Improvements of GNSS Receiver Performance Using Tightly Coupled INS Measurements

Transcription

1 Imrovements of GNSS Receiver Performance Using ightly Couled Measurements Christian Kreye, Bernd Eissfeller, horsten üc Institute of Geodesy and Navigation (IfEN) University FAF Munich Neubiberg (Germany) BIOGRAPHIES Christian Kreye received a diloma in Geodesy in 998 from the University of Hannover. He has started as a research associate at the Institute of Geodesy and Navigation of the Federal Armed Forces in 999. His research area has been different alications of integrated GNSS/ systems, with secial emhasis on low-cost. Prof. Dr.-Ing. Bernd Eissfeller is Full Professor and Vice-Director of the Institute of Geodesy and Navigation at the University of the Federal Armed Forces Munich. He is resonsible for teaching and research in the field of GPS/GONASS and inertial technology. ill the end of 993 he wored in industry as a roject manager in the develoment of GPS/ navigation systems. From 994- he was head of the GNSS aboratory. horsten üc is a research associate at the Institute of Geodesy and Navigation at the University of the Federal Armed Forces Munich. He studied Electrical Engineering at the universities in Stuttgart and Bochum. Since 998 his research area has been the integration of inertial navigation system with differential GPS ABSRAC Aart from sace and control segment the GNSS erformance deends on several arameters, e.g. signal roerties, multiath conditions, signal interrutions, signal-to-noise ratio, dynamics and receiver errors lie thermal noise and oscillator instabilities. he imact of these influences are based on the behaviour of the Phase oc oo (P) and the Delay oc oo (D) imlemented in the GNSS receiver. o guarantee availability of a recise navigation solution the loo errors have to be significantly lower than the loc thresholds. he first art of the aer resents investigations about the interaction of loo bandwidth, signal-to-noise ratio, accelerations, oscillator Allan variance and oscillator vibrations. In order to decrease the influence of receiver noise the tracing loo bandwidth has to be reduced. Consequently under high dynamic conditions or in a jamming environment the dynamics of the seudoranges have to be removed from the loo signal to ee them in loc. If the vehicle dynamics on the other hand is measured by a low cost, the range velocities, accelerations and higher derivatives can be comuted by aid of inertial data. In this context investigations about the different couling rinciles between GNSS and are resented. With tightly couling filtered the -data is fed bac directly into the GNSS receiver code and carrier tracing loos. A ossible filter design for low cost -data and its transfer function are described. In oosite to the unaided receiver the tightly couled sensors are no longer deending on the accelerations but on the accelerometer bias and gyro drift behaviour of the. he Schuler frequency dominates the jitter value of the loos. he aer demonstrates that a gyro rate of aroximately /h is necessary to suort the hase loc loo. Concerning the delay loc loo a gyro rate of /h is sufficient. In next art of the aer a simulation tool for a tightly couled GPS/-Sensor is described. Used algorithms and sensor models are reviewed. he data flow beginning with original data and ending with the integrated navigation solution is resented. First simulation results are also shown. Advantages and roblems of the couled sensor system are mentioned. he develoment of an exerimental tightly couling system using the MIE GPS Architect receiver an the ION N- Inertial Measurement Unit is the ey oint of the final section of the aer. Besides hardware and software asects methods of time synchronisation and data rocessing are addressed. INRODUCION GNSS Receiver Measurement Errors and racing hresholds he GNSS signal acquisition and tracing rocess is a two-dimensional signal relication rocess. In order to

2 derive seudoranges between satellite and receiver references of PRN code and the carrier frequency lus Doler must be relicated and correlated with the incoming satellite signals. he signal generation is rovided by a numerical controlled oscillator (NCO). he frequency alignment is controlled in a closed loo configuration. A basic structure of a delay (D) or hase loc loo (P) for a single SV is resented in figure. σ B C / N C / N λ π with λ carrier wave length (3) Deending on values for loo bandwidth and signal-tonoise ratio the resulting thermal noise can be resented in figure : x ( ) hermal Noise x( ) - oo- F() E() NCO K Allan Deviation Vibrations Figure. Rudimental delay or hase loc loo Only if this control loo is in loc seudoranges and carrier hases can be obtained. When total GNSS measurement errors exceed the tracing threshold the receiver looses loc. hese errors deend on carrier frequency f, signal to noise ratio C/N, reintegration time, loo bandwidth B, oscillator stability and receiver dynamics. As it can be seen in literature [e.g. WARD 998] the rules-of-thumb concerning the tracing thresholds of P and D can be exressed by the following formulas: e t σ P σ σ A σ V 5 () 3 e( t) C σ D σ σ A σ V () 3 6 with σ P P error σ D D error σ hermal noise (-sigma) σ A Allan variance oscillator jitter (-sigma) σ V Oscillator vibrations (-sigma) e(t) Dynamic stress error Chi length C In relation to a GNSS lie GPS using the -band the tracing thresholds can be fixed to 8 mm for the hase loc loo and 5 m (Y) or 5 m (C/A) for the delay loc loo resectively. As the first comonent of formula and the influence of the thermal noise should be investigated. Concerning the hase loc loo it can be obtained by the equation: ( ) Figure. P thermal noise In order to decrease the thermal noise a reduction of the filter bandwidth is necessary. he errors caused by the oscillator are neglected in many accuracy investigations. he value for the Allan variance oscillator jitter conforms to its stability and can increase to 3 mm. he frequency variations caused by the receiver dynamics are called oscillator vibrations. Also this error can reach values u to.5 mm using a random vibration ower curve of 4.8 g /Hz from 5 Hz to 5 Hz (. mm vibration amlitude). Esecially in high dynamic environments the loc-in condition of the receiver loos cannot be et. In this case the dynamic stress error e(t) violates the tracing threshold of the control loos. e (4) ω ( t) x A reduction of the dynamic stress error without external aiding is only ossible if the natural frequency of the loo filter ω can be increased. his however means a higher loo bandwidth at the same time. Higher thermal noise would be the result. he following grahs demonstrate the behavior of the total P-error in relation to different loo bandwidths:

3 Figure 3. otal P-Error, loo bandwidth 5 Hz Figure 6. otal D-error, loo bandwidth.75 Hz As the thermal noise level of the code is much higher than the value of the carrier the loo bandwidth must be smaller. he lower robustness against dynamic stress can be comensated by carrier aiding of the code tracing loo. he variations of the total D error concerning the Y-Code is demonstrated in figures 5 and 6. Figure 4. otal P error, loo bandwidth 5 Hz In oosite to the hase loc loo in the delay loc loo there are effectively only two erformance characteristics: the code loo thermal noise error and the maximum line-of-sight dynamic threshold. he thermal noise error can be comuted by: B σ (5) C C N C N GNSS/ COUPING PRINCIPES Because of different error influences based on sace segment, signal roagation or receiver noise the accuracy otential of a GNSS system is not sufficient for all alications. Additionally the availability and data frequency can be inadequate. herefore many navigation or surveying systems use integrated GPS/ systems to combine the only short-term stability of an sensor and the long-term stability but noise behavior of an GNSS sensor. Different couling rinciles of GPS/ systems are ossible. Most GPS/ integrations are loosely couled, giving u a great deal of erformance in return for simlicity of integration. Using this rincile the osition and velocity estimates of a GNSS receiver are used as observations in an filter, thereby cascading two navigation filters (see figure 7; : osition, velocity, attitude). An estimation of errors, a reduction of GPS-noise and a bridging of GPS outages are ossible. Yet single seudoranges cannot be used to udate the filter. Inertial Navigation System Navigation Processor Errors GNSS Receiver PV(A) Figure 5. otal D-error, loo bandwidth. Hz Figure 7. oosely couling rincile

4 In a closely couling architecture only one navigation filter generates the integrated navigation solution with the information of GPS observations and navigation data (see figure 8). In this case the GPS measurements of seudoranges and deltaranges are usable even when fewer than four satellites are traced. Knowing the osition lost ambiguity values can be restored more quicly. Inertial Navigation System Navigation Processor Errors rocess the loo is closed by use of conventional control theory. aided GNSS receiver Inertial Navigation System hermal Noise s ( t) s( t) - Detector (active) NCO oo (assive) Inertial data RCVR cloc-, sensor errors e(t) Phase- or Delay oc oo GNSS Receiver Pseudoranges Pseudorange-Rates Figure 8. Closely couling rincile But in an observation environment with a low signalto-noise ratio and high receiver dynamics no seudoranges can be comuted caused by transgression of the tracing thresholds of the P or D. In order to avoid this ind of GNSS outage additionally information (e.g. velocity) is added into the tracing loos of the receiver. hereby the receiver dynamics can be subtracted from the rest of the signal. he dynamic stress error is minimized. Furthermore the loo bandwidth can be decreased to reduce the thermal noise error influence. his ind of GPS/ integration is called tightly couling (see figure 9). Other advantages of this method are the fast reacquisition of lost satellite signals and the better anti-jamming erformance. As the information influence the signal rocessing of the GNSS receiver the realization of this couling rincile requires secial hardware equiment (see in the following sections of the aer). Delay oc oo Phase- oc oo Aided GPS Receiver Pseudo-RangeRate Pseudo-RangeRate Ehemeris Range Rate RangeRate Figure 9. ightly couling rincile Process Comuter filter Errors Navigation Processor Inertial Data Inertial Navigation System In ultra-tightly or deely couling the filter is integrated on tracing loo level, i.e. each loo D and/or P is combining the early late detector outut with transformed and corrected range and range rate data from the in an otimal filter. After this Figure. Ultra-tightly or deely couling rincile he basic rincile may be described as follows for a st order digital D, where the state model may given as follows (rocess noise not considered) τ Τ Τ with τ Τ Τ B ( 4 B ) 4 B τ Τ Τ loo estimate of seudorange geometric range (satellite receiver) geometric range-rate (satellite receiver) tracing loo bandwidth re-detection integration interval ( msec for GPS) he following observations are used to u-date the loo filter : i) Detector outut (non-coherent E ) D Q E a c Q I I ( d)( Τ τ ) ε E with a amlitude of GPS signal c chi length d correlator sacing Q, I quadra-hase, in-hase signal E, early, late ε detector noise ii) derived range and range - rate with b cloc c Τ Τ bˆ d dt cloc bˆ cˆ cloc d dt ε cˆ ε ε cloc error estimate from central filter inertial error estimate from central filter

5 his aroach may be easily generalized for the carrier loo, carrier aiding and higher order loos. HEOREICA SIMUAION In order to study the imrovements of GPS erformance using the tightly couling architecture a detailed investigation of the receiver loos behaviour have to be made concerning the additional information. In comarison to figure the extension of e.g. the P configuration is resented in figure. x ( ) G() x Inertial Navigation System hermal Noise x( ) ( ), x ( ) i i - - Allan Deviation oo- F() NCO Vibrations Figure. racing loo with inertial aiding K E() Using a theoretical simulation of the closed loo answers about the following questions should be available: What ind of filter function has to be used decreasing the error influence of the? Is it ossible to reduce the filter bandwidth and therewith the thermal noise? What ind of is suitable to guarantee the tracing threshold conditions? Some fundamental nowledge about inertial navigation and signal rocessing are combined to develo a simlified but meaningful mathematical descrition of the discussed control loo. he derivation of an observation equation is based on the fundamental equation of inertial navigation. he secific force a can be comuted by: a x g ; GM g g x x 3 R R with: a x g R secific force osition gravitation earth radius (6) Including a simle error model with an accelerometer bias B and a gyro drift D and restricted to the horizontal lane it can be written (t time): g a δ a x x ; δa B g D t R (7) A transformation from time to frequency domain using the alace transformation and resorting yields to the following equation of inertial velocity: x i ( ) A g R x B g D ( ) 3 (8) In signal rocessing the closed loo configuration resented in figure can be described as: [ K F( ) ] X ( ) K F( ) x ( ) K F ( ) G( ) xi (9) hereby the added term contents the influence of inertial aiding. Design ie it is shown in the equation 9 the effect of inertial data deends on the used filter function G(). In relation to the tightly couling rincile the dynamics of the seudoranges should be subtracted from the total signal. According to this criterion G() have to be derived. he dynamical tracing error E() is comuted by: E ( ) x ( ) x( ) () A combination of equation 8 and 9 yields to: E ( ) () K F K F K F ( ) G( ) ( ) G( ) K F( ) ( ) B 3 g R K F X ( ) ( ) G( ) g K F( ) 4 In order to guarantee the indeendency of E() and X () the first term must be zero. his is ossible if g Γ K F( ) G( ) R () Γ designates in equation the couling coefficient between GNSS and. A value of one means the full integration caacity, zero signify the unaided GNSS receiver. Using equation the filter function G() can be derived: D

6 ( ) G Γ K F with ω ( ) ( ω ) S 3 g R s Schuler frequency (3) he amlitude resonse G(ω) of this filter function concerning a second order tracing loo can be determined by: G ( ω) 4 Γ ω (4) K α ω ( ω ω ) In order to evaluate the filter characteristics the result of equation 4 has been lotted in figure. An asymtotical trend can be seen at the Schuler frequency. Otherwise long term variations are suressed. According to the chosen filter bandwidth high frequencies u to Hz ass the filter nearly unchanged. As a basic aroach this might be a reasonable solution based on the nown short term stability of data. In a next ste secial filters have to be develoed qualified for the characteristic error erformance of the used inertial sensor. S First a second order hase loc loo is investigated. he following grahs should demonstrate the influences of loo filter bandwidth B and error erformance (accelerometer bias B, drift D)concerning the total loo error (-σ). he current values are secified in the figure descrition. he couling coefficient is fixed by.. Figure 3. P; B : 5 Hz, D:. /h, B:. m/s.5.4 G(ω).3.. B. Hz B. Hz ω [Hz] Figure. Amlitude resonse of filter B5. Hz If the resented filter function G() is inserted in equation and after retransformation from frequency to time domain according to equation 4 the dynamical tracing error of the tightly couled sensor can be derived (see equation 5). Assuming a couling coefficient of one the dynamic tracing error is indeendent of the seudorange acceleration. Instead of this errors lie acceleration bias B and gyro drift D influence the tracing error. e ( t) ( Γ ) ω Γ B Γ D g sin( ωs x cos( ωs t) ω ω ω S t) Figure 4. P; B : 5 Hz, D:. /h, B:. m/s Because of the tightly couling architecture the Perror no longer deends on the seudorange acceleration but on the time. he Schuler frequency dominates the error behavior of the tracing loo. If the loo bandwidth is decreasing an increase of tracing error can be recognized. At times of maximum amlitude the error transcends the tracing threshold already at higher signal-to-noise ratios C/N. (5) Simulation results Using the dynamic tracing error resented in equation 5 the rules of thumb concerning the tracing thresholds of P and D (equation and ) can be evaluated for a tightly couled configuration.

7 Figure 5. P; B : 5 Hz, D:. /h, B:. m/s technology (MEMS), /h is only ossible with a fiber otic gyro (FOG). Higher accuracy requirements are only rovided by more exensive ring laser gyros (RG) of high recision. he following statements characterize the P investigations: oday MEMS-technology cannot meet the requirements to suort the hase loc loo. Imrovements of dynamical indeendence are disturbed by a high error level. A P suort using high recision allow filter bandwidths down to.5 Hz. But exensive sensors limit the area of alication. As a comromise with fiber otic gyros can be used. ie it is resented in figure 7 the P is et in loc with an reduced bandwidth of Hz. he aiding of the code loc loo is less critical than of the P, but because of different chi length it must be distinguished between Y- and C/A-Code. An evaluation of equation yields to the following results concerning the tightly couling configuration (figure 8 and 9): Figure 6. P; B : 5 Hz, D:. /h, B:. m/s Figure 7. P; B : Hz, D:. /h, B:. m/s ie it is resented in figure 4, 5 and 6 using a more accurate inertial sensor reduce the amlitude of the Schuler frequency, so that the bandwidth can decreased again. But higher accuracy of the means increased costs. While a gyro rate of /h can be rovided by an with micro electrical mechanical Figure 8. D (Y); B :. Hz, D:. /h, B:. m/s An inertial sensor with a gyro drift of. /h is able to suort an D with a loo bandwidth of. Hz tracing the Y-Code. he tracing threshold of nearly 5 m is not be violated down to an signal-to-noise ratio of db-hz. he loose-of-loo condition concerning the C/A-Code rests with nearly 5 m. If a decrease of filter bandwidth down to.5 Hz is sufficient gyro rates of. /h ee the D in loc. In oosite of the P to suort a code tracing loo with data the today available MEMS technology meets the necessary requirements of tightly couling.

8 Figure 9. D(C/A); B :.5Hz, D:. /h, B:. m/s NUMERICA SIMUAION OO In order to verify the theoretical results and to reare a lanned ractical test system a numerical tightly couling simulation tool (ICOSIM) is in develoment. For these uroses an existing imlemented model of a generic GNSS receiver is extended and adated with an simulation tool. Esecially answers concerning the following questions are exected: What ind of filters are ossible with different error models? What haens in case of GNSS signal outages caused by the environment? What means a GNSS outage concerning the system accuracy? Can the reacquisition time be reduced? What ind of feedbac effects have to be exected because of error estimation using observations of the combined system? Environment Geometry Signal Generator GNSS Receiver Channel # Code and Phase Channel #n Code und Phase Prozessor Generated Position Constellation Geometry Signal Error Modell Pseudorange # Velocity, acceleration Pseudorange #n Velocity, acceleration Udate Current Wayoint Geometry Overview of the Control and Data Flow in ICOSIM In figure the control and data flow of the main rocessing modules is shown. he original data consists of three arts: first the environment that allow to simulate GNSS signal outages caused by different objects lie buildings or trees, second the constellation containing information about the satellite geometry and the GNSS signal structure and last the current wayoint data to describe the simulated osition of the integrated GNSS/ system over time. Using this information the ossible ranges between satellites and receiver are calculated and transformed in time and hase delays of the incoming GNSS signals. Furthermore in the signal generator the resective received signals s(t) are comuted (see equation 6). s ( t) a c( t τ ) sin( ωt θ ) (6) with: c( ) code bit τ time delay θ hase delay At the same time in the simulation element the receiver wayoints are transformed into high frequency accelerations and gyro rates in a body fixed coordinate system. Using a secific error model systematic (accelerometer bias and gyro drift) and stochastic errors (noise) are added to the inertial data. Furthermore geocentric acceleration and velocity in an earth fixed coordinate system are obtained by a stradown algorithm. In combination with the current satellite constellation the accelerations of the ranges can be calculated. In order to derive the code and hase measurements from the incoming GNSS signals they have to be correlated with a relica signal generated by the receiver model. his is only ossible if both carrier and code frequency and the hase delay is continuously traced in the code and hase loc loos of each receiver channel. he mathematical model describing the loos consists of two couled non-linear ordinary differential equations (7) and (8). Following the tightly couling rincile the acceleration of the seudoranges generated by the and transformed from meter to wavelength are added to the simulated hase loc loo (eq. 8). If carrier aiding is taen into account also the code tracing loo is suorted by the information. he continuous solutions of these differential equations allow statements about the current stability conditions of the simulated control loos. Figure. Diagram of the tightly couling simulation tool

9 τ B θ B ( d) [ i i q q ] c e c S N l δs τ e l λ πc θ ( ξω dt ω ) i q θ 4ξω ( ω ) δsθ θ S N (8) In equation 7 and 8 i and q are the in- and quadrahase comonents concerning the early e, late l and unctual signal. he i and q comonents are the result of the correlation rocess in the receiver and can be aroximated by: i r q r η R η R ( τ r) sin( θ ) ( τ r) cos( θ ) (7) (9) he r taes on the values, D/ and D/ for the unctual, early and late channel resectively. he symbols in equation 7, 8, 9 mean: B noise bandwidth of the D C chi length of the code λ carrier wavelength c roagation velocity τ code loo error θ hase loo error D correlator sacing S/N signal-to-noise ratio δs detector noise (code and hase) θ line-of-sight (hase) integration time of the correlator ξ daming factor ω characteristic frequency of the P R( ) correlation function of the code he osition rocessor calculates a least square solution, using the ranges that the Ds and Ps in the receiver deliver. Smoothing with hase measurements is also ossible. he navigation solution of an unaided inertial sensor is drifting away over time. Esecially the low-cost sensors that should be used in a ractical system show only a short time stability. herefore in fixed intervals the results of the osition rocessor are used in a filter to imrove the navigation solutions and to estimate the simulated error arameters of the inertial sensors. Concerning a ractical system an evaluation of this feedbac constellation is ossible. First results of the simulation element emhasize the influence of the receiver movement concerning the total seudorange dynamics. In this case (see figure ) slow land navigation is simulated along a testath over 7 seconds resented in figure. he seudorange acceleration to a satellite with degree elevation without any receiver movement is a nearly constant value of.4 m/s². [m/s] Figure. Course and velocity during the testath During the vehicle movement this acceleration change esecially in turns for relatively large values (see fig. ). herefore also in land navigation the jer stress should not be neglected esecially if the bandwidth of the tracing loos is decreased. [m/s^] Pseudorange Acceleration (static case) Pseudorange Acceleration (dynamic case) Figure. Influence of receiver dynamics PROOYP SYSEM DEVEOPMEN he theoretical bacground of tightly couling and the results of the theoretical and numerical investigations define the configuration of the ractical system. Both sensors GPS receiver and the inertial measurement unit have to meet secial requirements. First the internal code and hase tracing rocess of the receiver must be directly modified by the system develoer. herefore the imlementation of the control loos have to be changeable. It must be ossible to read out data from other sources and add them to the tracing loos. For this uroses we use the GPS Architect receiver card from MIE. It is a develoment system intended for GPS receiver design oerating in conjunction with a PC. he receiver software is written in C, can modified on a PC and then reloaded to the RISC rocessor of the receiver card. A bloc diagram of the GPS with all elementary comonents (RF front end, correlator, RISC rocessor, memory) and interfaces is shown in figure 3.

10 Figure 3. card Diagram of the GPS Architect receiver As we investigated in the theoretical investigations a gyro drift of the nearly /h is necessary to ee the hase tracing loos of the GNSS receiver in loc. his accuracy can be rovided by a fiber otic gyro. Additionally the integration of GNSS and for dynamic alications requires an exact timing synchronization of the sensors. herefore the ossibility of external synchronization of the measurement frequency is an imortant roerty for this alication. For this reasons we use the N- Inertial Measurement Unit from ION. It rovides the inertial data by a RS 485 data bus using the SDC rotocol with an outut rate of 4 Hz. In order to integrate the GPS and data they have to be synchronized to a common time table herefore a time information must be added to every data set. herefore a timing controller is used which is synchronized to GPS time by a PPS ulse. First of all a generated Hz cloc controls the data samling rocess. Additionally a Hertz ulse resets the internal time tag counter of the. hese counter values are one element of every data set. Finally the internal latching of data is signalized by a digital ulse which can be temorally fixed by the timer card. hereby the time synchronization can be controlled. PC Position,velocity RS 3 RS 3 Inertial data Figure 4. GPS 485 Predicted range-rate Stradown algorithm filtering bias estimation Caculation of range-rates rediction PPS ime imer card RS 485 Satellite orbits Inertial data Synchronisation signal (-RS 485) Mitel GPS ARM (RISC) Position, velocity, attitude, time itton N- Data flow of the tightly couled system In figure 4 an overview about the basic comonents and the data flow of the tightly couled ractical test system is resented lie it is lanned today. he central comonent of the test system is a software acage running on a PC. he following tass must be erformed: ransformation of inertial data into navigation data using a stradown algorithm ime synchronisation of GPS and data Estimation of errors using integrated osition and velocity as observations Calculation of seudorange, velocity and acceleration out of navigation data and satellite orbits Prediction of data at GPS tracing time In order to decrease the comlexity of the integrated system some simlifications are used. As the MIE GPS Architect does not rovide the PPS signal a second ordinary GPS card is used transferring the satellite orbit information at the same time. As a second simlification the estimation of the errors uses osition and velocity as observations instead of seudoranges. herewith the rincile of tightly couling is not exactly be et. However concerning to the evaluation of imrovement caabilities and difficulties correct results can be exceted using the resented system configuration OUOOK On condition that the lanned theoretical and ractical investigations of the tightly couled system show the unredictable difficulties an increase of GNSS erformance is ossible. he basis for a develoment of a -aided GNSS receiver might be established. Including a rice decrease of accurate sensors caused by further develoments of the MEMS technology the use of tightly couled GNSS sensors could become a standard for dynamical alications. It might be exected that the integration of GNSS and systems become more closer in the future. But this only seems to be ossible within the scoe of a tight cooeration with a manufacturer of signal rocessing units of GNSS receivers. ACKNOWEDGEMEN he investigations and develoments of a tightly couled GNSS/ sensor are founded by DR (Deutsches Zentrum fuer uft und Raumfahrt e. V.), FKZ 5 NA 99. REFERENCES Britting, Kenneth R. (97) Inertial Navigation Systems Analysis, Wiley-Interscience

11 DiEsosti, Raymond et al.(998) he Benefits of Integrating GPS, and PCS; Proceedings ION-GPS 998 Eissfeller, Bernd (997) Ein dynamisches Fehlermodell für GPS Autoorrelationsemfänger; Schriftenreihe Studiengang Vermessungswesen, Universität der Bundeswehr München, Heft 55 Eissfeller, Bernd (989) Analyse einer geodätischen raumstabilisierten Inertiallatform und Integration mit GPS, Schriftenreihe Studiengang Vermessungswesen, Universität der Bundeswehr München, Heft 37 Kalan, Elliott D. (996) Understanding GPS- Princiles and Alications, Mobile Communications Series; Artech House, Boston, ondon andry, Rene Jr. (999) New echnique to Imrove GPS Receiver Performance by Acquisition and racing hreshold Reduction; Research and echnology Organisation (RO), Meeting Proceedings 43 (International Conference on Integrated Navigation Systems) MIE (997) GPS Architect Channel GPS Develoment System; version DS March 997, Senott, Jim et. al.(997) Robustness of ightly Couled Integrations for Real ime Centimeter GPS Positioning; Proceedings ION-GPS 997 itterton, D.H. and Weston, J..(997) Stradown Inertial Navigation echnology; IEE Radar, Sonar, Navigation and Avionics Series, he avenham Press Ward, Ph. (998) Performance Comarisons between F, P and a Novel F-Assisted P Carrier racing oo under RF Interference Conditions; Proceedings ION-GPS 998, Winel, Jon O. et al (998) SNSS Simulated Navigation Satellite System; Simulating a Generic GNSS Receiver in Virtual Environments; Proceedings of ION-GPS 998