performs subsample based estimation with the wideband pilot symbols. These algorithms use a first peak and maximum peak detection algorithms, for detecting the first and maximum arriving signal. However, the 20 MHz bandwidth limits the rough synchronization and allows only a sample based estimation within 15 meters, which hinders accurate positioning in indoor environments. Thus, an oversampling approach is used, to tackle this issue which results in a significant error reduction. However, due to hardware imperfections of the test-bed and channel errors, such as non-line-of-sight and multipath propagation, the TDOA measurements are noisy and biased. Thus, to obtain better positioning results, this bias has to be predicted and mitigated. Several approaches exist to model time of arrival based ranging in indoor environments. All of these statistical models depend on bandwidth, carrier frequency and are time-invariant. However, for navigation applications an evaluation of the multipath and non-line-of-sight error for a moving receiver is essential. Hence, to improve the positioning accuracy, we derive in this paper a time-variant TDOA error model based on a measurement campaign of an outdoor-to-indoor channel. The evaluation of this measurement campaign yields an autoregressive error model which allows us to predict the TDOA error for a moving MT. By using this model within the particle filter yields promising results in terms of error mitigation. Each particle itself models the multipath and non-line-of-sight error according to the obtained time-variant error model. Additionally, we compare these results to the more general approach by assuming an uncorrelated error. In the final paper, we will provide a detailed description of the 3GPP-LTE downlink signal structure, which we use for TDOA based positioning. We will discuss and describe the applied algorithms for timing (pseudo-range) estimation. Furthermore, we will describe the 3GPP-LTE test-bed and the scenario in detail. Additionally, the particle filter used for the positioning estimation will be described in detail. Especially, we will analyze the measurement results and the derived time-variant model to predict and mitigate the multipath and non-line-of-sight error in the particle filter