Position, Navigation, and Timing Technologies in the 21st Century

GNSS SVs transmit all necessary states and parameters to the receiver in the navigation message (e.g. SV position, clock bias, ionospheric model parameters, etc.). In contrast, cellular BTSs do not transmit such information. Therefore, navigation frameworks must be developed to estimate the states and parameters of cellular BTSs (position, clock bias, clock drift, frequency stability, etc.), which are not necessarily known a priori. Several navigation frameworks have been proposed. One such framework is to have a dedicated station that acts as a mapper, which knows its states (from GNSS signals, for instance), is estimating the unknown states of cellular BTSs, and is sharing such estimates with navigating receivers. Another framework is to simultaneously estimate the states of the receiver and cellular BTSs in a radio simultaneous localization and mapping (radio SLAM) manner [26–29].

This chapter discusses how cellular signals could be used for PNT by presenting relevant signal models, receiver architectures, PNT sources of error and corresponding models, navigation frameworks, and experimental results. The remainder of this chapter is organized as follows. Section 38.2gives a brief overview of the evolution of cellular systems. Section 38.3discusses modeling the clock error dynamics to facilitate estimating the unknown BTSs’ clock error states. Section 38.4describes two frameworks for navigation in cellular environments. Sections 38.5and 38.6discuss how to navigate with cellular code‐division multiple access (CDMA) and LTE signals, respectively. Section 38.7discusses a timing error that arises in cellular networks: clock bias discrepancy between different sectors of a BTS cell. Section 38.8highlights the achieved navigation solution improvement upon fusing cellular signals with GNSS signals. Section 38.9describes how cellular signals could be used to aid an INS.

Throughout this chapter, italic small bold letters (e.g. ) represent vectors in the time domain, italic capital bold letters (e.g. ) represent vectors in the frequency domain, and capital bold letters represent matrices (e.g. X).

Cellular systems have evolved significantly since the first handheld cell phone was demonstrated by John F. Mitchell and Martin Cooper of Motorola in 1973. The first commercially automated cellular network was launched in Japan by Nippon Telegraph and Telephone (NTT) in 1979. This first generation (1G) was analog and used frequency division multiple access (FDMA). The second generation (2G) transitioned to digital and mostly used time‐division multiple access (TDMA), which later evolved into 2.5G: General Packet Radio Service (GPRS) and 2.75G: Enhanced Data Rates for GSM Evolution (EDGE). The third generation (3G) upgraded 2G networks for faster Internet speed and used CDMA. The fourth generation (4G), commonly referred to as LTE, was introduced to allow for even faster data rates. LTE used orthogonal frequency division multiple access (OFDMA) and featured multiple‐input multiple‐output (MIMO), that is, antenna arrays. Figure 38.1summarizes the existing cellular generations and their corresponding predominant modulation schemes.

This chapter focuses on using cellular CDMA and LTE signals for PNT. Table 38.1compares the main characteristics of (i) GPS coarse/acquisition (C/A) code, (ii) CDMA pilot signal, and (ii) three LTE reference signals: primary synchronization signal (PSS), secondary synchronization signal (SSS), and cell‐specific reference signal (CRS).

Figure 38.1 Cellular systems generations.

Source: Adapted from A. Elnashar, “Wireless Broadband Evolution,” http://www.slideshare.net/aelnashar/ayman‐el‐nashar, June 2011, accessed on: June 2019.

Table 38.1 GPS versus cellular CDMA and LTE comparison

* 1% of chip width

In 2012, the International Telecommunication Union Radiocommunication (ITU‐R) sector started a program to develop an international mobile telecommunication (IMT) system for 2020 and beyond. This program set the stage for 5G research activities. The main goals of 5G compared to 4G include (i) higher density of mobile users; (ii) supporting device‐to‐device, ultra‐reliable, and massive machine communications; (iii) lower latency; and (iv) lower battery consumption. To achieve these goals, millimeter wave bands were added to the current frequency bands for data transmission. Other salient features of 5G include millimeter waves, small cells, massive MIMO, beamforming, and full duplex [30, 31].

GNSS SVs are equipped with atomic clocks, are synchronized, and their clock errors are transmitted in the navigation message along with the SVs’ orbital elements. In contrast, cellular BTSs are equipped with less stable oscillators (typically OCXOs), are roughly synchronized to GNSS, and their clock error states (bias and drift) and positions are typically unknown. As such, the cellular BTSs’ clock errors and positions must be estimated. Therefore, it is important to model the clock error state dynamics. To this end, a typical model for the dynamics of the clock error states is the so‐called two‐state model, composed of the clock bias δt and clock drift , as depicted in Figure 38.2.

The clock error states evolve according to

(38.1)

Figure 38.2 Clock error states dynamics model.

Source: Z. Kassas, Analysis and synthesis of collaborative opportunistic navigation systems, Ph.D. Dissertation, The University of Texas at Austin, USA, May 2014. Reproduced with permission of Z. Kassas (University of Texas at Austin).

where the elements of are modeled as zero‐mean, mutually independent white noise processes, and the power spectral density of is given by . The power spectra and can be related to the power‐law coefficients , which have been shown through laboratory experiments to be adequate to characterize the power spectral density of the fractional frequency deviation y ( t ) of an oscillator from the nominal frequency, which takes the form [32, 33]. It is common to approximate the clock error dynamics by considering only the frequency random walk coefficient h −2and the white frequency coefficient h 0, which lead to and [34, 35]. Typical OCXO values for h 0and h −2are given in Table 38.2.