Review on Vector Tracking Application to GNSS Receiver
An Qi, Li Chuanjun，An Hongfei
Key Laboratory of Dynamics and Control of Flight Vehicle, Ministry of Education
School of Aerospace Engineering, Beijing Institute of Technology, Beijing, China
Email:
Abstract – GNSS receiver based on vector tracking loop joints information from all channels to estimate user’s position, velocity and time. In vector mode, signal tracking block and navigation processor are coupled together, so all channels can transmit and share information. The actual satellite ephemeris is adopted to verify the performance of tracking loop. Compared to STL, VTL improves weak signal tracking, anti-jamming ability and rapid signal recovery after a satellite blockage. This paper describes the development of vector tracking, vector tracking loop architecture, system modeling and algorithm. Major applications are summarized in terms of vector tracking.The aim of this review paper is to provide anextensive overview of existing vector tracking research and a guide for development in the future.
Keywords – global navigation satellite system, vector tracking loop, vectortracking application.
- INTRODUCTION
Global Navigation Satellite System (GNSS) is capable of providing users with positioning and timing service all around the world.In traditional receiver mode,there are two parts in each channel: delay lock loop (DLL) is used to estimate the pseudoranges and pseudorange rates, and phase lock loop (PLL) or frequency lock loop (FLL) is used to estimate carrier phase or Doppler shift .The results are forward to the navigation processor. Navigation processor uses the estimated information to estimate the receiver’s position velocity and time (PVT) finally [1,2]. A single tracking loop completes signal capture, tracking, bit synchronization, frame synchronization and navigation message extraction, so all tracking loops are independent from each other.Although the scalar tracking loops(STL)perform well under better power-to-noise density ratio and low dynamics, this architecture is still not optimal.
Nowadays, GNSS is employed widely in aviation field for positioning.In high dynamic condition, movement of satellite and receiver produces Doppler frequency. Tracking loop performance is undermined by varying Doppler offset. The traditional scalar tracking loop is unable to complete weak signal tracking task in complex electromagnetic condition.GNSS signal after spectrum spreading and long distance travelling is significantly weak. GNSS malfunction or disruption can cause safety problems. Traditional tracking loop is the weak point in a GNSS receiver as multipath, signal block and interference. On the contrary,vector tracking loop(VTL) based on multiple channels joint fully exploits the inherent information of each channel and the effective along each of ,,axes of user’s position increases with more satellite information. In this structure, each narrow bandwidth tracking loop is aided by a wide bandwidth common loop. This effectively solves the problem of signal tracking in weak signal environment.The research to VTL is benefit to improve the tracking performance of receiver [3].
Similar to STL, VTL consists of a vector delay loop (VDLL) and a vector carrier loop including one or both of a vector frequency lock loop(VFLL) and a vector phase lock loop (VPLL). In an ultratight integration strategy, VFLL architecture can perform more reliable Doppler measurements compared to VPLL. This paper mainly illustrates VDLL and VFLL.
- The Development of Vector Tracking
Vector delay lock loop concept was first proposed by Spilker in [4], it presents that in VDLL mode, Extended Kalman Filter (EKF) combines the tasks of signal tracking and navigation into a single algorithm, so there is no boundary between signal processing and navigation calculation. Noise is reduced in all channels making loop less likely to enter the non-linear tracking regions and the tracking performance is more robust. At the same time, the strong satellite signal can estimate the user’s PVT and the weak signal in turn can be predicted on the basis of the state of the user. This method has been extensively studied. There arethree main categories in VTL research: performance comparison of vector tracking loop and scalar tracking loop [5-14], VTL modeling and filter design [15-19] and application of VTL.
Matthew Lashley from Auburn University has presented a series researches of VDLL [5-11].Through performance comparison between VDLL and SDLL in several environments, VDLL implementation in GNSS receiver has more advantages. The ability of VDLL function at low ratio and operating during brief signal outages and rapidly reacquire blocked signals are explored. In relative research, covariance analysis and Monte Carlo simulation are used to analyze tracking threshold. A vector delay/frequency lock loop (VDFLL) is used to estimate the power of weak signals and analyze signal power of GPS C/A code signal in three different environments, and the results show that VDFLL can still achieve energy estimate even 10dBHz for GPS C/A code signal [12].
A great deal of work has been done on the modeling of VDLL based on EKF. This algorithm can reduce tracking threshold in theory, and the tracking threshold is lower if more satellite signals are tracked [7,8]. Researchers from FAF Munich University use navigation solution to model the measured pseudoranges, deduce the estimator and relative covariance matrix in SDLL and VDLL mode. According to quantitatively analysis, VTLL is more accurate than STLL when satellite signals are fading, as the experiments demonstrate [13].At present, modeling and filter design based on VDLL or VDFLL are studied intensively, most of which conduct simulation experiments [15-19].
Moreover, vector tracking application to Ultra-Tight GPS/INS integration is a hot subject of much research.GNSS can provide accurate position and time information.But GNSS signals are prone to jamming or attenuating. Once effective signals are weak, the position accuracy is deteriorated. While the inertial measurement units(IMU) with high accurate measurement rates and immune to outer interference, is able to provide measurements of the receiver’s acceleration and angular rates. In turn, the tracking performance can be improved since the tracking loop parameters can be optimized beyond GNSS-only case. Ultra-Tight GPS/INS integration fuses the traits of GNSS receiver and INS.Because of its ability of mitigating noise, improving accuracy and anti-jamming, it can improve performance of signal tracking in complex electromagnetic environment [20-24].
- tne basic principles of VTL
A.Structure of VTL
In order to track the visible satellites, the receiver must have knowledge of the signal code phase at the time of transmission and Doppler frequency before demodulating the navigation messages from the signal and compute the receiver’s PVT. Demodulating process requires the local harmonic signal and received signal must have pretty good phase alignment.
GPS receivers employ tracking loops to track satellite signals independently.DLL is used to estimate the PRN code phase of the incoming signal in order to generate a replica to remove the pseudorandom. Carrier tracking loop is implemented to keep tracking the carrier frequency or phase.STL uses discriminator in code and carrier tracking loop to estimate code phase and Doppler frequency measurements [1] [25].The outputs of the loop filters provide corrections to their respective NCOs. The code NCOs control generator to produce three shifted visions of the replica code (E, P, L).The carrier NCOs generate in-phase (I) and quadraphase (Q)components of the baseband CW signal .The replica code and carrier signals are correlated with incoming signals, and the results are passed to discriminator to generate estimates of code phase and carrier offset. Tracking loop attempts to minimize the GPS pseudorandom code delay error and the Doppler frequency error for each satellite channel in a close loop. Tracking results are fed to navigation processor. STL architecture is shown in Fig.1.
Fig.1 Scalar Tracking Loop Architecture
As illustrated in Fig.1, STL processes signal follows one direction and there is no feedback from the navigation processor to correct the tracking loop. In fact, each channel of STL acts independently and the performance of tracking depends on the feature of discriminator and filter. If the signal becomes degraded or antenna is hidden during maneuver, the tracking process is easily interrupted and signal is lost. Although GNSS signals are from different satellites, they have the same receiver, e.g. dynamic stress and thermal noise.On the contrary, user’s PVT and satellite navigation message from ephemeris can be used to predict the visible satellite PRN code and Doppler frequency. VTL makes fully use of inherent correlation among channels and aids each other [16].
Based on the input to navigation processor, there are two different architectures in vector tracking loop: coherent and non-coherent. The coherent architecture includes centralized coherent and federated coherent architecture. In centralized coherent architecture mode, navigation processor receives correlator outputs as inputs to compute PVT solution, while federated coherent architecture adds a bank of pre-filters to filter correlator outputs.
In the non-coherent architecture, the navigation filter inputs results from discriminator.Fig.2 shows the architecture of VTL, and the cascaded structure of tracking and navigation filters of the conventional receiver is replaced in this architecture with a central navigationfilter [9]. The correlator outputs are used to generate code error and carrier frequency error measurements.
An EKF as its navigation filter corrects its state vector. Once code phase error and carrier frequency error measurements are updated and then fed into NCOs to control the generator aligned with the received signal [26, 27].The control loop is closed through the central EKF. Signals from different satellites are processed together, which effectively reduces the noise bandwidth of tracking loop and improves performance of user’s dynamic stress in a global loop closure, so dynamic, accuracy of receiver is improved.
Fig.2 Vector Tracking Loop Architecture
B.Vector Tracking Modeling
The vector loop navigation filter based on EKF combines information from all satellite to correct system states error. Implementation of non-coherent vector tracking loop, state vector and measurement vector play important roles in modeling of navigation filter. The states of EKF are receiver’s position, velocity, acceleration, clock bias and drift. Suppose the number of the visible satellite is, is the time tag. The state equation is shown as (1)
(1)
Where,
The state transform matrixis
(2)
:update time of navigation processor, usually varying from 20 ms to 1second based on implementations, noise levels and user dynamics
: process noise vector, defined as white Gaussian noise.
In VDLL mode, the outputs of the discriminators are used as the measurements for EKF. In order to reduce sensitivity of signal amplitude changes and dependence of the carrier lock loop, the measurement of code error uses outputs of early and late correlators. The discriminator can track phase difference within 0.5 chip. The early-late-power discriminator is defined by:
(3)
Where,
: integrate and dump interval;
,: in-phase early and quadrature early;
,: in-phase late and quadrature late.
The pseudorange errors are shown in (4):
(4)
Whereis code error, is frequency of C/A code and
is speed of light.
The linearized measurement equation is given by:
(5) Where are pseudorange errors of all the visible satellites.
is measurement matrix:
(6)
Where are the components of the line-of-sight unit vector from the user’s estimated position to the nth satellite.
: ;
: numberof visible satellite
:geometric range of satellite i
: a vector of additive measurement noise
The navigation states can be computed based EKF algorithm by the equations:
Prediction:
(7)
Prediction mean square error:
(8)
Kalman gain:
(9)
Estimation:
(10)
Estimation mean square error:
(11)
Where the measurement noise covariancecan be deduced in[15],the disturbance noise covariance is related to the user’s movement in different situation.
- application of vector tracking
VTL is applied to the ultra-tight (or deep) integration of receiver with an inertial navigation system (INS) to improve overall tracking performance, which is the basic principle of GNSS/INS integration. Ultra-tight integration is an emerging integrated navigation and performs effectively in weak signal or high dynamic environment. In the Deep Integration architecture, Kalman filter is as the navigation processor and the Inertial Measurement Unit is used as an input to the navigation Kalman filter. Navigation information from GNSS receiver and IMU are conjunct and then used to control the NCOs in each channel. There are two different categories defined by architecture: concentrated Kalman filter and federal Kalman filter. Fig.3 shows a structure of federal deep integration.
Fig.3 Federal Deep Integration Architecture
VDLL is implemented to estimate the PRN code phase and VFLL is implemented to remove Doppler offset. A pre-processing filter is employed to smooth the discriminator outputs reducing demand for real-time processing. Because of independent of each pre-processing channel, it is available to improve failure detection, correctionand redundancy management. It is an efficient approach to avoid wrong information providing to navigation filter. In addition, it benefits to construct more accurate mathematical modeling to estimate residuals and optimal estimate both receiver tracking performance and INS navigation solutions[28-30].
Anti-jamming is another application of vector tracking. The power of GNSS signal received near the surface of the earth is extremely low. Signal is vulnerable to suffer intentional or unintentional interference. Traditional GNSS receivers use STL architecture. Each channel tracks one satellite. The errors calculated by discriminator are filtered and then provide feedback to NCOs. Because of outer signals, thermal noise and dynamic stress, local replica PRN code do not align with the received code completely. In VTL mode, the errors from discriminator do not provide feedback to NCOs directly. Pseudorange errors and pseudorange-rate errors are used as an intermediate variable in navigation processing. Navigation processor directly controls the feedback to align the receiver’s replica signals with satellite signals. Because of the more centralized control, the VTL architecture is considered to be more robust to jamming. Thus, this technique is more accurate navigation performance during jamming[31,32].
- conclusions
Recent years, there has been incredible innovation in vector tracking technique, especially in system modeling and model parameter calculation. The advances of vector tracking in weak signal processing, enhanced anti-jamming ability and continuously tracking after a satellite blockage are tested. But in harsh environment,a fault from a channel of the receiver will corrupt the tracking of other channelsin vector tracking loop based on discriminator. Advanced vector loop, adaptive tracking loop, is proposed to improve the stability of vector tracking loop [33,34]. An adaptive navigation filter lessens the effect of channels with low-quality signals. This method adjusts noise covariance matrix according to measurement, and Kalman filter gain changes correspondingly. So it enables to achieve tracking stability. Aiming at loss of lock of VTL, there is lack of detailed theoretical analyses. The tracking threshold is deduced by scalar tracking experience. This is an orientation of future research.
In practical application, it is very difficult to implement vector tracking on a real-time platform because vector tracking algorithm or ultra-tight GNSS/INS integrated navigation algorithm needs to process a mass of data immediately. If it is applied in traditional application-specific integrated circuits (ASICs), it will consume a significant number of hardware resource [35]. Field-programmable gate array (FPGA) plays an important role in system on chip approach because of its speed, capability of performing multiple high frequency operations in parallel, short design cycle. The FPGA vendors also provide a series of intellectual property cores invoked directly. In addition, digital signal processor (DSP) is a capable device of platform for software receiver. DSP is powerful at performing arithmetic tasks. Framework of FPGA &DSP is hopeful to offer more support for the implementation of vector tracking technique.
ACKNOWLEDGMENT
This paper issupported by the National Natural Science Foundation of China and the key laboratory of dynamics and control of flight vehicle, Ministry of education, school of aerospace engineering, Beijing Institute of Technology.
REFERENCES
[1]Kaplan E, Hegarty C. Understanding GPS: Principles and Applications.Norwood Massachusetts, USA: Artech House, 2006.
[2]JUANG Jyh-ching，CHEN Yu—hsuan. Phase／frequency tracking in a GNSS software receiver. IEEE Journal of Selected Topics in Signal Processing，2009，3(4):651-660.
[3]Lashley M, Bevly D M. What are vector track loop and what their benefits and drawbacks.Inside GNSS, May/June 2009,4(3):16-21
[4]SpilkerJ.J, Fundamentals of Signal Tracking Theory,in Global Positioning System:Theory and Applications, Parkinson B.W, Spilker J.J.,Editor.1994,American Institute of Aeronautics and Astronautics,Inc:Washington,DC.pp.245-327.
[5]Lashley M,Bevly D M. Comparison of traditional tracking loops and vector based tracking loops for weak GPS signals.ION GNSS 2007.September,2007,pp.789-795.
[6]Lashley M,David M.Bevly, John Y.Hung.A valid comparison of vector and scalar tracking loops.IEEE/ION PLANS,Indian Wells,USA,May 4-6,2010.
[7]Lashley M,David M.Bevly, John Y.Hung.Performance analysis of vector tracking algorithms for weak GPS signals in high dynamics.IEEE J SelTop Signal Proc,2009,3(4):661-673.
[8]LashleyM.Modeling and Performance Analysis of GPS Vector Tracking Algorithms,Auburn,Alabama,Auburn University,2009.
[9]Lashley M,Bevly D M.Analysis of discriminator based vector tracking algorithm.ION NTM,San Diego,USA,January 22-24,2007.
[10]Lashley M,Bevly D M.Performance comparison of deep integration and tight coupling.ION GNSS,Poland,USA,September 19-23,2011.