Tgp chi Khoa hoc v^ Cong ngh| 102 (2014) 064-068 Evaluation of Advanced RAIM Algorithm for GPS and Galileo Signals Integrity Monitoring Danh gia thuat toan ARAM dl giam sat tinh toan ven cua tin hieu GPS va Galileo Hieu Trung TRAN'*, Tung Hai TA', Letizia LO PRESTf ' NAVIS Centre Hanoi University of Science and Technology, No Dai Co Viet, Hanoi, Viemam ^ Politecnico di Torino, Corso Duca degUAbruzzi 24, Torino, Italy Received: April 01, 2014, accepted: August 25, 2014 Abstract Receiver Autonomous integrity Monitoring (FiAIM) is a navigation integrity assessment technology in aviation non-precision appnDach phases of flights With recent development of GNSS, the interest in further developi RAIM for more demanding operations has been increased The Advanced RAIM algorithm pnDposed by GEA one of the new FiAIM algonthms, utilizing the multi-GNSS environment to improve the performance of RAIM this paper, the ARAIM algonthm is implemented with real data input, containing both GPS and new Galileo £ signal Implementation results of ARAIM are also compared with that of traditional FiAIM, showing tlie improvement of ARAIM and possibility of real-life implementation Keyword, GNSS, integrity, mulii-GNSS, ARAIM, Galileo, GPS Tom tat Thu$t toan RAIM dwac sir dung nganh hang khdng di xac dmh tinh toan ven thdng tin dinh vj cac q trinh bay khdng ddi hoi chinh x^c cao NhOrig phat triin gin ddy cua cdng nghe dinh vi sir di^ng v§ tmh dS m nhiiu hirdng nghien ciru mdi cho FiAIM, md ring img dijng cua cdng nghe cho nhOng hoat d0ng yeu c^ dd chinh x^c cao han Thuat to^n RAIM Cii tiin (AFiAIM) di xuit bdi GEAS la thuat toan FiAIM thi he mdi, sd dijng tin hi$u dinh vi da h§ thing di cai thi$n hieu nang cua RAIM Trong bdi bao nay, thuat todn AFiAIM dirQz then khat su" dung tin hieu thwc, gom tin hieu GPSva Galileo El Kit qua thwc nghidm cua ARAIM cho thiy s cai thien Idn vi hieu n§ng ddi vdi FiAIM va kha nang triin khai thuc ti Tu khoa' GNSS, tinh toan ven thong tin, dinh vj da he thong, ARAIM, Galileo, GPS iDtroductioQ Receiver Autonomous Integrity Monitoring (RAIM) algonthm was developed m the early year of GPS-based navigation to ensure integrity for en route through non-precision approach phases of flight [1] Recent development of GNSS, such as the upcoming of Galileo, Beidou as well as the new signals for GPS, led to the increasing interest in further developing RAIM for higher demanding phases of flights, most notably LPV-200 [2] LPV-200 is a Localizer Performance with Vertical guidance (LPV) approach procedure, m which guidance is provided down to a minimum decision height of 200 ft,, or 60.96 m It requires the Vertical Alarm Limit (VAL) of 35 m: any vertical position error larger than VAL must be alerted to the pilot within 6s, The Advanced RAIM algorithms are the new generation of RAIM, utilizing the new multi-GNSS environment to provide better perfiirmance and availability for RAIM, aiming at providing LPV-200 availability world wide ' Corresponding Author: Tel +84-1234 138 476 Email: hieu.ttantrung-navis@hust edu.vn In this paper, the ARAIM algorithm proposed by GEAS [6] will be studied and discussed Being an extension of the Solution Separation (SS) method [5j, the algorithm proposes new threat model to provide better availability for Protection Levels, The algonthm is also implemented, using input of real data including GPS and multiple new Galileo satellites to test the performance and availability of the algorithm The paper is organized as follow Section provides an overview of the ARAIM algorithm Section describes the process of the algorithm, step by step Section presents and analyses the implementation results and compare them with those of traditional RAIM algorithm The paper concludes with a summary on the topic ARAIM overview RAIM algorithms can be divided into two groups, based on the input for the consistency check The first group performs the check on the redundancy ranging measurement This group is represented by the traditional Chi-square method by R, G Brown [3] Recent work in this group is the notable new RAJM method developed by Young Lee [4], considering the T^p chi Khoa hifc va Cong nghe 102 (2014) 064-068 multiple-fault wide faults assumption, detecting constellation- Tbe second group performs the check on the results of the PVT calculation process This group is represented • by the Solution Separation (SS) method [5], monitoring the behavior of multiple satellite subsets on positioning result to mle out any faulty satellite The ARAIM method was developed by the GNSS Evolutionary Architectiire Stiidy (GEAS), aimed at satistying the requirement for LPV-200, and is a modification of the SS method [6] The method is further improved by the Working Group C ARAIM Technical Subgroup of the EU-US Cooperation on Satellite Navigation [7] The mam extensions of ARAIM are the assumption on the threat model, the • Protection Levels (PLs) calculation, and the Integrity I Support Message (ISM) Instead of assuming only one , fault may occur at anytime as the traditional RAIM [8], ARAIM assumes that there maybe multiple faults, resulting from the fauh rate of each individual satellite in view Moreover, while the tiaditional RAIM algorithm calculates the PLs solely based on the geometry of the solution, the ARAIM algonthm also takes into account other fectors, such as the standard deviation of the position fixes, the fault rates of each satellites, etc, ARAIM also use extemal parameters regarding error sources, broadcasted via the ISM, Step Fault-tolerant positions In this step, the position solution is calculated for each fault mode The fault-tolerant position for fault mode k, denoted x^'^^, is calculated with the weighing matnx defined as: W(i,i) W(i.i) ^OViBS^ -Cf;K'-OVi ^Sh (3) Define: (4) The vanance of x f ' and A ^ f \ denoted a^"^^ and o-gf^ respectively, can be deduced from the positioning solution; q — 1,2,3 corresponds to East, North, Up component of the coordinate, respectively Step Solution Separation (SS) test and Chi-square (CS) I Algorithm Description Step A ll-in-view position solution The ali-in-view position is calculated using all the visible satelhtes not being flagged as feulty in the previous steps and have vaUd set of input parameters A weighted least-mean square (LMS) estimation is performed each iteration to find the all-in-view position, denoted x^'*^ The update AJ is given by [7]: AJc = {H'^WHT^H'^Whp For the satellite faults, the maximum number ^satmax of concurent faults is detemuned using the fault rates in the ISM Then, all subsets of up to l^satmax satelhtes are formed, each corresponds to a possible case of fault, called fault mode The indices of the satellites m subset k is denoted S^ The analysis for constellation faults is similar Total number offault modes for both types is denoted Nf^j^ (I) The SS tests' purpose is to check and find the presence of bad satelhtes The SS test is based on the deviation of feult-tolerant positions if a faulty satellite exists [5] Let satelhte j be faulty x^°^ and x^"^ E [x''''|7 S Sfc] should be close together because these positions are under the effect of the faulty satellite, while i ' " " ' G [x^"'' \jESj^] deviates away from the group Thus, the SS test, denoted T/fq is defined as: The geometry matnx H is a (Nsat) x (3 + Wponst) matrix for PVT calculation scheme [8j, where N^at 's the number of satellites, and Nconst is the number of satellite consteUations, The weighing matrk W is calculated as follow: W = c.«ao = (2) whereffyg^_£,CT^^ppj,(Tuser,i^^ variance of the clock and ephemeris accuracy, tropospheric delay and user contribution error, respectively, for the i-th satellite [7] Step Determine the faults to be monitored In this step, all possible cases offault are listed There are two types offault independent simultaneous satellite feults and constellation-wide feults where Kfa,q is a coefficient representing the false alarm rate of the test [7], T^,, is the test threshold If any of the test fails, exclusion must be attempted; otherwise the process can continue with PL calculation The CS test checks for consistency fault and the validity of the to-be-calculated PL The test statistic is calculated as: (6) ^accH(tl^W^^^H)-^H'^Wacc^Ap (7) Tap chi Khoa h^c va Cdng nghe 102 (2014) 064-068 parameters for ISM are referenced from Appendix Jof [9] and Annex B of [7] Caccii = f^^HE.i + (^iropo.i + (^user.i where (TURE,; 'S the variance of the clock and ephemeris error for satellite i, provided in the ISM The test threshold is defmed as: Xl-3-N, i^-PpA t.) (6) where Xaeg^-^ ^ 's the inverse of the cdf of a chisquare distnbution with deg degrees of freedom; N, is the number of constellations; PFACHIZ 'S the predefined false alarm probability for Chi-square test If the test is failed, i.e.^^ > T^z, tiie PL will be declared invalid and the system is not available for navigation purpose Step Exclusion The exclusion step finds and excludes faulty satellites when SS tests failed A list of failed fault modes is sorted in increasing subset size and decreasing T^q For each subset in the list, steps I to are performed on the remaining satellites If the tests on step are all passed, the satellites in the subset are flagged as faulty Step Protection Levels If all the test in step are passed, the PL calculation is executed The VPL is the solution to the equation [7]: Two experiments were executed The first case uses fault-free data input to evaluate the availability of the ARAIM algorithm for LPV-200 operations The second case uses faulted data input to check the capability of the algorithm to detect faults and alert the users The output of the process mcludes ftt positioning result, the Chi-square test to check the validity of the PLs, and the PLs resuh 4.1 Fault-free data input The fu^st result is the Chi-square test output, shown on Figure I The detection threshold is correlated to the number of satellites, as this number determines the degree of fi^eedom for the Chi-square distiibution The test statistic is well below the threshold, stating the validity of the Protection Levels Note that the more satellites is used, the higher tesi statistic can be obtained, because using more satellites would lead to more noise in the PVT calculation process, the risk is higher •P-i ^.ll 25 - Y, P/.«'2 VPL - Tl.-, - b^ (8) ax PllMI„„r na an lai Ma ixe ^m ifai ta - P„m where Q is the left side of the cdf of the Normal distribution, bq' is the effect of nominal biases on Xq [9], PHMIyERT is the predefined integrity budget for the vertical component, P„^ is the integrity budget of unmonitored satellites and constellations [7] The HPL is computed by solving an equation similar to (8), replacing coefficients for horizontal coordinates The method to solve the equations for PLs is provided in [7] and [9], Performance Analysis In this section, the mam implementation results of the ARAIM algorithm are analyzed The input data was recorded on March 27*, 2013 on the premises of NAVIS Centre when all four Galileo in-orbit validation satellites were visible at the same time, along with about 10 GPS satellites [10, 11] The Fig Chi-square test result The test statistic values at the start and the end of the experiment shows some disturbance This is due to the unstable presence of several satellites at low elevation angles These satelhtes alter the geometry of the solution, thus disturb not only the Chi-square test statistic but also, as shown on later figures, the Protection Levels The VPL values are calculated by solving (8), using the binary search method described in [7] and [9] The HPL values are computed in similar manner TTB PLs are depicted on Figure 2, The VPL is higher than HPL, correspondmg to the fact that the error on the vertical direction is always higher than on tta horizontal direction, due to the satellite disti-ibution it each direction The VPL in this case ranges from 12ii to 17m, well underthe required 35m VAL of LPV-200 I ap chi Khoa hoc va Cong n g h | 102 (2014) 064-068 To demonstrate the improvement in term of availability of ARAIM with respect to the tiaditional RAIM algorithm, the PLs are also calculated by traditional RAIM using the method described in [8] The PLs of RAIM are shown on Fig It is clear that PLs of ARAIM is much better than those of the traditional RAIM algonthm, which ranges from 90 m to 220 m The improvement of availability here is possible thanks to the mentioned extensions in threat model assumptions, PLs calculation method, etc The positioning results after the ARAIM process and fault exclusion are showTi on Fig The fixes are I close together in a cloud of roughly meters in radius '4.2 Fault-injected data input 17 of GPS during the time period of t - 700 ^ 800 This fauh heavily affected the consistency of the sateflite measurements during the period, or in other word, the measurement of satellite PRN 17 does not agree with those of other satellites This fault injection therefore should affect the results of the consistency check - the Chi-square test Consequently, the Chi-square test results on Figure increased drastically, well above the detection threshold during this period This fact indicates that the system has a fault, resulting in the PLs during this period is invalid, and the positioning solution is unavailable Note that in this test, the exclusion step of the algorithm had been disabled, so that the biased measurement could affect the Chi-square output To demonstrate the fault detection capability of ,ARAIM algorithm, a bias is injected to satellite PRN Fig ARAIM Protection Levels Fig Traditional RAIM Protection Levels h"SE!™ • "1»H1!" -""" Fig Position feces after ARAIM process Fig Chi-square test result when a manual consistency fault is injected PRN 17 during t = 700 -^ 800 Tap chi Khoa h^c v^ Cdng nghf 102 (2014) 064-068 r IS M^^baorsvaiae-tH., | !" J Fig, 6, Chi-square test result during faulted period (1=700-800) with fault exclusion enabled With exclusion enabled, tbe Chi-square test results came out as shown on Fig ARAIM have detected and excluded the faulty satellites, showing by the reduction of number of satellites while the test statistic values are still underthe threshold This means that despite the faulty input, the system still operates consistently Also, note that the test threshold reduces with the number of satellites, because this value is highly correlated with the number of satellites, as defined in (6) The PLs values after exclusion are shown on Fig The reduction of number of satellites increased the PLs, meaning a slight degradation in term of availability This is understandable because less satellites may reduce the quality of the positioning output However, it is important to note that, the VPL is about 14 m during the faulty penod, which still satisfies the required VAL of 35 m of LPV-200 In other word, ARAIM can detect and exclude the faulted satellite to ensure the integnty of the system, Conclusion RAIM is a technology to assert integnty in aviation However, with recent development of GNSS, new RAIM algorithms have been developed to exploit the multi-GNSS scenario, extendmg the use for more demanding operation, such as LPV-200 In this paper, the ARAIM algorithm was discussed and implemented with real multi-GNSS data The algorithm is an extension of SS method, works well with muhi-GNSS input, can protect against multiple faults The demonstration results show improvement in performance and availability of the ARAIM algonthm with respect to the tiaditional RAIM, The PLs calculated by ARAIM satisfy the requirements of LPV-200 no »D gc Fig Protection Levels durmg faulted penod, with faull exclusion enabled [I] C-129, T S O, T, Airborne Supplemental Navigation Equipment Using the Global Positioning System (GPS), 1992 [2] ICAO, Annex 10, Aeronautical Telecommunications, Volume I (Radio Navigation Aids), Amendment 84, published 20 July 2009, effective 19 November 2009 [3] Brown, R G,, and Chm, G "GPS RAIM, Calculation of Threshold and Protection Radius Using Chi-Square Methods - A Geometric Approach", The Institute of Navigation Monograph Series, Vol V, (1998) 155-78 [4] Lee, Y C "New advanced RAIM wilh inq)roved availability for detecting constellation-wide faults, usmg two independenl constelianons" Proceedings of the 2012 International Technical Meeting of The Institute of Navigahon, Newport Beach, CA, January 2012 [5] Parkinson, B W., and Spilker, J J Global Positioning System: Theory and Applications (volume One), vol Aiaa, 1996 [6] GEAS Panel, Phase of the GNSS Evolutiooaiy Architecture Study [7] Working Group C, ARAIM Technical Subgroup, Interim Report, December 2012 [8] Kaplan, E D„ and Hegarty, C J Understanding GPS; principles and applications, Artech house, 2005 [9] Blanch, J., Walter, T., Enge, P , Lee, Y., Pervan, B , Rippl M„ and Spletter, A, "Advanced RAIM User Algoritluii Description: hitegrity Support Message Processing, Fault Detection, Exclusion, and Protection Level Calculation", ic Proceedings of the 2012 Global Navigation Satellite Systems Conference of Ihe Institute of Navigation (ION GNSS 2012) [10] "Viehiamese University Aimounces First Galileo-Only Positionmg", Inside GNSS, April 2013, available at: http://www.insidegnss.com/node/35l5 [II] "Four Galileo Birds Sighted over Asia", GPS World April 2013, available at: http://gpsworldconi/four-gaiileo-birds-sighted-over-asia ... First Galileo- Only Positionmg", Inside GNSS, April 2013, available at: http://www.insidegnss.com/node/35l5 [II] "Four Galileo Birds Sighted over Asia", GPS World April 2013, available at: http://gpsworldconi/four-gaiileo-birds-sighted-over-asia... capability of ,ARAIM algorithm, a bias is injected to satellite PRN Fig ARAIM Protection Levels Fig Traditional RAIM Protection Levels h"SE!™ • "1»H1!" -""" Fig Position feces after ARAIM process... mam implementation results of the ARAIM algorithm are analyzed The input data was recorded on March 27*, 2013 on the premises of NAVIS Centre when all four Galileo in-orbit validation satellites