Interference and multipath suppression with spacetime adaptive beamforming for safetyoflife maritime applications
 432 Downloads
Abstract
In this work, we present mitigation algorithms to protect GNSS receivers against malicious interference. Maritime applications with an antenna arraybased receiver are considered as a use case. A twostage mitigation algorithm, that tackles multipath and radio frequency interference (RFI), caused by personal privacy devices (PPD) or additive white Gaussian noise (AWGN) interferers is presented. Our approach consists of a prewhitening step, followed by a spacetime adaptive principle component analysis (PCA) beamformer that uses a dimensionality reduction (i.e. compression) method based on Canonical Components (CC) with a bank of signalmatched correlators. The algorithms are capable of suppressing strong RFI and separating highly correlated and even coherent multipath signals, thus achieving a reliable time delay and, therefore, pseudorange estimation performance. Finally, we evaluate and compare the proposed algorithms not only via numerical simulations but also with real data collected from a measurement campaign performed at DLR’s maritime jamming testbed in the Baltic sea. A complete description of the test platform and the scenarios is provided.
Keywords
Global navigation satellite systems (GNSS) Maritime navigation Interference mitigation Multipath mitigation Antenna array signal processing Beamforming1 Introduction
Global navigation satellite systems (GNSS) are essential for maritime applications. For these applications, GNSS receivers are coupled with inertial measurement units (IMUs), which fuse their respective measurements with the GNSS signal processing algorithms to produce more accurate and reliable positioning results. Information from GNSS units is subsequently integrated with electronic navigation charts (ENCs) and the automatic identification system (AIS), that larger ships are required to carry. Together they form the main positioning, navigation, and timing (PNT) unit necessary for ships of a certain class.
During the last years, studies indicated that the majority of maritime accidents are caused by human error [1]. A case that gathered widespread publicity was the sinking of Costa Concordia [2], where the captain misused the ship’s navigation system. A sufficient degree of automation can most likely avoid these kind of accidents. This can be achieved using trustworthy semiautonomous navigation systems to reduce the effect of human errors. In the light of those studies, the International Maritime Organization (IMO) developed and actively promotes the “eNavigation” concept [3], which aims for a wider and holistic integration of already existing and new electronic navigational tools to reduce navigation errors and hereby accidents.
One of the key elements of the “eNavigation” concept is the standardization of GNSS performance requirements with respect to positioning accuracy, integrity and signal availability. Current planning supports the introduction of terrestrial (differential GNSS) or satellitebased augmentation systems (SBAS), to satisfy the positioning requirements.
Prominent studies [4] indicate that GNSS signals in the maritime environment can be susceptible to severe degradation. The General Lighthouse Authorities of the United Kingdom and Ireland (GLA) in collaboration with the UK Ministry of Defense (MOD) have performed seatrials with the aim of identifying the effects of GPS jamming on safety and security with respect to navigation at sea [5]. One of the important results of this study was that interference in the GNSS bands not only affects the position outcome of the onboard GNSS receivers. The ship’s digital situation awareness, chart stabilization, digital selective calling and emergency communications could be affected, since they are all coupled and rely on GNSS.
The main and most prevalent reason is degradation due to radio frequency interference (RFI). First, due to the low signal power of GNSS satellite signals at the earth’s surface, the nominal operation of GNSS equipment is easily susceptible to unintentional interference [6]. Emissions of radio systems that either share the same frequency band, e.g. aviation Distance Measurement Equipment (DME) in Galileo E5 band, or operating in the neighboring frequencies and generate harmonics in the GNSS bands, e.g. terminals of Mobile Satellite Systems (MSS) might cause a significant performance degradation or even disable the usage of the GNSS bands.
Additionally, the availability of illegal socalled Personal Privacy Devices (PPDs) has risen over the years. Those type of devices intentionally transmit interfering signals in the GNSS bands. As a result, due to their relative high power in comparison with the GNSS signals, commercial GNSS receivers in a wide radius are jammed and cannot provide navigation information, thus threatening even critical infrastructure [7].
Finally, multipath signal components caused by reflections of the GNSS signals can cause a significant degradation of the positioning accuracy [8]. While that might not be the case in the open sea, vessels preparing for docking in the harbor and traversing in inland waters might experience a significant positioning error, since they have to navigate among tall and metallic structures, that can reflect GNSS signals.
Furthermore, large maritime organizations, such as the Lloyd’s [9] register but also the European Global Navigation Satellite Systems Agency (GSA) [10] state that navigation robustness and cybersecurity will be a major challenge and trend for the upcoming years.
Several singleantennabased RFI and multipath mitigation techniques for GNSS have been proposed in the literature. The simplest way is to design the antenna, such that it highly attenuates reception from low elevations, since these kinds of interference are expected to impinge from below [11, Chapter 17]. To resolve highly correlated multipath or RFI, rather long observation intervals are necessary if only the time domain is used [12]. Singleantenna approaches show good mitigation capabilities for stationary and nonestationary narrowband interferences in the frequency and time domain [13, 14, 15]. However, in case of broadband interference, their performance is limited [16].
Antenna arrays allow for mitigation techniques in the spatial domain [17, 18, 19]. By combining the spatial and time (spacetime) domain, the suppression capabilities can be further increased, especially for wideband and high dynamic interferences [17]. Furthermore, highresolution parameter estimation algorithms can jointly mitigate multipath and RFI, and provide highly accurate results [20, 21]. This is achieved by separating the LOS component from reflections [20]. However, these methods entail rather high complexity in the parameter estimation as multidimensional nonlinear problems have to be solved. Accurate modeling is mandatory.
In radar signal processing, spacetime adaptive processing (STAP) techniques are widely known for their mitigation capabilities of clutter [22, 23]. Our approach follows the principles of spacetime eigenrake receivers [24], but is tailored to timedelay estimation for GNSS. To mitigate RFI, we combine the blind RFI mitigation technique described in [25] with the adaptive spacetime method presented in [26]. Therein, a spacetime adaptive principle component analysis (STAPCA) using a compression method based on Canonical Components (CC) with a bank of signalmatched correlators [27] is described. Since the problem is linear and no modelorder estimation is required, the computational complexity is significantly reduced. Furthermore, adapted blockwise preprocessing algorithms—Forward–Backward Averaging (FBA) and/or Spatial Smoothing (SPS) [28]—are employed to even increase the decorrelation capabilities.
The remainder of the paper is organized as follows: first, a generic data model including multipath and radio frequency interference is introduced. Based on that, a precorrelation mitigation technique is described. As a next step, postcorrelation mitigation algorithms are motivated and derived. These lead to a description of the timedelay estimation in Sect. 4, followed by a softwarebased evaluation of the algorithms. Subsequently, the hardware platform used for the experimental proof of concept is described. A summary and conclusion complement the paper.
2 Notation

\(\mathbf {x}\): bold face lower case letters denote column vectors.

\(\mathbf {X}\): bold face capital letters denote matrices.

\(\mathrm {Re}\{\cdot \}\text { }(\mathrm {Im}\{\cdot \})\): real (imaginary) part of a complex scalar, vector or matrix.

\((\cdot )^T\text { }((\cdot )^H)\): the transpose (hermetian) of a vector or matrix.

\(\otimes\): the Kronecker product.

\(\Box\): the Khatri–Rao product.

\(\mathbf {1}_M\): a column vector of length M consisting of ones.

\(\mathbf {I}_Q\): a \(Q\times Q\) identity matrix.

\(\text {vec}(\mathbf {X})\): vectorized version of a matrix \(\mathbf {X}\), i.e. all columns of \(\mathbf {X}\) are stacked up.

\(\mathrm {E}\left[ \cdot \right]\): the expected value of a random variable, for which ergodicity is always assumed.

\(x{[}k{]}=x(kT)\): a sample at index k of a time continuous signal x(t) with sampling period T.
3 Data model
4 Precorrelation interference mitigation
5 Postcorrelation interference and multipath mitigation
The key ingredient for the postcorrelation mitigation technique described in the following section is the postcorrelation covariance matrix. First, the data model and basic definitions are introduced. This leads to a formal derivation of the covariance matrix. Given certain geometrical properties of the antenna array, the covariance matrix has a special structure, which is described in the next subsection. This enables spatial decorrelation techniques (FBA and SPS). These are presented in the last two subsections.
The following remark about notation and calculus methods used seems necessary: in contrast to the precorrelation domain, the postcorrelation domain has one additional dimension (due to the correlator bank consisting of Q correlators for each satellite channel, which will be described in the following). Naturally, the data could be described using tensors or manifolds. However, this would lay beyond the scope of this work. Therefore, whenever more than two dimensions occur, an “unfolding mechanism” (mainly based on Kronecker products, vectorization and selection matrices) is applied to enable the use of ordinary matrix calculus.
5.1 Postcorrelation data model
5.2 Postcorrelation covariance matrix
5.3 Special decorrelation properties
5.4 Blockwise forward–backward averaging (FBA)
5.5 Blockwise spatial smoothing (SPS)
In the following subsection, we derive a blockwise 2D SPS scheme for the spacetime covariance matrix considering Uniform Rectangular Arrays (URAs) with \(M_x \times M_y\) elements. We define linear subarrays in x and ydirections with the same number of sensors. Therefore, we get the number of sensors for one subarray in xdirection \(M_{sub_{x}} = M_x  L_x + 1\), where \(L_x\) defines the number of linear subarrays in xdirection. In our case, \(M_{sub_{x}}\) is equal to the number of sensors in ydirection \(M_{sub_{y}}\) and also \(L_x\) is equal to \(L_y\). Then the number of rectangular subarrays is \(L_s = L_x L_y\) and each subarray contains \(M_\mathrm{sub} = M_{sub_{x}}M_{sub_{y}}\) sensor elements.
6 Timedelay estimation
7 Simulation results
Parameter setup for the simulations
Delay  Azimuth \((^{\circ })\)  Elevation \((^{\circ })\)  

Direct signal  N/A  8  80 
Distant MP  0.3 chips  133  20 
Close MP  0.3 chips  30  60 
Interference  N/A  120  10 
Before describing the simulation results, the following remark is provided: all examples are illustrative and focus only on the parameter variations mentioned before. Signaltomultipath, signaltointerference as well as spatial separation would have an effect on pseudorange measurements as well, but are not considered. However, an exhaustive parameter space exploration would lay far beyond the scope of this work.
Summary of the evaluated algorithms
Algorithm  Legend 

Regular singleantenna approach  Single 
Multicorrelator singleantenna approach  MULTI 
Precorrelation beamforming  PCA 
Spacetime adaptive beamforming  STAPCA 
The multicorrelator algorithm using one antenna suppresses the multipath in the temporal domain and has a lower pseudorange error than the traditional approach using only three correlators. The traditional approach, which reflects the stateoftheart implementation of current commercial maritime GNSS receivers, suffers the most and its pseudorange error grows significantly larger than the rest. After 25 s one RFI signal is also turned on. The singleantenna algorithms loose lock and do not yield a pseudorange estimate. Due to prewhitening, the multiantenna algorithms keep their performance after convergence of the tracking loops even when the RFI is turned on.
Surprisingly—in the presence of RFI—the pseudorange error observed for PCA and STAPCA in the analyzed scenario of Fig. 2 is lower than without RFI, i.e. in the time interval from 10 to 25 s. Both subspace projection methods form spatial zeros in the direction of the RFI source. Sidelobes are created as well. This has an effect on the distant multipath component: it is filtered out when the RFI is turned on. However, one cannot generalize based on that specific simulation setup and outcome.
8 Multiantenna GNSS receiver test platform

The power consumption has to be low enough to record long enough.

To avoid fire, the heat produced by the devices has to be acceptable.

The downmixing has to be done for all channels synchronously in parallel.

A calibration signal is necessary to measure and, therefore, compensate different latencies of the cables connecting the antennas, that are in general not equal.

All devices (i.e. mixers and ADsamplers) have to be synchronized using a common local oscillator frequency.

The sampling rate needs to be high enough to capture the whole GPS L1 band.

The storage (i.e. RAIDdevice) has to be fast and big enough to allow for high bit resolutions of the sampler and several minutes of recording time.

The number of output channels used for replay has to correspond to the number of inputs the (arraybased) realtime receiver uses.

For postprocessing, the IF and the sampling rate of the D/A converters have to be compatible with the receivers that are to be stimulated.
The signal processing chains in the digital domain—for postprocessing in software or realtime—can be designed arbitrary in principle. In the context of this work, the processing flow described in the first sections is implemented. The different parts of the platform are described in the following subsections.
8.1 Antenna array
8.2 Frontend
8.3 Sampler
8.4 Replay
9 Experimental evaluation of the test platform
The platform was tested in a maritime measurement campaign in the week from June, 13th to June, 17th 2016, in DLR’s maritime jamming test bed in the Baltic sea, close to Hiddensee, Germany. The location of the test environment is illustrated in Fig. 7 (indicated by the red star). The general setup of all scenarios consists of two vessels. One vessel is equipped with the jammer setup and is anchored at a fixed point. The other vessel is equipped with the first part of the receiver test platform. Different experiments have been performed to demonstrate the mitigation capability for two interference types. Additive white Gaussian noise (AWGN) interference as well as PPDs have been used. To evaluate the algorithms and setup until their limit is reached, an amplifier with a maximum boost of 4 W was necessary.
9.1 AWGN interference mitigation
The first experimental setup is an AWGN broadband (15 MHz) interference scenario. The vessel, which is equipped with the receiver test platform holds an approximately static position in the main lobe of the interference transmit antenna of the jammer vessel. For the first 30 s of the experiment the interference is turned off. After 30 s the AWGN interference with approximately 20 dB jammertosignal ratio (JSR) is turned on. Over 50 s the interference power is continuously increased such that a final JSR of approximately 60 dB is achieved after 80 s.
The singleantenna algorithm looses lock to all satellites, right after the interference is turned on. Compared to STAPCA, PCA looses more satellites at a lower JSR and looses lock at approximately 65 s.
9.2 PPD interference mitigation
10 Summary and conclusion
In this work, we presented a set of mitigation algorithms for multipath and RFI interference attacks. The techniques are based on antenna arrays. First, a twostage algorithm for a joint RFI interference and multipath mitigation was presented. It involves a prewhitening operation, followed either by STAPCA or a multicorrelator approach. Different combinations have been evaluated and compared. This has been done via software simulations using synthetic data and with real data recorded during a measurement campaign carried out at DLR’s maritime jamming testbed in the Baltic sea. The platform used to collect the data has been described as well as basic measurement scenarios with different interference types. STAPCA is able to tolerate more than 55 dB in case of PPD interference and more than 60 dB for AWGN noiselike interference.
Footnotes
 1.
It should be mentioned here that the contribution of all other satellites (except the one under consideration) is included in the noise term as well. This is to a large extent justified by the design criteria of the PRN sequences (near zero crosscorrelation and power level below the noise floor).
 2.
Which is the case for GPS/L1/CA.
 3.
In other words, left (right) centro hermetian matrices are conjugate axis symmetric w.r.t. the middle horizontal (vertical) axis.
 4.
In other words, centro hermetian matrices are conjugate point symmetric w.r.t. the middle point of the matrix, which may also lay in between entries if the dimensions are even. This is automatically the case for all array configurations, where we can find a center of point symmetry in the geometric layout of the elements.
 5.
The assumption here is, that the LOS component is the strongest one and the LOS component is visible in all windows. This implies, that the steering of the DLL is sufficient to keep the control loop in a steady state.
Notes
Acknowledgements
The measurement campaign was carried out in cooperation with our colleagues from the department for Nautical systems. We greatly acknowledge the support, especially from Stefan Gewies, in preparing and performing the campaign. The research leading to the results in this paper is part of the project EMSec (Echzeitdienste für die Maritime Sicherheit Security) and has received funding from the program Research for Civil Security of the Federal Ministry of Education and Research (BMBF) of the Federal Republic of Germany under the Grant FKZ 13N12744. This support is greatly acknowledged.
References
 1.Seafarer’s International Research Centre (SIRC) of Cardiff University, and Allianz SE, Safety and Shipping 19122012, From Titanic to Costa Concordia: An insurer’s perspective from Allianz Global Corporate and Specialty, München, London (2012)Google Scholar
 2.Italian Ministry of Infrastructures and Transports (Marine Casualties Investigative Body): Cruise Ship COSTA CONCORDIA, Marine casualty on January 13, 2012 (Report on the safety technical investigation), Pisa (2013)Google Scholar
 3.International Maritime Organization, NAV 54/25 Annex 12: Strategy for the development and implementation of enavigation, London (2008)Google Scholar
 4.John A.: Volpe National Transportation Systems Center: Vulnerability assessment of the transportation infrastructure relying on the Global Positioning System, Cambridge (2001)Google Scholar
 5.Grant, A., Williams, P., Ward, N., Basker, S.: GPS jamming and the impact on maritime navigation. J. Navigat. 62(2), 173–187 (2009)CrossRefGoogle Scholar
 6.Musumeci, L., Samson, J., Dovis, F.: Experimental assessment of Distance Measuring Equipment and Tactical Air Navigation interference on GPS L5 and Galileo E5a frequency bands. In: Satellite Navigation Technologies and European Workshop on GNSS Signals and Signal Processing, (NAVITEC), 2012 6th ESA Workshop on, pp. 1–8 (2012)Google Scholar
 7.Divis, D.A.: Redacted DHS report details privacy jammer risks. In: Inside GNSS, Online (2016)Google Scholar
 8.Kaplan, E.: Understanding GPS–Principles and Applications, 2nd edn. Artech House, Norwood (2005)Google Scholar
 9.Lloyd’s Register Group Limited, QinetiQ and University of Southampton: Global Marine Technology Trends 2030, Online (August 2015)Google Scholar
 10.European Global Navigation Satellite Systems Agency (GSA): GNSS Market Report, Online (2015)Google Scholar
 11.Teunissen, P., Montenbruck, O. (eds.): Springer Handbook of Global Navigation Satellite Systems. Springer, New York (2017)Google Scholar
 12.Ward, P.W., Betz, J.W., Hegarty, C.J.: Interference, multipath, and scintillation. In Understanding GPS: Principles and Applications, pp. 243–299. Artech House, Boston (2005)Google Scholar
 13.Guner, B., Johnson, J.T., Niamsuwan, N.: Time and frequency blanking for radiofrequency interference mitigation in microwave radiometry. IEEE Trans. Geosci. Remote Sens. 45(11), 3672–3679 (2007)CrossRefGoogle Scholar
 14.Niamsuwan, N., Johnson, J.T., Ellingson, S.W.: Examination of a simple pulseblanking technique for radio frequency interference mitigation. Radio Sci. 40(5) (2005)Google Scholar
 15.Milstein, L.B.: Interference rejection techniques in spread spectrum communications. Proc. IEEE 76(6), 657–671 (1988)CrossRefGoogle Scholar
 16.Gao, G.X., Sgammini, M., Lu, M., Kubo, N.: Protecting GNSS receivers from jamming and interference. Proc. IEEE 104(6), 1327–1338 (2016)CrossRefGoogle Scholar
 17.Castaneda, M.H., Stein, M., Antreich, F., Tasdemir, E., Kurz, L., Noll, T.G., Nossek, J.A.: Joint spacetime interference mitigation for embedded multiantenna GNSS receivers. In: Proceedings of ION GNSS 2013, Nashville, TN (2013)Google Scholar
 18.SecoGranados, G., FernndezRubio, J.A., FernndezPrades, C.: ML estimator and hybrid beamformer for multipath and interference mitigation in GNSS receivers. IEEE Trans. Signal Process 53(3), 1194–1208 (2005)MathSciNetCrossRefGoogle Scholar
 19.Vagle, N., Broumandan, A., JafarniaJahromi, A., Lachapelle, G.: Performance analysis of GNSS multipath mitigation using antenna arrays. J. Glob. Position. Syst. 14(1), 4 (2016)CrossRefGoogle Scholar
 20.Antreich, F., Nossek, J.A., SecoGranados, G., Swindlehurst, L.A.: The extended invariance principle for signal parameter estimation in an unknown spatial field. IEEE Trans. Signal Process 59(7), 3213–3225 (2011)MathSciNetCrossRefGoogle Scholar
 21.SelvaVera, J.: Efficient Multipath Mitigation in Navigation Systems. Ph.D. dissertation, Department of Signal Signal Theory and Communications, Universitat Politcnica de Catalunya, Spain (2004)Google Scholar
 22.Melvin, W.L.: A STAP overview. IEEE Aerosp. Electron. Syst. Magn. 19(1), 19–35 (2004)CrossRefGoogle Scholar
 23.Ward, J.: Spacetime adaptive processing for airborne radar. In: 1995 International Conference on Acoustics, Speech, and Signal Processing, vol. 5, pp. 2809–2812 (1995)Google Scholar
 24.Brunner, C., Haardt, M., Nossek, J.A.: On spacetime rake receiver structures for WCDMA. In: Proceedings Proc. 33rd Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, CA (1999)Google Scholar
 25.Sgammini, M., Antreich, F., Kurz, L., Meurer, M., Noll, T.G.: Blind adaptive beamformer based on orthogonal projections for GNSS. In: Proceedings of ION GNSS 2012, Nashville, TN (2012)Google Scholar
 26.Troetschel, F., Antreich, F., Nossek, J.A.: Spacetime adaptive principle component analysis for timedelay estimation. In: 8th IEEE Sensor Array and Multichannel Signal Processing Workshop (SAM). A Coruna, Spain (2014)Google Scholar
 27.SelvaVera, J.: Efficient Multipath Mitigation in Navigation Systems. PhD Thesis, Department of Signal Signal Theory and Communications, Universitat Politcnica de Catalunya, Spain (2004)Google Scholar
 28.Pillai, S.U., Kwon, B.H.: Forward/backward spatial smoothing techniques for coherent signal identification. IEEE Trans. Acoust. Speech Signal Process. 37, 8–9 (1989)CrossRefGoogle Scholar
 29.Sgammini, M., Antreich, F., Meurer, M.: SVDbased RF interference detection and mitigation for GNSS. In: Proceedings of ION GNSS+ 2014, Tampa, FL (2014)Google Scholar
 30.Marinho, M.A.M., Antreich, F., de Costa, J.P.C.L., Nossek, J.A.: A Signal Adaptive Array Interpolation Approach with Reduced Transformation Bias for DOA estimation of Highly Correlated Signals. In: Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2014, Florence (2014)Google Scholar
 31.Marinho, M.A.M., Costa, J.P.C.L.D., Antreich, F., Menezes, L.D.: Unscented transformation based array interpolation. In: Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2015, Brisbane (2015)Google Scholar
 32.Navstar GPS Joint Programm Office: Navstar GPS Space Segment/Navigation User Interfaces (2006)Google Scholar
 33.Appel, M., Iliopoulos, A., Fohlmeister, F., Marcos, E.P., Cuntz, M., Konovaltsev, A., Meurer, M., Antreich, F.: Experimental validation of GNSS repeater detection based on antenna arrays for maritime applications. CEAS Space J (2018)Google Scholar
 34.Heckler, M., Cuntz, M., Konovaltsev, A., Greda, L., Dreher, A., Meurer, M.: Development of robust safetyoflife navigation receivers. Microw. Theory Tech. IEEE Trans. 59(4), 998–1005 (2011)CrossRefGoogle Scholar
Copyright information
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.