ARAIM: Utilization of Modernized GNSS for Aircraft-Based Navigation Integrity Alexandru (Ene) Spletter Deutsches Zentrum für Luft- und Raumfahrt (DLR), e.v. The author gratefully acknowledges the support of the FAA Program Office during his PhD work at
Big Issues in Today s Aviation Capacity Air traffic congestion in US and Europe (increased demand for air travel) Efficiency Rising oil prices (increased demand and scarcity) The push for green not just operating existing system more efficiently, but creating new propulsion, operations, manufacturing concepts Safety Ever-increasing (security) standards standards Introduction Motivation Background Contributions Model 2
Advantages of Satellite Navigation Ability to define an optimal flight path Reduced amount of airspace needed by each airport Fuel savings and reduced environmental footprint Eliminating dependence on ground navaids Ground equipment costs (e.g. real estate, equipment installation and maintenance) are skyrocketing with the increase in capacity Introduction Motivation Background Contributions Model 3
Outline Introduction & Background Goal: Optimize navigation performance with an unaided global navigation satellite constellation for aviation vertical guidance Advanced RAIM Algorithm Development Model and Method Validation Summary & Conclusions Introduction Motivation Background Contributions Model 4
Motivation o For precision airplane navigation with current singlefrequency GPS, expensive ground + space equipment is needed for augmenting the satellite positioning signals o Aircraft-Based Augmentation Systems (ABAS) combined with GPS alone are currently not suitable for aviation precision approaches o As modernized GPS (L1-L2-L5 civilian signals), Galileo and additional GNSS constellations become fully operational, the positioning integrity workload can be redistributed more efficiently between the user, space and ground segments of the navigation system Introduction Motivation Background Contributions Model 5
Photo courtesy: Traian Vuia Timisoara International Airport, Romania Background Global Navigation Satellite Systems (GNSS) Ground and Aircraft Based Augmentation Systems (GBAS and ABAS) Autonomous Navigation and Civil Aviation Safety 6 Introduction Motivation Background Contributions Model
Satellite Distance Measurements Each satellite provides two types of range measurements: Code phase measurements» Unambiguous range, 1m resolution Carrier phase measurements» Ambiguous range, 1cm resolution GNSS satellites 7 Introduction Motivation Background Contributions Model
Global Positioning System (GPS) Oops! GPS User GPS User Runway not to scale Introduction Motivation Background Contributions Model 8
Range Measurement Errors Provides two range measurements Code phase measurements» Unambiguous range, 1 m resolution Carrier phase measurements» Ambiguous range, 1 cm resolution GPS satellites Satellite clock & ephemeris error User Range Errors Satellite clock (2 m) Ephemeris/Orbit (2 m) Ionospheric delay Ionosphere (1~5 m)» Can be removed by a dual frequency GPS receiver Troposphere (0.1~1 m) Multipath and Receiver Noise (0.5~1m for code and 0.5~1cm for carrier) Tropospheric delay Receiver noise Ground Multipath Background
Differential GPS Advantage: Accuracy DGPS User DGPS User GPS Reference stations not to scale Introduction Motivation Background Contributions Model 10
GPS Augmentation System Advantage: Integrity Providing users with error bounds that guarantee safe navigation limits OR a timely warning otherwise. Alert Limit Error Bound Runway not to scale Introduction Motivation Background Contributions Model 11
Aviation Navigation Requirements 35m VAL equivalent: LPV 200 GLS not to scale Introduction Motivation Background Contributions Model 12
Global Navigation Satellite Systems Present: (regional) GNSS Augmentation Systems Ground-Based Augmentation System (GBAS)» SCAT-I (Norway), LAAS (U.S.), GRAS (Australia) Space-Based Augmentation System (SBAS)» WAAS (U.S.), EGNOS (E.U.), MSAS (Japan), GAGAN (India) [A worldwide SBAS would have difficulty meeting the 6 s Time-To-Alert] Future: (worldwide) GNSS will include Modernized GPS (U.S.) Glonass (Russia), Galileo (E.U.) Beidou (China) not to scale ( broadcasting in multiple frequency bands ) Introduction Motivation Background Contributions Model 13
RAIM - description and background - Receiver Autonomous Integrity Monitoring uses measurements from redundant satellites in order to protect the user against large positioning errors This research explores a comprehensive threat space, including multiple simultaneous faults, constellation failures, and measurement biases Introduction Motivation Background Contributions Model 14
Potential Satellite Fault GNSS Satellites Protection Level (PL) VPL User not to scale HPL Runway Introduction Motivation Background Contributions Model
Research Goal Optimize the performance of an unaided multifrequency GNSS constellation from a vertical integrity standpoint for aviation precision approaches Introduction Motivation Background Contributions Model 16
Overview of RAIM Methods Sequential Algorithms (Multiple Epoch) Nikiforov, I.V., New Optimal Approach to Global Positioning System/Differential Global Positioning System Integrity Monitoring, Journal of Guidance, Control, and Dynamics, v.19, no.5, 1996. [Measurements are correlated on a time-scale longer than the Time-To-Alert] Snapshot Algorithms (Single Epoch) Classical Least-Squares RAIM Brown, R.G., A Baseline GPS RAIM Scheme and a Note on the Equivalence of Three RAIM Methods, Navigation, v.39, no.3, 1992. Solution Separation RAIM Brown, R.G., McBurney, P., Self-Contained GPS Integrity Check Using Maximum Solution Separation, Navigation, v.35, no.1, 1988. Generalized Likelihood Ratio (Multiple Hypothesis) Ober, P.B., Integrity Prediction and Monitoring of Navigation Systems, Integricom Publishers, 2003. Other derived flavors of RAIM focused on improving fault detection, position error or navigation availability. Introduction Motivation Background Contributions Model 17
Overview of RAIM Methods Snapshot Algorithms Classical Least-Squares RAIM Brown, R. G., A Baseline GPS RAIM Scheme and a Note on the Equivalence of Three RAIM Methods, Navigation, v.39, no.3, 1992. Solution Separation RAIM Brown, R. G., McBurney, P., Self-Contained GPS Integrity Check Using Maximum Solution Separation, Navigation, v.35, no.1, 1988. Generalized Likelihood Ratio (Multiple Hypothesis) Single-Constellation / Single Fault Ober, P.B., Integrity Prediction and Monitoring of Navigation Systems, Pervan, Integricom B., Pullen, Publishers, S. and Christie, 2003. J., A Multiple Hypothesis Approach to Satellite Navigation Integrity, Navigation, v.45, no.1, 1998. Other derived flavors of RAIM focused on improving fault detection, position error or navigation availability. Multiple-Constellation / Multiple Fault Emerging class of algorithms called Advanced RAIM (ARAIM) Introduction Motivation Background Contributions Model 18
Contributions I. ARAIM Algorithm Development Developed the capability of handling multiple faults, a well-defined comprehensive threat model, VPL prediction capabilities and proposed an original Fault Detection and Exclusion (FDE) method well adapted to the MHSS RAIM algorithm. II. Determination of Parameter Space (over which ARAIM results meet aviation performance requirements) Explored new issues and opportunities that will come with a multiple constellation GNSS, identifying the key parameters that affect the overall performance: total number of available satellites, the bias and variance of ranging errors under conditions assumed to be nominal. III. Validation of the Error Model Using the type of dual frequency measurements currently available (GPS L1/L2 semicodeless), developed a method to compute a position solution and showed how theoretical predictions can be adapted to model errors at a new frequency. IV. Verification of the Algorithm with Real Data Based on data recorded during flight inspection, made available by the FAA Tech Center, validated the proposed FDE method, testing it against real measurements, as well as artificially injected satellite failures. Introduction Motivation Background Contributions Model 19
Contribution I Algorithm Development Multiple Hypothesis Solution Separation (MHSS) RAIM Algorithm» Real-Time VPL» Predicted VPL Fault Detection and Elimination (FDE) Method 20 Introduction Motivation Background Contributions Model
Modeling Range Errors Nominal error distribution model consists of zero-mean noise (allowing a Gaussian overbound) and biases in each channel. Abnormal errors, which are present in faulty range measurements will be modeled as fault biases. i i b i f i Introduction Motivation Background Contributions Model 21
Vertical Protection Level (VPL) VPL = Gaussian term + Bias overbound Introduction Motivation Background Contributions Model 22
Real-Time VPL Interval Integrity Risk: over all fault modes, P(HMI) 10-7 per approach VPE > VPL VPE VPL HMI VPL Introduction Motivation Background Contributions Model 23
Multiple Hypothesis Algorithm Assign prior probabilities to Fault Modes Allocate the integrity risk between Modes Take union of subset VPL intervals Hypotheses: { H 0, H 1, H 2,, H j, } Mode 0 causes Integrity Fault Mode 1 causes Integrity Fault Mode 2 causes Integrity Fault Mode j causes Integrity Fault V V V V V + + + + + + P(HMI 0 ) P(HMI 1 ) P(HMI 2 ) P(HMI j ) Fault Tree Equation: Total Integrity Risk P( HMI ) P( HMI ) P j prior j 0 Introduction Motivation Background Contributions Model m P(HMI j ) H ( H j ) 24
Solution Separation VPL For each Mode, compute a position solution and error bound for the SUBSET that is not affected by the fault ( j) VPL K ( j) ( j) ( Palloc) v n i 1 ( j) S 3, i B max, j ( j) v (0) VPL ( j) VPL Vertical direction! z 0 ( j) v z j z VPL m max j 0 ( j ) VPL Introduction Motivation Background Contributions Model 25
I. Algorithm Development Multiple Hypothesis Solution Separation (MHSS) RAIM Algorithm Real-Time VPL Predicted VPL Fault Detection and Elimination (FDE) Method Introduction Motivation Background Contributions Model 26
VPL Prediction Equation SUBSET VPL + SOLUTION SEPARATION = = GAUSSIAN TERM GAUSSIAN TERM + + BIAS TERM BIAS TERM Predicted VPL n ( j) ( j) ( j) ( j) ( j) ( j) VPL K( Palloc) v S3, i Bmax, j SS bss i 1 Introduction Motivation Background Contributions Model 27
VPL Prediction Algorithm Since the VPL is a direct function of the position measurement residuals, a tool has been developed for predicting VPL values ahead of time, when a critical navigation operation is about to begin The Prediction VPL is expected to be an upper bound on the Real-Time VPL, given the satellite geometry and model parameters Introduction Motivation Background Contributions Model 28
I. Algorithm Development Multiple Hypothesis Solution Separation (MHSS) RAIM Algorithm Real-Time VPL Predicted VPL Fault Detection and Elimination (FDE) Method Introduction Motivation Background Contributions Model 29
MHSS FDE Algorithm Compute the VPL for the given all-in-view configuration as usual Also compute VPLs after eliminating up to k satellites for all possible such partial configurations Choose the minimum of all these possible VPLs If this minimum VPL is smaller than the original all-in-view VPL, then a fault has been recorded on the satellites eliminated from the corresponding subset Introduction Motivation Background Contributions Model 30
Simulation with MHSS Simulation Parameters User Grid 5 deg lat 10 deg long Constellation Size (optimized orbit spacing) 24 GPS satellites 30 Galileo satellites Elevation Mask Angle 5 deg (GPS) 10 deg (Galileo) Period 150 sec/ 24 hours approach total Satellite: Constellation: A Priori Fault 1 x 10-4 / 1 x 10-7 / Probability approach approach At each user location, the 99.5% largest VPL over the simulation period is mapped, and the maps are then colored by interpolation between grid points availability Introduction Motivation Background Contributions Model 31
URA = 1m, Bias = 0m, Failure priors of 10-4 / 10-7 80 World VPL Maps 60 40 Galileo GPS + only Galileo (24 (30 active (54 satellites) Latitude (deg) 20 0-20 The VPL value is a function of user location, time of day and the current constellation configuration. -40-60 -80-150 -100-50 0 50 100 150 Longitude (deg) Latitude (deg) 80 60 40 20 0-20 URA = 1m, Bias = 0m, Failure priors of 10-4 / 10-7 Latitude (deg) 80 60 40 20 0-20 < 7 < 10 < 15 < 20 < 25 < 30 < 35 < 50 > 50 99.5% VPL (m) 4863.03 m avg. URA = 1m, Bias = 0m, Failure priors of 10-4 / 10-7 -40-40 -60-60 -80-80 -150-100 -50 0 50 100 150 Longitude (deg) -150-100 -50 0 50 100 150 Longitude (deg) < 7 < 10 < 15 < 20 < 25 < 30 < 35 < 50 > 50 99.5% VPL (m) 10.54 m avg. < 7 < 10 < 15 < 20 < 25 < 30 < 35 < 50 > 50 99.5% VPL (m) 2784.19 m avg. Multiple constellations are necessary to ensure desired performance. Introduction Motivation Background Contributions Model 32
Real-Time and Predicted VPL Real Time VPL URA = 1m, Bias = 0m, Failure priors of 10-4 / 10-7 Predicted VPL URA = 1m, Bias = 0m, Failure priors of 10-4 / 10-7 80 80 60 60 40 40 Latitude (deg) 20 0-20 Latitude (deg) 20 0-20 -40-40 -60-60 -80-80 -150-100 -50 0 50 100 150 Longitude (deg) -150-100 -50 0 50 100 150 Longitude (deg) < 7 < 10 < 15 < 20 < 25 < 30 < 35 < 50 > 50 99.5% VPL (m) 10.81 m avg. < 7 < 10 < 15 < 20 < 25 < 30 < 35 < 50 > 50 99.5% VPL (m) 17.48 m avg. Introduction Motivation Background Contributions Model 33
Fault Detection and Elimination (FDE) (10m Fault Bias on Most Critical SV) Satellite Failure Present Faulty Satellite Removed URA = 1m, Bias = 0m, Failure priors of 10-4 / 10-7 URA = 1m, Bias = 0m, Failure priors of 10-4 / 10-7 80 80 60 60 40 40 Latitude (deg) 20 0-20 Latitude (deg) 20 0-20 -40-40 -60-60 -80-80 -150-100 -50 0 50 100 150 Longitude (deg) -150-100 -50 0 50 100 150 Longitude (deg) < 7 < 10 < 15 < 20 < 25 < 30 < 35 < 50 > 50 99.5% VPL (m) 17.36 m avg. < 7 < 10 < 15 < 20 < 25 < 30 < 35 < 50 > 50 99.5% VPL (m) 12.88 m avg. 34 Introduction Motivation Background Contributions Model
Contribution I Algorithm Development Multiple Hypothesis Solution Separation (MHSS) RAIM Algorithm» Real-Time VPL» Predicted VPL Fault Detection and Elimination (FDE) Method Developed: o the capability of handling multiple faults; o a well-defined comprehensive threat model; o VPL prediction capabilities, guaranteeing a performance level even in the absence of actual measurements. Proposed an original Fault Detection and Exclusion (FDE) method that minimizes the VPL values. Introduction Motivation Background Contributions Model 35
Contributions I. ARAIM Algorithm Development Developed the capability of handling multiple faults, a well-defined comprehensive threat model, VPL prediction capabilities and proposed an original Fault Detection and Exclusion (FDE) method well adapted to the MHSS RAIM algorithm. II. Determination of Parameter Space for Initial Assumptions (over which desired ARAIM performance is achievable) Explored new issues and opportunities that will come with a multiple constellation GNSS, identifying the key parameters that affect the overall performance: total number of available satellites, the bias and variance of ranging errors under conditions assumed to be nominal. III. Validation of the Range Error Model Using the type of dual frequency measurements currently available (GPS L1/L2 semicodeless), developed a method to compute a position solution and showed how theoretical predictions can be adapted to model errors at a new frequency. IV. Verification of the Algorithm with Real Data Based on data recorded during flight inspection, made available by the FAA Tech Center, validated the proposed FDE method, testing it against real measurements, as well as artificially injected satellite failures. Introduction Motivation Background Contributions Model 36
User Range Accuracy (URA) URA characterizes Gaussian errors due to satellite clock and orbit uncertainties VPL as a function of user location VPL as a function of user location 80 80 60 60 40 40 Latitude (deg) 20 0-20 Latitude (deg) 20 0-20 -40-40 -60-60 -80-80 -150-100 -50 0 50 100 150 Longitude (deg) -150-100 -50 0 50 100 150 Longitude (deg) < 7 < 10 < 15 < 20 < 25 < 30 < 35 < 50 > 50 VPL (m) with MASK = 5deg, URA = 1m, 99.5% < 7 < 10 < 15 < 20 < 25 < 30 < 35 < 50 > 50 VPL (m) with MASK = 5deg, URA = 2m, 99.5% Failure priors: 10-4 /satellite and 10-7 /constellation per each approach Introduction Motivation Background Contributions Model 37
Parametric Studies - Ranging Errors Two important user-independent parameters are the satellite ranging accuracy and the measurement bias. Vertical Protection Level (m) 100 90 80 70 60 50 40 30 20 10 Parametric study for ranging accuracy min 1% avg 99% max Vertical Protection Level (m) 100 90 80 70 60 50 40 30 20 10 min 1% avg 99% max Parametric study of biases 0 0 1 2 3 4 User Ranging Accuracy (m) 0 0 1 2 3 4 5 7 10 Max nominal pseudorange bias (m) 38 Introduction Motivation Background Contributions Model
Parametric Studies Fault Probability Another important user-independent parameter is the a priori measurement fault probability. 30 25 20 Parametric study for SV failure priors min 1% avg 99% max VPL (m) 15 10 5 0 10-7 10-6 10-5 10-4 10-3 10-2 SV failure prior Introduction Motivation Background Contributions Model 39
Contribution II Determination of Parameter Space for Initial Assumptions (over which desired ARAIM performance is achievable) Explored new issues and opportunities that will come with a multiple constellation GNSS, identifying the key parameters that affect the overall performance: total number of available satellites, the variance and bias of ranging errors under conditions assumed to be nominal. o White Noise error term: URA < 1 m o Bias error component: max Bias < 2 m o Satellite Fault Prior: Prob < 10-3 / approach Introduction Motivation Background Contributions Model 40
Contributions I. ARAIM Algorithm Development Developed the capability of handling multiple faults, a well-defined comprehensive threat model, VPL prediction capabilities and proposed an original Fault Detection and Exclusion (FDE) method well adapted to the MHSS RAIM algorithm. II. Determination of Parameter Space for Initial Assumptions (over which desired ARAIM performance is achievable) Explored new issues and opportunities that will come with a multiple constellation GNSS, identifying the key parameters that affect the overall performance: total number of available satellites, the bias and variance of ranging errors under conditions assumed to be nominal. III. Validation of the Range Error Model Using the type of dual frequency measurements currently available (GPS L1/L2 semicodeless), developed a method to compute a position solution and showed how theoretical predictions can be adapted to model errors at a new frequency. IV. Verification of the Algorithm with Real Data Based on data recorded during flight inspection, made available by the FAA Tech Center, validated the proposed FDE method, testing it against real measurements, as well as artificially injected satellite failures. Introduction Motivation Background Contributions Model 41
Modeling Range Errors Nominal error distribution model consists of zero-mean noise (allowing a Gaussian overbound) and biases in each channel. Abnormal errors, which are present in faulty range measurements will be modeled as fault biases. i i b Estimate measurement noise ε variance for satellite i as: σ i2 = URA 2 + σ i,tropo2 + σ i,iono-free 2 i ( Error model agreed upon in the FAA GEAS workgroup in 2007 ) f i Introduction Motivation Background Contributions Model 42
Raw Measurements Cycle Slip Detection Carrier Smoothing Model Validation Scope: Validate a range measurement error model for dual frequency measurements with currently available GPS L1-L2 data. Data Collection Setup: An Ashtech Z-Extreme and a NovAtel OEM4 receivers were connected to the aircraft antenna of the FAA Boeing 727 during LAAS Flight Tests at Memphis, Tennessee on September 19-20 2006. Apply Corrections Determine Range Errors Validate Modeled Sigmas Introduction Motivation Background Contributions Model VPL Prediction 43
Range Residuals Statistics Standard Deviation (m) 3.5 3 2.5 2 1.5 1 0.5 Range Error sigmas by Elevation max L1 error L2 error L2/L1 ratio 0 0 10 20 30 40 50 60 70 80 90 degrees Standard Deviations for residual ranging errors. Error distributions are illustrated for each of the 5 elevation bins. Introduction Motivation Background Contributions Model 44
Error Model Calibration 3 2.5 Modeled and Measured Range Errors Ashtech Novatel Theoretical L2 Calibrated Standard Deviation (m) 2 1.5 1 0.5 0 0 10 20 30 40 50 60 70 80 90 Elevation (deg) Introduction Motivation Background Contributions Model 45
Contribution III Error Model Validation Using the only type of dual frequency measurements currently available (GPS L1/L2 semicodeless), developed a method to compute a position solution and showed how a theoretical prediction can be adapted to model errors at a new frequency. Validated model has to be conservative in overbounding errors, as it currently covers a broad array of situations, from flight at altitude to near landings at any possible airport. Introduction Motivation Background Contributions Model 46
Contributions I. ARAIM Algorithm Development Developed the capability of handling multiple faults, a well-defined comprehensive threat model, VPL prediction capabilities and proposed an original Fault Detection and Exclusion (FDE) method well adapted to the MHSS RAIM algorithm. II. Determination of Parameter Space for Initial Assumptions (over which desired ARAIM performance is achievable) Explored new issues and opportunities that will come with a multiple constellation GNSS, identifying the key parameters that affect the overall performance: total number of available satellites, the bias and variance of ranging errors under conditions assumed to be nominal. III. Validation of the Range Error Model Using the type of dual frequency measurements currently available (GPS L1/L2 semicodeless), developed a method to compute a position solution and showed how theoretical predictions can be adapted to model errors at a new frequency. IV. Verification of the Algorithm with Real Data Based on data recorded during flight inspection, made available by the FAA Tech Center, validated the proposed FDE method, testing it against real measurements, as well as artificially injected satellite failures. Introduction Motivation Background Contributions Model 47
Before Fault Insertion Over all in-flight measurement sets: Integrity: No HMI event: VPE VPL 100% of the time! Availability: (VPL 35m) Predicted VPL 81.5% Real-Time VPL 85.4% Introduction Motivation Background Contributions Model 48
SV Position Error (norm) vs. Time 700 courtesy: Prof. Boris Pervan, Illinois Institute of Technology 600 SV Unhealthy 500 Error(meters) 400 300 200 100 NANU Scheduled Outage Time 0 12 14 16 18 20 22 0 2 Time (starting atnoon April10th) 49
Simulated SV Failures In order to test the capabilities of the MHSS FDE RAIM algorithm, ramp range errors were inserted into the recorded flight inspection data. Successive tests were performed, with the slope of the error gradually increasing from ±1cm/s to ±1m/s. Larger errors are more rapidly identified. As error detection is performed solely based on the snapshot of current measurements, the response to either step or ramp errors will be identical. Introduction Motivation Background Contributions Model 50
Insertion of Ramp Error (Ashtech PRN 16) Measured Error Inserted Fault 400 Residual Range Error 400 Residual Range Error 350 350 300 250 300 250 200 150 100 50 0-50 0 1000 2000 3000 4000 5000 6000 7000 8000 seconds meters 200 150 100 50 0-50 0 1000 2000 3000 4000 5000 6000 7000 8000 seconds A satellite clock runoff was simulated on PRN16, a well-visible satellite throughout the entire set of observations. Introduction Motivation Background Contributions Model 51
Fault Detection and Elimination Undetected Fault Detection Process Real-Time VPL and Error from True Position FDE VPL Interval and Position Error meters 1500 1000 500 All-in-view VPE All-in-view VPL 100 75 50 35 Healthy subset VPE Healthy subset VPL VPE : VPL ratio 0 0 2000 4000 6000 8000 10000 1 0.5 Error/ VPL Ratio before Elimination 0 0 2000 4000 6000 8000 10000 seconds 0 0 2000 4000 6000 8000 10000 0 0 2000 4000 6000 8000 10000 Introduction Motivation Background Contributions Model 1 0.5 Error/ VPL Ratio after Elimination seconds 52
Worst Observed Performance with Large Ranging Errors (i.e. Faults) Number of Satellites Position Solution All-in-view VPL 1 Fault Eliminated 2 Faults Eliminated N 3 N = 4 N = 5 HIGH N = 6 100m HIGH N = 7 50m 100m N/A N = 8 35m 50m N/A N = 9, 10 25m 15-50m N/A N 11 5-15m 10-35m 15-50m Even having 7-8 satellites in the position solution is not sufficient to provide availability under all geometries in the presence of faulted measurements. Introduction Motivation Background Contributions Model 53
After Fault Insertion Measurement sets contain an inserted fault about 90% of the time! Integrity: HMI occurs whenever VPE > VPL All-in-view Real-Time VPL: VPE VPL 100% of the time! MHSS RAIM - FDE Algorithm: VPE VPL 100% of the time! Availability: (VPL 35m) 21.9% under the no fault assumption, which increases to 81.5% after FDE is applied. Introduction Motivation Background Contributions Model 54
Summary A significant improvement in autonomous integrity monitoring will be available for satellite navigation once multiple modernized GNSS constellations become operational. The MHSS algorithm with FDE capabilities would enable LPV-200 landings at all runway ends in the world without the need for a satellite-based or ground-based augmentation system (SBAS or GBAS). Fault Detection can be employed both to eliminate large erroneous measurement biases and also to improve the performance under nominal conditions. Introduction Motivation Background Contributions Model 55
Contributions 1. ARAIM Algorithm Development Developed the capability of handling multiple faults, a well-defined comprehensive threat model, VPL prediction capabilities and proposed an original Fault Detection and Exclusion (FDE) method well adapted to the MHSS RAIM algorithm. 2. Determination of Parameter Space (over which ARAIM results meet aviation performance requirements) Explored new issues and opportunities that will come with a multiple constellation GNSS, identifying the key parameters that affect the overall performance: total number of available satellites, the bias and variance of ranging errors under conditions assumed to be nominal. 3. Validation of the Error Model Using the type of dual frequency measurements currently available (GPS L1/L2 semicodeless), developed a method to compute a position solution and showed how theoretical predictions can be adapted to model errors at a new frequency. 4. Verification of the MHSS Algorithm with Real Data Based on data recorded during flight inspection, made available by the FAA Tech Center, validate the proposed FDE method, testing it against real and artificially injected satellite failures. Introduction Motivation Background Contributions Model 56
Publications 1) Alexandru Ene, Juan Blanch, Todd Walter, J. David Powell Validation of Multiple Hypothesis RAIM Algorithm Using Dual-frequency GNSS Signals Presented at the Toulouse Space Show/ ENC-GNSS 08, Toulouse, France, 22-25 April 2008. 2) Georg Schroth, Alexandru Ene, Juan Blanch, Todd Walter, Per Enge Failure Detection and Exclusion via Range Consensus Presented at the Toulouse Space Show/ ENC-GNSS 08, Toulouse, France, 22-25 April 2008. 3) Juan Blanch, Alexandru Ene, Todd Walter, Per Enge An Optimized Multiple Hypothesis RAIM Algorithm for Vertical Guidance Presented at ION GNSS 07, Ft. Worth, TX, 25-28 September 2007. 4) Alexandru Ene Multiple Hypothesis RAIM with Real-Time FDE and Forecasted Availability for Combined Galileo-GPS Vertical Guidance Presented at ENC-GNSS 07, Geneva, Switzerland, 28-31 May 2007. 5) Alexandru Ene, Juan Blanch, J. David Powell Fault Detection and Elimination for Galileo-GPS Vertical Guidance Presented at ION NTM 07, San Diego, CA, 22-24 January 2007. 6) Alexandru Ene Further Development of Galileo-GPS RAIM for Vertical Guidance Presented at ION GNSS 06, Ft. Worth, TX, 26-29 September 2006. 7) Alexandru Ene, Juan Blanch, Todd Walter GPS-Galileo RAIM for Vertical Guidance Presented at ION NTM 06, Monterey, CA, 18-20 January 2006. 8) Alexandru Ene, Di Qiu, Ming Luo, Samuel Pullen, and Per Enge A Comprehensive Ionosphere Storm Data Analysis Method to Support LAAS Threat Model Development Presented at ION NTM 05, San Diego, CA, 24-26 January 2005. 9) Ming Luo, Samuel Pullen, Alexandru Ene, Di Qiu, Todd Walter, and Per Enge Ionosphere Threat to LAAS: Updated Model, User Impact, and Mitigations Presented at ION GNSS 04, Long Beach, CA, 21-24 September 2004. 10) Dennis Akos, Alexandru Ene, and Jonas Thor A Prototyping Platform for Multi-Frequency GNSS Receivers Presented at ION GPS/GNSS 03, Portland, Oregon, 9-12 September 2003. Introduction Motivation Background Contributions Model 57
Future Work Continue validation efforts for error model and algorithm with larger amounts of data Implement alternative ways to screen satellite geometries to minimize VPE in addition to VPL Integrate ARAIM algorithm with an actual GNSS receiver in real time, and validate with (real or simulated) radio signals from multiple constellations Conduct flight tests to validate Real-Time algorithm against RTCA-229D aviation certification requirements (for both vertical and lateral guidance) Introduction Motivation Background Contributions Model 58
LAX Sunset Airplane Landing by Flip Barrientos
- APPENDIX - The Use of GNSS for Safe Airplane Navigation Integrity, Continuity and Availability Studies September 3, 2008 The author gratefully acknowledges the support of the FAA Program Office.
Aviation Navigation Requirements» Accuracy: 0.95 (95%) position error bound» Integrity: 0.9999999 (1-10 -7 ) position error bound = PL» Availability: 0.995 (99.5%) of the time require PL AL» Continuity: 0.999996 (1-4x10-6 ) of the time availability should not be lost while the approach is in progress Integrity: Probability of vertical error larger than 35 m without annunciation within 6 seconds must be less than 10-7 /approach Continuity: Probability of loss of service must be less than 4x10-6 per 15 seconds 61 Introduction Motivation Background Contributions Model
Aviation Navigation Requirements Introduction Motivation Background Contributions Model 62
Continuity Equation Continuity Term = Kcont*sigma + Bcont*rss(S3,3) Pr(CONT) > (1-4*10-6 ) / approach Loss of Continuity Continuous Operation Introduction Motivation Background Contributions Model 64