September 2014 -1- GPS/IRS HYBRIDIZATION IENAC T12 Anne-Christine Escher Paul Thevenon September 2014 September 2014 -2- GPS NAVIGATION AND GNSS FOR CIVIL AVIATION September 2014 -3- Outline 1. GPS principles 1. Positioning principle 2. GPS segments 2. GPS measurements 1. Measuring the distance 2. Satellite position and time (navigation message) 3. User position 1. GPS measurement model 2. GPS position computation 4. GPS precision 1. SPS accuracy 2. Error budget 3. Constellation geometry 5. GNSS for Civil Aviation 1. Reminder of GNSS definition 2. Development of GNSS in Civil Aviation 3. SiS & ICAO requirements 4. GNSS Autonomous integrity monitoring 5. Onboard autonomous integrity monitoring 6. Signal characteristics 6. Annex GNSS receivers Definitions Reference documents September 2014 1. GPS Principles Positioning Principle – GPS position estimation is based on “triangulation“ from satellites. – Basic principle • To measure the distances and relative velocities between satellites with known orbits and a receiver with unknown position to find the location and the velocity of the receiver. • Satellites only broadcasts to the user (no return link). – Measurements: • The measured distances are propagation delays between the satellite and the receiver. • The measured velocities are the Doppler offset of the received signals. – Result: • Thanks to these measurements, any user can determine the position and velocity of its antenna and synchronize its receiver oscillator with the GPS time reference. It is thus possible to estimate the PVT parameters: Position, Velocity and Time. -4- September 2014 1. GPS Principles GPS space segment: constellation – – – – Minimum of 24 satellites on 6 circular orbital planes with 55° inclination Orbit radius of ~26600 km (altitude ~20200 km) Orbit repetition: 12 siderial hours Ground tracks repeat every 24 siderial hours (= 23h 56m 04.0905s ) -5- September 2014 1. GPS Principles -6- GPS space segment: signals – Each satellite has 3 or 4 atomic clocks (Cesium and Rubidium, the last ones with Rb). – The satellite uses 2 or 3 carrier frequencies in the L band • L1: 1575.42 MHz in an ARNS band • L2: 1227.60 MHz not in an ARNS band • L5: 1176.45 MHz in an ARNS band for new satellites (Block IIF, 2009-?) – An ARNS (Aeronautical Radio Navigation Services) band is allocated by the ITU to aeronautical radionavigation systems. Providing a safety service, such allocations permit to claim protection from other systems. September 2014 1. GPS Principles -7- GPS space segment : signals and services – The signals are transmitted using spread spectrum techniques: a binary signal is the product of the Nav message • navigation message with a low data rate GPS signal • a pseudo-random noise sequence (PRN) PRN sequence with a very high rate. f c L1 C/A L1 P(Y) L5 Carrier freq 1575.42 MHz 1575.42 MHz 1176,.5 MHz Nav message 50 bps 50 bps 50 bps PRN sequence 1.023 MHz 10.23 MHz 10.23 MHz – Example: • The C/A code means Coarse/Acquisition, and was initially created to acquire more rapidly the P(Y) code. Actually, it is used by all civil receivers. • The P(Y) code is a PRN code that can either be the P (Precise) code, or the Y code. The P code P is public, the Y code is secret. This signal is used for (US + allied) military applications. • L5 is a civil signal, opening accuracies to the civil world reserved to military applications up to now: faster coderate + dual frequency processing. September 2014 1. GPS Principles GPS control segment – Functions : • Control the characteristics of the transmitted signals from each satellite • Compute ephemeris data (Keplerian elements) and satellite clock corrections • Upload the navigation message to each satellite – Structure : • It contains – control stations, – upload stations, – a master control station (MCS) in Colorado. • The MCS receives the measurements from each control station, and estimates the ephemerides and clock drift model parameters for each satellite. • There are 4 upload stations (1 to 3 upload per day per satellite). -8- September 2014 1. GPS Principles GPS control segment -9- September 2014 2. GPS Measurements - 10 - Measuring the distance (1/4) – Each satellite (SV) generates its own PRN code. • They are Gold codes whose correlation functions are very typical and interesting for SV identification and signals synchronisation. • The faster the code are transmitted, the more precise synchronisation can be AUTOCORRELATION OF 1 GPS GOLD CODE achieved. 1.2 INTERCORRELATION BETWEEN 2 DIFFERENT GPS GOLD CODES 0.08 0.06 1 0.04 0.8 0.02 0.6 0 0.4 -0.02 0.2 -0.04 0 -0.06 -0.20 200 400 600 800 DELAY [chips] 1000 1200 -0.08 0 200 400 600 800 DELAY [chips] 1000 1200 September 2014 2. GPS Measurements - 11 - Measuring the distance (2/4) – Each signal transmitted by the satellites is modulated using a PRN code. – The receiver (Rx) measures the delay between : • the direct signal received by the antenna coming from the satellite, and • a local replica of that signal generated by the receiver ๏ Asuming the Rx and the SV are generating the same PRN code at the same time, the process consists in detecting the correlation peak. September 2014 2. GPS Measurements - 12 - Measuring the distance (3/4) – SVs are equipped with atomic clocks, but it is not the case for the Rx: Rx clocks are far less accurate. ๏ There is a shift between the SV reference time and the Rx reference time. – The measured delay is referred to as pseudorange (PR). It is the sum of: • the propagation delay between the satellite and the receiver antennas, • plus the shift between the satellite time and the receiver time. September 2014 2. GPS Measurements Measuring the distance (4/4) – A reference time is defined (eg. GPS time). – The shift between the satellite and the receiver times is then the difference between: • the offset between the satellite time and the reference time, and • the offset between the receiver time and the reference time. Reference time scale – The shift between the satellite and the reference times can be corrected ๏ We get the corrected pseudorange ๐๐ ๐๐๐๐๐๐๐ก๐๐ = ๐ + โ๐ก๐ ๐ฅ . ๐ + ๐ - 13 - September 2014 2. GPS Measurements - 14 - Satellite position and time (1/2) – The propagation time depends upon the relative position between the receiver antenna and the satellites antenna. ๏ Assuming the position of the satellites is known as well as the shift between satellite and reference times, the user can estimate its own position and its time shift with respect to the reference time. – The GPS Control Segment estimates these 2 sets of parameters – satellite positions and time shift - for each satellite. – Because of the very stable oscillators onboard satellites, it is possible to have an accurate model of the satellite time to reference time drift. – Ths information is transmitted to the user through the navigation message contained in the signal broadcast by the satellites. September 2014 2. GPS Measurements - 15 - Satellite position and time (2/2) – Example: GPS navigation message • Binary data transmitted at 50 bps on L1 and L2 by each satellite • It is composed of 25 frames of 1500 bits that repeats. The total duration of the message is: 25*1500/50 = 750 s = 12,5 min • Each frame is decomposed into 5 sub-frames of 300 bits. • Each sub-frame is decomposed into 10 words of 30 bits Frame TLM HOW Clock Corrections /URA/Satellite Health (transmitting sat) Sub-frame 1 TLM HOW Ephemeris data (transmitting sat) Sub-frame 2 TLM HOW Ephemeris data (transmitting sat) Sub-frame 3 TLM HOW Almanach/Iono/UTC over 25 frames (all sats) Sub-frame 4 TLM HOW Almanach/Health over 25 frames (all sats) Sub-frame 5 Total duration of a sub-frame: 6 secondes September 2014 3. User Position - 16 - GPS measurements model The GPS corrected measurements general model is the following: ๐๐ ๐๐๐๐๐๐๐ก๐๐ = ๐๐๐ − ๐๐ข๐ ๐๐ + ๐๐ข๐ ๐๐ + ๐ with ๐๐๐ , the satellite location ๐๐ข๐ ๐๐ , the unknown receiver location ๐๐ข๐ ๐๐ , the unknown receiver clock bias Corrected pseudorange User clock bias Line of constant pseudorange September 2014 3. User Position - 17 - GPS position computation – From a set of pseudo-ranges, we can define a set of non-linear equations linking the observations and the user position: 1 1 ๐๐ ๐๐๐๐๐๐๐ก๐๐ = ๐๐๐ − ๐๐ข๐ ๐๐ + ๐๐ข๐ ๐๐ + ๐ 1 2 2 ๐๐ ๐๐๐๐๐๐๐ก๐๐ = ๐๐๐ − ๐๐ข๐ ๐๐ + ๐๐ข๐ ๐๐ + ๐ 2 โฎ ๐ ๐ ๐๐ ๐๐๐๐๐๐๐ก๐๐ = ๐๐๐ − ๐๐ข๐ ๐๐ + ๐๐ข๐ ๐๐ + ๐ ๐ – At least 4 equations are required to solve the 4 unknowns : (x,y,z)user and buser – The solving of this system can be done • at each epoch by using a Non-Linear Least Square estimation technique, usually, taking into account a weighting matrix ๏จ Weighted Least Square estimation. • using all previous epochs with a Kalman filter September 2014 4. GPS Precision Performance with GPS L1 C/A – [GPS SPS PS, 2008] specifies the following time/position performance: - 18 - September 2014 4. GPS Precision - 19 - Error budget - Atmospheric errors – Ionosphere: • Ionized layer of the upper atmosphere • Dispersive (error depends upon the frequency) ๏ can be corrected by dual frequency receiver • Spatial and temporal correlation of the error – Troposphere: • lower atmosphere, • Spatial and temporal correlation of the error Satellite at horizon ๏ goes through a thick layer – Corrections are broadcast in the navigation message Satellite at zenith ๏ goes through a thin layer Ionosphere 50-1000 km Troposphere 0-50 km September 2014 4. GPS Precision - 20 - Error budget - Multipath errors – Satellite signal reflections from the local environment interfere with the direct signal. – The result is a barrage of signals arriving at the antenna: the direct one and delayed replicas known as multipaths. – If the bounced signals are strong enough, they can confuse the receiver and cause erroneous measurements. Diffraction September 2014 4. GPS Precision - 21 - Error budget – Measurements Model – L1 C/A code pseudorange raw measurement at the output of the receiver DLL, for satellite ๐ Satellite to receiver distance ≈ 20,000km Orbit uncertainty Satellite clock bias ≈ 1km Tropo. delay ≈ 2-30m Iono. delay ≈ 0,15-30m Receiver thermal noise ≈ 1m L1 C/A code multipath error ≈ 0-150m for 1 ray ๐ถ1๐ = ๐ ๐ + ๐ โ ๐ก๐ข + ๐๐ ๐ธ ๐ − ๐ โ ๐ก๐๐ ๐ + ๐๐ก๐๐๐๐ ๐ + ๐๐๐๐๐๐ฟ1 ๐ + ๐๐๐๐๐ ๐๐ถ1 + ๐๐๐ถ1 ๐ Code meas. [m] User clock bias ≈ 300m Estimated by the user, as part of the navigation solution Partially compensated using error correction mathematical models May be mitigated using code-carrier smoothing – After corrections application, we get the corrected L1 C/A code pseudorange measurement for satellite ๐ ๐๐ = ๐ ๐ + ๐ โ ๐ก๐ข + ๐๐๐ ๐ด ๐ + ๐๐ก๐๐๐๐ ๐ + ๐๐๐๐๐๐ฟ1 ๐ + ๐๐๐๐๐ ๐๐ถ1 + ๐๐๐๐ถ1 ๐ ๐๐ ๐๐ ๐ธ๐๐ข๐๐ฃ๐๐๐๐๐ก ๐ ๐๐๐๐ ๐ธ๐๐๐๐ ๐๐ธ๐ ๐ธ : ๐๐ September 2014 4. GPS Precision - 22 - Error budget - User Equivalent Range Error (UERE) – UERE represents the impact of errors affecting the propagation of the signal or the signal processing on the pseudorange measurements after application of the errors correction or mitigation techniques – For instance for GPS, the following error budget is proposed by [GPS SPS PS, 2008]: (AOD = Age of data) September 2014 4. GPS Precision Constellation geometry – DOP – The DOP (Dilution Of Precision) value translates the quality of the constellation geometry of the visible satellites to estimate a position or time – It is a unitless parameter which only depends on satellites positions relative to user location. – ๐๐๐๐ = ๐๐๐ธ๐ ๐ธ ∗ ๐ท๐๐ Good DOP Poor DOP ๏ The lower the DOP, the more accurate the estimate - 23 - September 2014 5. GNSS for Civil Aviation - 24 - Reminder of GNSS definition – GNSS (Global Navigation Satellite System): • It is a theoretical concept elaborated by the FANS (Future Air Navigation Systems) committee from ICAO (International Civil Aviation Organization). • It is a part of the global concept of CNS/ATM (Communication Navigation Surveillance/Air Traffic Management). C, N and S support air traffic, while ATM supports its management. – ICAO definition of GNSS: GNSS is defined as a system able to estimate the position and time of the user, and that includes one or several satellite constellations, onboard receivers, and an integrity monitoring system, augmented if necessary, in order to reach required navigation performances for the desired aircraft operation. – The GNSS concept was widely inspired from the GPS and GLONASS systems that were just starting at that time. – Usual definition of GNSS: it is very common to refer to any global satellite-based navigation system as GNSS. September 2014 5. GNSS for Civil Aviation - 25 - GNSS development in Civil Aviation (1/2) – Feb. 94: ICAO Council recommended the use of satellite navigation technology while using as soon as possible existing constellations (GPS, GLONASS, augmentations). – 1994/1996 GPS GLONASS offers: • Oct. 1994: Letter from the US government to ICAO offering, without any direct user charges, an open worldwide positioning service to civil aviation with GPS. At least 6 years notice prior to termination. • June 1996: Similar letter from the Russian federation to ICAO proposing a GLONASS open service for a period of at least 15 years. • Offers were reiterated at various occasions, e.g. February 2007 (180th Session of the ICAO Council) [J. Nagle, ‘ICAO policy on GNSS, GNSS SARPs and global GNSS developments’] – 1998: Assembly resolutions A32-19 and A32-20 • A32-19: “Charter on the Rights and Obligations of States Relating to GNSS Services” • A32-20: “Development and elaboration of an appropriate long-term legal framework to govern the implementation of GNSS” September 2014 5. GNSS for Civil Aviation GNSS development in Civil Aviation (2/2) – GNSS Panel work has led ICAO to approve and publish in November 2002 in the Annex 10 to Chicago convention, GNSS Standards And Recommended Practices (SARPS) to support all phases of flight up to Cat I (Precision Approach). – In 2003, the 11th Air Navigation Conference recommends a worldwide transition to GNSS-based air navigation. - 26 - September 2014 5. GNSS for Civil Aviation - 27 - Signal In Space & ICAO requirements – Signal in Space (SiS) is the signal transmitted by satellites • It is part of the Navigation System Error (NSE) โ ICAO SiS requirements are set assuming the combination of GNSS elements and a faultfree GNSS user receiver. โ GNSS comprises • • • Core constellations. Augmentations: ABAS, SBAS and/or GBAS Onboard receivers โ Requirements are defined per operation as: • • • • Accuracy (-95%), integrity, continuity, and availability Operation Accuracy- 95% Integrity Continuity Availability En-route En-route, Terminal Initial Approach, Intermediate, NPA, Departure APV1 APV2 CAT1 ICAO GNSS standards were first published in November 2002 in Annex 10 of the Chicago Convention. September 2014 5. GNSS for Civil Aviation - 28 - Signal In Space & ICAO requirements – Accuracy: characterize typical behavior of the system in presence of nominal errors. – Integrity: Limit risk of abnormal behavior of the system due to errors resulting from system faults. It is composed of several parameters: – Maximum Tolerable Error / Alert Limit – Time to Alert – Integrity risk: probability that error > MTE MTE Pos error being detecting after TTA Actual position True position – Continuity: Limit risk of losing the service unexpectedly – Availability: fraction of time that one has Accuracy + Integrity + Continuity September 2014 5. GNSS for Civil Aviation - 29 - ICAO Signal In Space requirements (1/2) Table 3.7.2.4-1 in Annex 10, Chapter 3 of the Chicago Convention Operation Accuracy- 95% Integrity Continuity Availability Horizontal Vertical Time to Alert Integrity Risk En-route 3.7km N/A 5min 1-1x10-7/h 1-1x10-4/h to 1-1x10-8/h 0.99 to 0.99999 En-route, Terminal 0.74km N/A 15s 1-1x10-7/h 1-1x10-4/h to 1-1x10-8/h 0.99 to 0.99999 Initial Approach, Intermediate, NPA, Departure 220m N/A 10s 1-1x10-7/h 1-1x10-4/h to 1-1x10-8/h 0.99 to 0.99999 APV1 16m 20m 10s 1-2x10-7 in any approach 1-8x10-6 per 15s 0.99 to 0.99999 APV2 16m 8m 6s 1-2x10-7 in any approach 1-8x10-6 per 15s 0.99 to 0.99999 CAT1 16m 6m to 4m 6s 1-2x10-7 in any approach 1-8x10-6 per 15s 0.99 to 0.99999 September 2014 5. GNSS for Civil Aviation - 30 - ICAO Signal In Space requirements (2/2) Table 3.7.2.4-1 in Annex 10, Chapter 3 of the Chicago Convention Operation Alert Limit Horizontal Vertical En-route (oceanic/continental low density) 7.4km N/A En-route (continental) 3.7km N/A En-route, Terminal 1.85km N/A NPA 556m N/A APV1 40m 50m APV2 40m 20m CAT1 40m 35m to 10m September 2014 5. GNSS for Civil Aviation - 31 - GNSS autonomous integrity monitoring – Civil aviation requirements can be very stringent and up to now, the bare satellite navigation systems alone cannot be used as a means of navigation. – To ensure the levels required in terms of accuracy, integrity, continuity of service and availability, ICAO standards define different architectures to augment the basic constellations: • Some of them use control stations to monitor satellite signals and provide corrections: – GBAS (Ground-Based Augmentation System) – SBAS (Satellite-Based Augmentation System) • Others only use GNSS measurements redundancy or combine them with on-board navigation sensors: – ABAS (Aircraft-Based Augmentation System) – These solutions are based onto a Fault Detection (FD) or a Fault Detection and Exclusion (FDE) function • FD allows to perform the detection of signals anomalies that is to say to make sure of the integrity of the used signals • Upon detection, FDE allows to make sure of the continuity of the navigation after detection occurs. September 2014 5. GNSS for Civil Aviation - 32 - GNSS autonomous integrity monitoring – FD performances are measured by computing protection level - H/VPL - values. – The xPL is a bound of the positioning error which translates the quality of the positioning from an integrity point of view. ๏ The lower the xPL, the more confident the positioning. The xPL computed value depends on the constellation geometry and the UERE. HPL HAL Pos error Actual position – Example: True position In horizontal plane using RAIM for Fault Detection: ๏ HPL is a bound of the positioning error computed such as any error causing the horizontal positioning error to exceed HPL is guaranteed to be detected with a given probability (eg. 10-3 for NPA) that itself guarantees the Integrity Risk is satisfied. ๏ The GNSS means of navigation is claimed to be available, that is usable for navigation in the intended operation, whenever HPL<HAL provided it also fulfills the ICAO accuracy and continuity requirements. September 2014 5. GNSS for Civil Aviation Onboard autonomous integrity monitoring Receiver Autonomous Integrity Monitoring (RAIM) – Uses redundant GPS measurements only – RAIM availability allows ICAO aviation operations from en route down to NPA. – Widely implemented from general aviation to commercial aviation. Aircraft Autonomous Integrity Monitoring (AAIM) – Based on GPS measurements redundancy and onboard inertial information – AAIM availability allows ICAO aviation operations from en route down to NPA • For instance, Northrop Grumman AIME is certified as a primary means of navigation on Airbus aircrafts family • Supports Fault Detection and Exclusion (FDE) with an increased availability wrt RAIM and better anomalies detection performances. – Implemented on higher end commercial aviation only - 33 - September 2014 5. GNSS for Civil Aviation - 34 - Signal characteristics (1/2) – Core constellation satellites are to be observed at an elevation angle of at least 5° (mask angle) • To guarantee the power of the received signal in an unobstructed location (-160 dBW to -153 dBW) • To limit multipath effects • In ICAO Annexe 10, core constellation is defined as GPS+GLONASS [www.spacegeodesy.re.kr/sgd/sgd02.aspx] – WGS-84 is the position reference frame – Signal processing thresholds are set to • ๐ถ ๐0 = 29.93 ๐๐ต๐ป๐ง, for signal acquisition • and ๐ถ ๐0 = 32.4 ๐๐ต๐ป๐ง, for signal tracking and demodulation. This guarantees a WER of 10−3 – Today, only mono-frequency code pseudorange measurements are used for positioning on board C/A aircraft • GPS L1 C/A signals are used worldwide. September 2014 5. GNSS for Civil Aviation - 35 - Signal characteristics (2/2) – To be used for approach operations, a code-carrier smoothing must be implemented to mitigate multipath and noise impact on the used measurements • Indeed, carrier measurement are less affected by noise and multipath (typically, cm-level accuracy) • This filter has a 100s-constant time: the receiver must wait 3.6 times this time before using the smoothed code measurements for positioning • Using code-carrier smoothing implies to be able to detect cycle slip occurrences that would affect the carrier measurements Residual Multipath Error Std Model (including noise receiver) • Yet, non-smoothed measurements are used when implementing inertial hybridization 0.145 0.14 ๐๐๐๐_๐๐ = 0.11 + 0.13๐๐ฅ๐ −๐ธ๐/4 0.135 0.13 [m] – A multipath residual error model for GPS L1 C/A provides information on the residual accuracy after smoothing. • For protection level determination • For position computation (Weighted LS) 0.125 0.12 0.115 0.11 0.105 0.1 0 10 20 30 40 50 Elevation [deg] 60 70 80 90 September 2014 6. Annexes Onboard GNSS receivers – To operate Area Navigation (RNAV) procedures, onboard equipment is not only composed of the GNSS receiver, but also, • an RNAV computer to elaborate the guidance, • an aircraft navigation display • a Man-Machine Interface for crew interface, • and an updated navigation data base. – For general aviation aircraft, this equipment is gathered in 1 or 2 systems. – For air transport aircraft, • the core GNSS receiver may be replaced by a Multi-Mode System (MMR) which gathers many sensors and functionalities (GNSS and ILS receivers; FLS, GLS and MLS modes) for multiple phases of flight. • the RNAV computer is in the FMS (Flight Management System) which also includes the navigation data base and is connected to different navigation sensors as inertia. • the Man-Machine Interface is achieved through the MCDU (Multi Control Display Unit) - 36 - 6. Annexes September 2014 Definitions – ABAS - Aircraft-Based Augmentation System – is an augmentation system which augments and/or integrates the information obtained from the other GNSS elements with information on board the aircraft in order to ensure the SIS performance. – SBAS - Satellite-Based Augmentation System – is a wide coverage augmentation system in which the user receives augmentation information from a satellitebased transmitter. It is made up of the ground infrastructure, the SBAS satellites, and the SBAS airborne receiver. – GBAS - Ground-Based Augmentation System – consists of ground and aircraft elements. One ground station can support all the aircraft within its coverage providing them with approach data, corrections and integrity information for GNSS satellites in view via a VHF data broadcast (VDB). – The Signal In Space (SIS) is the aggregate of guidance signals arriving at the antenna of an aircraft assuming a fault-free receiver. - 37 - 6. Annexes September 2014 - 38 - Reference Documents ICAO document (SARPS) • International Standards and Recommended Practices, Annex 10 to the Convention on International Civil Aviation, Volume 1, Sixth Edition, 2006. Amendment 85. RTCA, Inc. Documents (RTCA DO) documents • MOPS for airborne supplemental navigation equipment using Global positioning System (GPS), DO 208, 1991 • MOPS for GPS Wide Area augmentation System Airborne Equipment, DO 229C, 2001 • MOPS for GPS Wide Area augmentation System Airborne Equipment, DO 229D, 2007 FAA Standard Technical Orders • TSO C129a, airborne supplemental navigation equipment using the global positioning system (GPS) • TSO C145a, airborne navigation sensors using the global positioning system (GPS) augmented by the Wide Area Augmentation System (WAAS) • TSO C146a, standalone airborne navigation sensors using the global positioning system (GPS) augmented by the Satellite Based Augmentation System (WAAS) • TSO C145c, airborne navigation sensors using the global positioning system augmented by the Satellite -Based Augmentation System Other documents • Global Positioning System Standard Positioning Service - Performance Standard, 4th edition, DoD, September 2008 September 2014 - 39 - INERTIAL NAVIGATION AND GPS/IRS HYBRIDIZATION September 2014 - 40 - Outline 1. Inertial navigation 1. INS/IMU components 2. Inertial navigation principle 3. Sensors 4. Navigation problem presentation 5. Commonly used frames 6. Inertial navigation solution 7. Inertial navigation properties 8. Platform implementation 2. Illustration of IRS classical implementation for aircraft 1. Simulation assumptions 2. Sensors measurements modeling 3. IRS mechanisation in the NED navigation frame 4. Simulation results 3. GPS/IRS hybridization 1. Principle and schemes 2. Hybridation architectures 3. Overview of Kalman filtering 4. Example of GPS/IRS tight coupling using KF 5. Integrity Monitoring 6. Conclusion September 2014 1. Inertial Navigation INS/IMU components Inertial Measurement Unit IMU ๐๐/๐ผ ๐ ๐๐/๐ผ ๐ computer [Systron dqi-sd] – An Inertial Navigation System (INS) or Inertial Measurement Unit (IMU) is composed of accelerometers, gyrometers and a computer. • • It aims at estimating the mobile position, velocity and attitude in an appropriate navigation frame, (N). Besides the computer may also estimate a series of parameter that are useful for aircraft piloting and guidance: • Ground speed • Track angle (angle between North and the route) • Flight Path Angle (angle between horizontal plane and the route) - 41 - September 2014 1. Inertial Navigation Inertial navigation principle – It is based on dead-reckoning navigation • • After the alignment of (p) w.r.t. (N) has been achieved, ie attitude determination, the accelerometric measurements are integrated to get the mobile velocity in the navigation frame. Depending on the platform implementation, this alignment may be performed in either a mechanical or an analytical manner. – Strong points: • • • autonomous – does not depend on external aid in comparison to GPS non-radiating – in comparison to conventional NAVAIDS insensitive to external perturbations – meteorological perturbations, jamming, RF interference, multipath – Drawbacks: • • Time drift of the position estimate Cost due to the used technologies to get sensors of great accuracy and great reliability - 42 - September 2014 1. Inertial Navigation Sensors Accelerometers: principle – Obtain the vehicle acceleration – An accelerometer senses the vehicle specific force along its sensitive axis. Its output has to be corrected to take into account the gravity acceleration. – Ex: what would be the specific force measured by a free-falling object? – Two major designs: • open-loop or vibratory accelerometers • closed-loop or force-rebalance accelerometer (illustration): standard and most accurate design - 43 - September 2014 1. Inertial Navigation - 44 - Sensors Accelerometers: examples of technologies Pendulous Integrated Gyro (PIG) Accelerometer - 25 PIGA Titan II Pendulous accelerometer Axe sensible Axe pendule ๏ฒ f http://www.nasm.si.edu/images/collections/media/full/T20090008003.JPG Micro-Electronical Mechanical System (MEMS) accelerometer Analog Device ADXL202 http://www.machinegrid.com/2008/12/accelerometers-for-your-robot- September 2014 1. Inertial Navigation - 45 - Sensors Accelerometers: error model ๐1 = ๐1 + ๐ + ๐0 + ๐๐๐๐ ๐ ๐๐๐๐ ๐1 ๐1 + ๐12 ๐2 + ๐13 ๐3 + ๐๐ ๐ ๐ ๐๐๐๐ ๐๐๐๐ก๐๐ ๐๐๐๐๐๐๐๐๐ก ๐๐๐๐๐ ๐ก๐๐๐๐๐๐๐ก๐ข๐๐ ๐๐๐๐ f1 f component along the sensitive axis f2,f3 f "cross-axis" components k12, k13 are the sensitivity factors n Noise – sensor sentitivity threshold k0 bias k1 Scale factor - linear kTT Temperature difference in comparison to calibration - bias These parameters characterize the accelerometer accuracy +โฏ September 2014 1. Inertial Navigation - 46 - Sensors Accelerometers: technologies and applications Technologies VBA Pendulous accelerometer MEMS PIGA Bias (mg) 10 munitions 1 Flight controls 0.01 0.1 Tactical missiles aircraft navigation Launchers AHRS Applications 0.001 Marine navigation Strategic missiles [source Thales] September 2014 1. Inertial Navigation Sensors Gyrometers: principle – Sense the vehicle inertial rotation rate: • • In the stable platform case, they measure the platform rotation which is free from the mobile. ๏ The platform frame is kept in a known direction by mechanical means, typically a system of gimbals and a gyroscopically controlled feedback loop. In the strap-down case, the platform is linked to the mobile and follows its angular movements. – Main designs • • Vibratory gyro, based on mechanical effects Optical gyro (illustrated), based on the Sagnac effect - 47 - September 2014 1. Inertial Navigation - 48 - Sensors Gyrometers: examples of technologies Mechanical and Fiber-optic gyros http://www.aerospaceweb.org/question/weapons/ guidance/gyroscope.jpg Rotor-gyro and housing Ring laser gyroscopes Aerospace Topflight ADIRU sheet, Thales Avionics, 2008. http://www.laserfest.org/lasers/images/inngyroscope.gif MEMS gyroscope – GS 12 Robotis September 2014 1. Inertial Navigation - 49 - Sensors Gyrometers: error model ๐1 = ๐1 + ๐ + ๐0 + ๐๐๐๐ ๐ ๐๐๐๐ ๐1 ๐1 + ๐12 ๐2 + ๐13 ๐3 + ๐๐ ๐ ๐ ๐๐๐๐ ๐๐๐๐ก๐๐ ๐๐๐๐๐๐๐๐๐ก ๐๐๐๐๐ ๐ก๐๐๐๐๐๐๐ก๐ข๐๐ ๐๐๐๐ ω1 f component along the sensitive axis ω2, ω3 f "cross-axis" components k12, k13 are the sensitivity factors n Noise – sensor sentitivity threshold k0 bias k1 Scale factor - linear kTT Temperature difference in comparison to calibration - bias kg G-dependant bias + ๐๐ ๐1 … September 2014 1. Inertial Navigation Sensors Gyrometers/accelero: error model – A mathematical model can be defined, it will involve: • • • • noise: random additive error on the measurement, bias: slow-varying error in the measurement, scale factor: error depending on the amplitude of the measurement, mis-alignment error: error in the alignment of the sensor measurement axes from the orthogonal platform axes. ๏ But, its exact expression depends on the technology. – The total effect of the sensor bias and the scale factor error shows up the sensor drift. – All these errors characterize the sensor accuracy. - 50 - September 2014 1. Inertial Navigation - 51 - Sensors Gyrometers: technologies and applications Technologies Ring Laser gyros Vibrating FOG ESG Floated MEMS 100 munitions DTG 10 Flight controls 1 0.1 Tactical missiles Drift (degree/hr) 0.001 0.01 aircraft navigation Marine navigation Launchers AHRS Applications Strategic missiles [source Thales] September 2014 1. Inertial Navigation Sensors Sensor grades Sensor grades are used to describe the application range targeted by a sensor. They indicate their performances. Grade Position drift Cost Application Marine 1.8 km/day 1 M$ Military ships, submarines, intercontinental missiles, spacecrafts Aviation / Navigation 1.5 km/hour 100 k$ Military and commercial aviation Intermediate 15 km/hour 20-50 k$ Small aircraft, helicopter Tactical Ok for a few minutes 2-30 k$ Guided weapon, UAV Consumer / Automotive Not integrated in IMU, but rather sold as individual sensors 1-10$ Pedometer, ABS, active suspension, airbags Groves 2013 – Principles of GNSS, Inertial and Multisensor Integrated Navigation Systems, Artech House - 52 - September 2014 1. Inertial Navigation - 53 - Sensors Navigation-grade sensors typical figures Typical strapdown sensor performance requirements for a navigation-grade IRS [Palmqvist, 1997]: Performance Parameter Gyro Bias Uncertainty ๏จdeg hr ๏ฉ Requirements 0.01 Gyro Random noise deg 0.003 ๏จ hr ๏ฉ Gyro Scale Factor Uncertainty ๏จ ppm๏ฉ Acceleration Bias Uncertainty ๏จ๏ญg ๏ฉ Acceleration Scale Factor Uncertainty ๏จ ppm๏ฉ Sensor Alignement Uncertainty ๏จsec ๏ฉ 10 50 100 10 September 2014 1. Inertial Navigation - 54 - Navigation problem presentation – The sensors are implemented onto the gyro-accelerometric platform. • Accelerometers sense the vehicle rectilinear acceleration. Each one measures, along its sensitive axis, the inertial specific force of the mobile in the platform coordinate frame. • Gyrometers sense the vehicle rotational motion which is the mobile angular velocity with respect to the inertial frame. It is also expressed in the platform coordinate frame – The measurements corresponds to the specific force or angular velocity of the mobile (m) with regard to an inertial reference frame (I), expressed in the platform frame (p). – What we want to obtain is the mobile (m) characteristics with regard to the terrestrial frame (e), and possibly expressed in the navigation frame (N). ๏The navigation problem consists in ๐๐/๐ฐ ๐ ? ๐๐/๐ฌ ๐๐/๐ฌ ๐ต ๐๐/๐ฐ ๐ ๐๐/๐ต September 2014 1. Inertial Navigation - 55 - Commonly-used frames (1/3) Z๏บ ZT n R O X ๏น e M – Terrestrial frame (ECEF) – (E): (O,XT,YT,ZT) YT – Navigation frame – (N): eg. the NED coordinate frame (M,n,e,v) Y L ωIE.t n v λ – Space-fixed or inertial frame (ECI) - (I): (O,X,Y,Z) XT – Body or mobile frame - (m): (M,x,y,z) orientation with respect to the NED is defined by the attitude angles (Euler angles): • roll (Φ) – rotation around (n), • pitch (θ) – rotation around (e), • heading (ψ) – rotation around (v). Greenwich meridian x X Y Z ๏ z ๏ฑ v e y Xsens Mti IMU – Gyro-accelerometric platform frame - (p): • The gyrometers’s position defines the platform trihedron. • Accelerometers are placed along the axes of that trihedron. – (p) and (m) are assumed to be aligned in this ๐ ๐ ๐ lecture: ๐๐/๐ฐ = ๐๐ ๐/๐ฐ , ๐๐/๐ฐ = ๐๐/๐ฐ September 2014 1. Inertial Navigation Commonly-used frames (2/3) • Inertial navigation equations can be solved in any frame (I), (E) or (N). – (I) has the most simple equations. – (E) or (N) being non-inertial frames, they introduce Coriolis force. – (N) origin also adds a transport term, since the origin of (N) moves wrt (I). – However, (N) is often the desired solving frame for users. • Solving the inertial navigation equations needs to change the coordinates frame, notably, to express the specific force measurement ๐๐ ๐/๐ฐ in the desired solving coordinate frame (N) ๐๐ต ๐/๐ฐ . ๐ • This is done thanks to a coordinate transformation matrix ๐๐ต ๐/๐ฐ = ๐น๐๐๐ต . ๐๐/๐ฐ - 56 - September 2014 2. Illustration of IRS Implementation for Aircraft Commonly-used frames (3/3) Expression of the change of coordinates matrix Rm2N The components of a vector expressed in the mobile frame may be expressed in the navigation frame by the series of the three rotations: ๏ฉ x๏ข ๏น ๏ฉ cos๏น sin ๏น 0๏น ๏ฉ x ๏น ๏ช y๏ข๏บ ๏ฝ ๏ช๏ญ sin ๏น cos๏น 0๏บ ๏ ๏ช y ๏บ ๏ช ๏บ ๏ช ๏บ ๏ช ๏บ ๏ช๏ซ z ๏ข ๏บ๏ป ๏ช๏ซ 0 0 1๏บ ๏ช z ๏บ ๏ฑ๏ด๏ด๏ด๏ฒ๏ด๏ด๏ด๏ณ๏ป ๏ซ ๏ป N rotation๏จ๏น / z ๏ฉ Rm 2 N ๏ฉcos๏น ๏ cos ๏ฑ ๏บ ๏๏น ๏๏ ๏๏ฑ ๏๏ ๏๏ฆ ๏ ๏บ ๏ช๏ช sin๏น ๏ cos ๏ฑ ๏ช๏ซ ๏ญ sin ๏ฑ – Attitude angles estimates: ๏ฉ x๏ข๏ข ๏น ๏ฉcos ๏ฑ 0 ๏ญ sin ๏ฑ ๏น ๏ฉ x๏ข ๏น ๏ช y๏ข๏ข๏บ ๏ฝ ๏ช 0 1 0 ๏บ๏บ ๏ ๏ช๏ช y๏ข๏บ๏บ ๏ช ๏บ ๏ช ๏ช๏ซ z ๏ข๏ข ๏บ๏ป ๏ช๏ซ sin ๏ฑ 0 cos ๏ฑ ๏บ๏ป ๏ช๏ซ z ๏ข ๏บ๏ป ๏ฑ๏ด๏ด๏ด๏ฒ๏ด๏ด๏ด ๏ณ rotation๏จ๏ฑ / y ' ๏ฉ ๏ญ sin๏น ๏ cos ๏ฆ ๏ซ cos๏น ๏ sin ๏ฑ ๏ sin ๏ฆ cos๏น ๏ cos ๏ฆ ๏ซ sin๏น ๏ sin ๏ฑ ๏ sin ๏ฆ cos ๏ฑ ๏ sin ๏ฆ ๏ฑ ๏ฝ ๏ญ arcsin ๏จRm 2 N ๏จ3,1๏ฉ๏ฉ ๏ฆ Rm 2 N ๏จ3,2 ๏ฉ ๏ถ ๏ท๏ท ๏จ ๏ฉ R 3 , 3 ๏จ m2 N ๏ธ ๏ฆR ๏จ2,1๏ฉ ๏ถ๏ท ๏น ๏ฝ arctan ๏ง๏ง m 2 N ๏ท ๏จ Rm 2 N ๏จ1,1๏ฉ ๏ธ ๏ฆ ๏ฝ arctan ๏ง๏ง 0 0 ๏น ๏ฉ x๏ข๏ข ๏น ๏ฉu ๏น ๏ฉ1 ๏ช v ๏บ ๏ฝ ๏ช0 cos ๏ฆ sin ๏ฆ ๏บ ๏ ๏ช y๏ข๏ข๏บ ๏ช ๏บ ๏ช ๏บ ๏ช ๏บ ๏ช๏ซ w๏บ๏ป m ๏ช๏ซ0 ๏ญ sin ๏ฆ cos ๏ฆ ๏บ๏ป ๏ช๏ซ z ๏ข๏ข ๏บ๏ป ๏ฑ๏ด๏ด๏ด๏ฒ๏ด๏ด๏ด ๏ณ rotation๏จ๏ฆ / x '' ๏ฉ sin๏น ๏ sin ๏ฆ ๏ซ cos๏น ๏ sin ๏ฑ ๏ cos ๏ฆ ๏น ๏ญ cos๏น ๏ sin ๏ฆ ๏ซ sin๏น ๏ sin ๏ฑ ๏ cos ๏ฆ ๏บ๏บ ๏บ๏ป cos ๏ฑ ๏ cos ๏ฆ - 57 - September 2014 1. Inertial Navigation - 58 - Inertial navigation solution in (I) frame Z • Let ๐ be the mobile location: ๐ ๐ผ = ๐๐ with ๐ the Earth’s center R • In the inertial frame, the accelerometric measurement provides: ๐2 ๐ผ ๐ผ ๐ผ ๐ = ๐ − ๐ ๐ + ๐๐ ๐๐ก 2 ๐ ๐๐๐๐๐๐๐ ๐๐๐๐๐ ๐กโ๐๐ก ๐ค๐๐ข๐๐ ๐๐ ๐๐๐ ๐ข๐๐๐ ๐๐ฆ ๐๐๐ ๐๐๐๐๐ ๐๐๐๐๐๐๐๐๐๐๐ก๐๐ M O X ๐๐๐๐ ๐ข๐๐๐๐๐๐ก ๐๐๐๐๐ • The first Newton’s law is applied to estimate the vehicle inertial acceleration: ๐2 ๐ผ ๐ = ๐ ๐ผ + ๐๐ผ ๐ 2 ๐๐ก • The inertial velocity and position are deduced through double integrations • There are 2 issues in the case of INS measurements: - The measurement are expressed in the (m) frame ๏จ coordinate transformation - The gravity is absent from the measurement ๏จ gravity model Y September 2014 1. Inertial Navigation - 59 - Inertial navigation solution in (I) frame ๐ ๐2๐ผ (-) ๐ ๐๐/๐ผ ๐ ๐๐/๐ผ 1. Update attitude 2. Transform specific force frame Groves 2013 – Principles of GNSS, Inertial and Multisensor Integrated Navigation Systems, Artech House ๐ผ ๐๐/๐ผ 3. Update velocity ๐ผ ๐ฃ๐/๐ธ (−) ๐๐ผ ๐ 4. Update position ๐ผ ๐๐/๐ผ (-) ๐ ๐2๐ผ (+) ๐ผ ๐ฃ๐/๐ผ (+) ๐ผ ๐๐/๐ผ (+) September 2014 1. Inertial Navigation - 60 - Inertial navigation solution in (N) frame ๐ ๐2๐ (-) ๐ ๐๐/๐ผ ๐ ๐๐/๐ผ 1. Update attitude 2. Transform specific force frame Groves 2013 – Principles of GNSS, Inertial and Multisensor Integrated Navigation Systems, Artech House ๐ ๐๐/๐ผ 3. Update velocity ๐ ๐ฃ๐/๐ธ (−) ๐๐ (−), ๐ฟ๐ (−), โ๐ (−) ๐๐ ๐ 4. Update position ๐ ๐2๐ (+) ๐ ๐ฃ๐/๐ธ (+) ๐๐ (+), ๐ฟ๐ (+), โ๐ (+) September 2014 1. Inertial Navigation - 61 - Inertial navigation solution – Inertial navigation requires computer initialization (position, velocity, attitude) – Gravity model: Gravity acceleration compensation depends on the current location (latitude & height over the ellipsoid) • • A gravity loop is implemented to improve the acceleration estimate The gravity acceleration correction uses the estimated point provided at the end of the computation chain ๏จ๏ฉ g I Rˆ 2 d ˆI R dt 2 ~I f d ˆI R dt ๏ฒ t ๏ฝ0 ˆ d ˆI R R dt ๏ฒ I t ๏ฝ0 Rฬ I September 2014 1. Inertial Navigation Inertial navigation properties Position performance – The accuracy of the computed instantaneous position at the IRS output is due to • initialisation errors, • accelerometric and gyrometric measurements errors, • errors in the (m)-to-(N) frames change of coordinates definition, • Gravity model uncertainty • and approximations due to numerical integrations. – For long-range air-transport aircraft performance must be lower than 2 NM/hr (95%) - 62 - September 2014 1. Inertial Navigation - 63 - Inertial navigation properties Position performance example We assume a strapdown inertial system mounted in a vehicle travelling at a constant speed and at constant height above the Earth: East velocity: 500 kt (~926km/hr) • Latitude: 45° Illustration of the middle-term contribution error models. Initial azimuth error (1 mrad) Initial speed error (0.3 m/s) East gyro bias (0.01 °/hr) Vertical gyro bias (0.01 °/hr) Accelerometer bias (50 ๏ญg) Total RSS error 6 Position Error [km] • North Position Error vs Time 4 2 [Titterton, Weston, 2004] 0 0 1 2 Time [h] 3 4 September 2014 1. Inertial Navigation - 64 - Platform implementation Stable platform – The sensor platform is free from the vehicle. • • Gyroscopes data are used to mechanically stabilized the sensor platform w.r.t. the navigation frame trihedral (N) (use of actuators). Sensor measurements are directly in the navigation frame. ๐ Inertial sensors computer ๐ ๐๐ Command laws for platform stabilization Roll, pitch, heading Gravity Compensation Integration Rฬ d ˆ R dt September 2014 1. Inertial Navigation - 65 - Platform implementation Stable platform illustration Schematic of a three-gimbal platform www.accessscience.com, Encyclopedia article: Inertial navigation system Photo of Internals of a Gimballed Inertial Platform INS Mechanism. © GEC Marconi http://www.al-nasir.com/www/Jamie/Articles/Technology/... GimballedINSPhoto.gif September 2014 1. Inertial Navigation - 66 - Platform implementation Strapdown platform (IRS) – IRS is for Inertial Reference Sensor. – The sensor platform is fixed with respect to the vehicle. • • The sensor platform (p/m) is analytically (ie. virtually) aligned with the navigation frame trihedron (N) Accelerometric measurements are in the platform coordinate frame. computer (p/m) to (N) alignment Accelero. Gravity Compensation Integration Gyro. ๏ฒ Attitude angles Rฬ d ˆ R dt Roll, pitch, heading September 2014 1. Inertial Navigation - 67 - Platform implementation Strapdown platform illustration Strap-down inertial navigation system. Litton LTN-92 laser gyro INS Inertial Navigation Unit, INU www.accessscience.com, Encyclopedia article: Inertial navigation system (After A. Lawrence, Modern Inertial Technology, 2d ed., Springer, 1998) Command and http://www.es.northropgrumman.com/solutions/ltn92/assets/LTNDisplay Unit , CDU 92_Ring_Laser_Gyro_Inertia.pdf Mode Selector Unit, MSU September 2014 1. Inertial Navigation - 68 - Platform implementation On board aircraft implementations (1/2) – On board high end commercial aviation aircraft, the Inertial Reference Unit is implemented in ADIRU – Air Data Inertial Reference Unit. ADIRU is divided in 2 parts, each one can work separately in case of failure in the other • • the Air Data Reference (ADR) part which supplies barometric altitude, airspeed, mach, angle of attack, temperature and overspeed warnings. the Inertial Reference (IR) part which supplies attitude, flight path vector, track, heading, accelerations, angular rates, ground speed and aircraft position. – Localisation may be output either as pure IRS position with baro-hybridized vertical axis or GPS/IRS/baro position. – Hybridization processes are implemented inside the ADIRU. Aerospace Topflight ADIRU sheet, Thales Avionics, 2008. September 2014 1. Inertial Navigation - 69 - Platform implementation On board aircraft implementations (2/2) – On board business, regional and smaller air-transport aircraft, the Inertial Reference Unit is implemented as AHRS – Attitude & Heading Reference System. AHRS system may consist of: • Gyrometers, accelerometers - with lower performance than for ADIRU (intermediate grade) - and/or inclinometers, and a Magnetic Sensing Unit, to provide user with attitude and heading (magnetic and true). • It may also provide GPS/AHRS/altitude/AirData velocity and position. • Hybridization processes are implemented inside the AHRS. AHRS diagram example Extract of DO 334 - RTCA, 2012 September 2014 - 70 - Part 2 Illustration of IRS implementation for aircraft 1. 2. 3. 4. Simulation assumptions Sensors measurements modeling IRS mechanisation in the NED navigation frame Simulation results September 2014 2. Illustration of IRS Implementation for Aircraft Simulation assumptions – Computation of a baro-IRS trajectory with Commercial Aviation grade performance assumptions for the sensors, based on a simulated flight path. • IRS provides the pilot with IRS horizontal position and velocity, as well as roll, pitch and heading angles Baro-altimeter data provide the pilot with vertical position, and are used to aid the IRS in the vertical velocity and accelaration computations Gravity loop is not implemented • • Roulis 0.9 0.2 [rad] 0.4 0 0.85 -0.2 0.8 0 1000 2000 3000 4000 5000 6000 4000 5000 6000 4000 5000 6000 Tangage 0.5 0 0.005 0.015 0.02 0.025 0.03 Longitude [rad] Trajectoire de reference sur l'axe vertical 4 6 x 10 0.01 0.035 0.04 [rad] 0.75 -0.005 0 -0.5 0 1000 2000 4 3000 Cap 5 [rad] Altitude [m] Latitude [rad] Trajectoire de reference dans le plan horizontal 0.95 2 0 0 1000 2000 3000 Temps [s] 4000 5000 6000 0 -5 0 1000 2000 3000 Temps - 71 - September 2014 2. Illustration of IRS Implementation for Aircraft - 72 - Sensors measurements modeling The IMU and baro-altimeter sensors mesurements are deduced from the reference path using time difference. The sensors performance are modelled as additionnal errors on ๐ ๐ the ideal measurements ๐๐/๐ผ , ๐๐/๐ผ , โ๐๐๐๐ • 1st order Gauss-Markov process (bias ๐๐ ) Time constant 300s Standard deviation (1σ) 25μg ba ~m f mI f mIm Gyrometers • ~m ๏ท mI m ๏ท mI Accelerometers 1st order Gauss-Markov process (bias ๐๐ ) Time constant 3600s Standard deviation (1σ) 0.01°/hr bg hbaro ~ hbaro Baro-altimeter • Additive white-centered Gaussian noise (๐๐๐๐๐ ) nbaro ๏จ๏ณ ๏ฝ 15m๏ฉ September 2014 2. Illustration of IRS Implementation for Aircraft - 73 - IRS mechanisation in the NED navigation frame General scheme (w/o gravity loop) The sensor trihedron, (๐), is assumed to be exactly aligned with the mobile trihedron, (๐) outputs Attitude Acceleration Velocity Position Specific force wrt to the Inertial referential frame + Acceleration Integration Velocity Integration + Accelerometer Triad + Attitude Computation Inertial Corrections + Gyrometer Triad IRU Gravity Compensation + - computer September 2014 2. Illustration of IRS Implementation for Aircraft - 74 - IRS mechanisation in the NED navigation frame Local gravity acceleration modeling – The local gravity force vector is the vector to which a ‘plump bob’ would align itself when held above the Earth. Ignoring any gravity anomaly, it will be assumed aligned with the vertical: ๐๐ ๐ ๐ = 0,0, ๐๐ ๐, โ ωIE Earth surface ( WGS-84 ) n Fe g ๐ gl R λ – It is sum of the mass attraction force and the centripetal acceleration: ๐๐ ๐ ๐ โ ๐๐ ๐ + ๐น๐๐ ๐ with ๐น๐๐ ๐ = −๐๐ผ๐ธ × ๐๐ผ๐ธ × ๐ ๐ – Gravity model example: ๐๐ ๐, โ = ๐๐ ∗ ๐ ๐, 0 . ๐+โ 2 ∗ where ๐๐ ๐, 0 = e ๐บ0 1 + ๐sin2 ๐ 1 − ๐ 2 sin2 ๐ September 2014 2. Illustration of IRS Implementation for Aircraft - 75 - IRS mechanisation in the NED navigation frame Mobile velocity estimate The component form of the inertial navigation dynamic equation in the N-frame is ๏ฆ ๏ถ ๏ง ๏ท ๏ฉ ๏น v ๏ง ๏ท e ๏ช R ๏ซh ๏บ ~ ๏ง ๏ท ๏ฉ fx ๏น N 0 ๏น ๏ cos ๏ฌ ๏ฉv๏ฆn ๏น ๏ฉ ๏ฉ ๏น ๏ฉ vn ๏น ๏ช ๏บ E ๏ง ๏ท ๏ช~ ๏บ vn ๏ช v๏ฆ ๏บ ๏ฝ R ๏บ ๏ท ๏ด ๏ชv ๏บ ๏บ ๏ซ 2๏ช fy๏บ ๏ซ ๏ช 0 ๏บ ๏ญ ๏ง๏ช ๏ญ 0 e m 2 N ๏ช ๏ช ๏บ ๏ช ๏บ ๏ช ๏บ๏ท ๏ช e ๏บ ๏ช RE ๏ซ h ๏บ ๏ง ~ ๏ช fz ๏บ ๏ช๏ซ v๏ฆv ๏บ๏ป ๏ช๏ซ g l ๏จ๏ฌ , h ๏ฉ๏บ๏ป ๏ช๏ซ ๏ญ ๏ E sin ๏ฌ ๏บ๏ป ๏ท ๏ช๏ซ vv ๏บ๏ป ๏บ ๏ง ๏ช ve ๏ซ ๏ปm ๏ป ๏ป ๏ฑ๏ด๏ฒ๏ด ๏ณ ๏ฑ๏ด๏ฒ๏ด๏ณ tan ๏ฌ ๏บ N N N ๏ง๏ช ๏ท local gravity force v๏ฆmE vmE ๏ท IE R ๏ซ h ๏ช ๏บ N ๏ง ๏ซ๏ฑ๏ด ๏ท ๏ด๏ฒ๏ด๏ด ๏ณ๏ป ๏ง ๏ท N Errors due to IMU NI ๏จ๏ฑ๏ด๏ด๏ท๏ด ๏ธ ๏ด๏ด ๏ด๏ด๏ด๏ด๏ฒ๏ด๏ด๏ด๏ด๏ด ๏ณ mesurement errors Errors due to the gravity Coriolis acceleration in ๏จ N ๏ฉ model uncertainties In the E-frame, the aircraft location is characterized by its latitude λ, its East longitude ฯ, and its height h above the reference ellipsoid. Those time differentiate ๏ฉ 1 ๏น equations are nonThey verify: 0 0 ๏ช ๏ฉ๏ฌ๏ฆ ๏น ๏ช R N ๏ซ h ๏ช ๏บ ๏ช ๏ช๏ช๏ฆ ๏บ ๏ฝ ๏ช 0 ๏ช h๏ฆ ๏บ ๏ช ๏ซ๏ป๏ป 0 ๏ช ๏ฆ mE p ๏ซ 1 ๏จRE ๏ซ h ๏ฉ ๏ cos ๏ฌ 0 ๏บ ๏บ ๏ฉv n ๏น 0 ๏บ ๏ ๏ช๏ช v e ๏บ๏บ ๏บ ๏ช v v ๏บ๏ป ๏ญ 1๏บ ๏ซ๏ป ๏บ vmE N ๏ป linear: they are solved thanks to numerical means that introduce additional errors onto the velocity estimate. September 2014 2. Illustration of IRS Implementation for Aircraft - 76 - Simulation results -3 2 Horizontal path ๏ค ๏ฌ [rad] 0.9 0 -2 -4 0.85 reference IRS 0 4 6 x 10 0.01 Longitude [rad] Vertical path 0.02 0.03 0.04 ๏ค ๏ฆ [rad] -0.01 2000 3000 4000 5000 6000 5000 6000 4000 5000 6000 4000 5000 6000 4000 5000 6000 5000 6000 0.01 0 -0.01 0 1000 2000 3000 4000 Altitude error estimate 20 reference IRS [m] 4 0 -20 2 0 1000 2000 -3 1000 2000 3000 Time [s] 4000 5000 6000 2 ๏ฆ [rad] 0 0 1000 0.02 0.8 0.75 -0.02 0 Longitude error estimate ๏ค h [m] Latitude [rad] 0.95 Latitude error estimate x 10 Roll error estimate x 10 0 -2 -4 0 1000 2000 -3 2 ๏ฑ [rad] – We observe an important drift in the horizontal position error estimate 0 1000 2000 3000 Heading error estimate 0.02 ๏น [rad] ๏ As expected, the vertical error remains bounded 3000 Pitch error estimate x 10 0 -2 ๏ Several kilometers 3000 Time [s] 0.01 0 -0.01 0 1000 2000 3000 temps(s) 4000 September 2014 - 77 - Part 3 GPS/IRS Hybridization 1. 2. 3. 4. 5. 6. Principle and schemes Hybridation architectures Overview of Kalman filtering Example of GPS/IRS tight coupling using KF Integrity Monitoring Conclusion September 2014 3. GPS/IRS Hybridization - 78 - Principle and schemes Why coupling? ๏ Complementary performance of the two systems. System or means Strong points Weak points type of information GPS, GLONASS Long-term accuracy, worldwide absolute positioning Integrity, continuity, RF interference 3D position and velocity Time INS Short term accuracy, immunity from perturbations Error drift, cost, relative positioning 3D position and velocity Attitude calibration (re) alignment GNSS INS (re) acquisition resistance from RF interference, jamming integrity monitoring availability continuity aiding September 2014 3. GPS/IRS Hybridization Principle and schemes Fix point at Toulouse location during 30 min with σGPS=12.5m and navigation-grade IRS - 79 - September 2014 3. GPS/IRS Hybridization - 80 - Hybridation architectures GPS/IRS loose coupling = Position-domain hybridization Signal preprocessing and sampling PRN Correlators and tracking loops Navigation computer GPS navigation solution PR1 GPS receiver We assume GPS stands for the true location. GPS position is used to help estimating inertial errors and correct IRS navigation solution. There is redundancy of output navigation solutions. But, need for at least 4 GPS measurements and the distribution of GPS navigation solution needs to be well known. Inertial Sensor measurement pre-processing Navigation and integration algorithm Hybridized navigation solution Integration process IRS navigation solution Accelerometers IRS computer Gyrometers IRS platform Figure extracted from GEBRE-EGZIABHER (D.) – What is the difference between ‘loose’, ‘tight’, ‘ultra-tight’ and ‘deep’ integration strategies for INS and GNSS ?, InsideGNSS, Jan-Feb 2007, pages 28-33 September 2014 3. GPS/IRS Hybridization - 81 - Hybridation architectures Forward configuration Inertial Reference System IRS External aiding source GPS ๐ฅ๐บ๐๐ผ๐ ๐ = ๐ฅ๐ผ๐ ๐ − ๐ฟ ๐ฅ๐ผ๐ ๐ + + - IRS errors estimate + - Kalman Filter Forward configuration Only a correction is applied to the IRS estimates. No risk of propagating error modes from one sensor to the other. But, the KF state propagation model tends to become less and less accurate with time, due to IRS drift. ๏จ Inertial error estimate is getting less and less accurate, whatever the GPS measurements. September 2014 3. GPS/IRS Hybridization - 82 - Hybridation architectures Backward configuration Inertial Reference System IRS IRS errors estimate External aiding source GPS + - Kalman Filter Backward configuration IRS calibration is done using the KF state estimates. The KF state model is far less sensitive to IRS drift: this improves the inertial error estimate. But, risk of propagating error modes from one sensor to the other. But, the KF state propagation model tends to become less and less accurate with time, due to IRS drift. ๏จ Need to detect failures, especially GPS ones, as soon as they occur. September 2014 3. GPS/IRS Hybridization - 83 - Hybridation architectures GPS/IRS tight coupling = range-domain hybridization Signal preprocessing and sampling PRN Correlators and tracking loops PR1 GPS receiver Inertial Sensor measurement pre-processing Navigation and integration algorithm Hybridized navigation solution Integration process Accelerometers Gyrometers IRS platform We assume GPS stands for the true location. GPS measurements are used to help estimating inertial errors and correct IRS navigation solution. Even with less than 4 satellites, GPS will improve the hybridized solution. ๏จ Inertia provides the state propagation model; GPS the observation one. Figure extracted from GEBRE-EGZIABHER (D.) – What is the difference between ‘loose’, ‘tight’, ‘ultra-tight’ and ‘deep’ integration strategies for INS and GNSS ?, InsideGNSS, Jan-Feb 2007, pages 28-33 September 2014 3. GPS/IRS Hybridization - 84 - Hybridation architectures Architecture Loose Advantages • Provides stand-alone GPS and INS solutions • Simpler implementation Drawbacks Comment • Requires 4 satellites • Requires GPS position error distribution • Integrity monitoring = RAIM (GPS only) • Works with <4 satellites Tight • Requires access to GPS • KF filter better tuned thanks pseudoranges to better error modelisation measurements of pseudoranges • More complex • Monitoring functions uses implementation INS measurements • Widely used in current systems • Integrity monitoring = AAIM (hybridized) Forward • INS drift accumulates • Requires higher grade of INS sensors Backward • Possible propagation of error modes between INS and GPS • Requires GNSS fault detection functions September 2014 3. GPS/IRS Hybridization Overview on Kalman Filtering (1/3) Underlying dynamic system model (w/o control-input) ๐ฅ ๐ =๐น ๐ ๐ฅ ๐−1 +๐ฃ ๐ ๐ฆ ๐ =๐ป ๐ ๐ฅ ๐ +๐ค ๐ • • ๐ฅ is the state to estimate, ๐ฅ ๐ ∈ โ๐ ๐ฆ is the measurement, ๐ฆ ๐ ∈ โ๐ • ๐ฃ is the process noise which is assumed to be drawn from a centered multivariate normal distribution, ๐ฃ ๐ ~๐ 0, ๐ ๐ ๐ค is the observation noise which is also assumed to be drawn from a centered multivariate normal distribution, ๐ค ๐ ~๐ 0, ๐ ๐ • • • ๐น is the ๐ × ๐ state transition matrix ๐ป is the ๐ × ๐ observation matrix - 85 - September 2014 3. GPS/IRS Hybridization Overview on Kalman Filtering (2/3) Assumptions • • Discrete-time, linear and Gaussian problem The initial state, and the noise vectors at each step {๐ฅ 0 , ๐ค 1 , … , ๐ค ๐ , ๐ฃ 1 … ๐ฃ ๐ } are all assumed to be mutually independent Solution • KF is a recursive estimator that provides the Minimum Mean Squared Error estimator from the past noisy measurements ๏ At each instant ๐, ๐ฅ ๐ is defined as ๐ฅ ๐|๐ = ๐ธ ๐ฅ ๐ |๐ฆ 0 … ๐ฆ ๐ • The state variance is defined as ๐ ๐ = ๐ ๐|๐ = ๐๐๐ฃ ๐ฅ ๐ |๐ฆ 0 … ๐ฆ ๐ ๏ Its stands for the estimation error variance: ๐๐๐ฃ ๐ฅ ๐ − ๐ฅ ๐ |๐ฆ 0 … ๐ฆ ๐ - 86 - September 2014 3. GPS/IRS Hybridization - 87 - Overview on Kalman Filtering (3/3) Implementation 1. Use of the state transition model 3. Compute optimal KF gain (trade-off between the trust in the state propagation model and the trust in the measurements) 4. Correct the predicted state 2. Compute the innovation (measurement residual) September 2014 3. GPS/IRS Hybridization Example of GPS/IRS tight coupling using KF (1/7) Definition of the inertial errors The error terms are defined as the difference between the true value and the measured or estimated one: Parameter State Model Position error ๏คp Dynamics navigation equation Velocity error ๏คv Dynamics navigation equation A priori knowledge of the accelerometric noises Gravity model uncertainty Attitude error ๏ค๏ฒ Attitude evolution equation A priori knowledge of the gyrorometric noises Accelerometer error ๏คf Assumed stochastics processes to model biases Gyrometer error ๏ค๏ท Assumed stochastics processes to model drifts - 88 - September 2014 3. GPS/IRS Hybridization - 89 - Example of GPS/IRS tight coupling using KF (2/7) Propagation model of the IRS navigation errors The general inertial error equations in the N-frame are the following Position • The position differential equation is in form of ๐ = ๐1 ๐, ๐ฃ • The position error propagation equation is thus ๐ฟ ๐ = ๐1 ๐, ๐ฃ − ๐1 ๐๐ผ๐ ๐ , ๐ฃ๐ผ๐ ๐ = ๐ป๐1 ๐, ๐ฃ . ๐ฟ๐, ๐ฟ๐ฃ or ๐ฟ ๐ = ๐ฟ๐ฃ + ๐๐ธ๐ ๐ × ๐ฟ๐ ๐ Velocity • The velocity differential equation is in form of ๐ฃ = ๐2 ๐, ๐ฃ, ๐, ๐ • The velocity error propagation equation is thus ๐ฟ ๐ฃ = ๐2 ๐, ๐ฃ, ๐, ๐ − ๐2 ๐๐ผ๐ ๐ , ๐ฃ๐ผ๐ ๐ , ๐๐ผ๐ ๐ , ๐๐ผ๐๐ = ๐ป๐2 ๐, ๐ฃ, ๐, ๐ . ๐ฟ๐, ๐ฟ๐ฃ, ๐ฟ๐, ๐ฟ๐ or (with simplifications) ๐ ๐ฟ ๐ฃ = ๐ ๐2๐ ๐ฟ๐ + ๐๐ผ๐ ๐ + ๐๐ผ๐ธ ๐ × ๐ฟ๐ฃ + ๐ฟ๐ × ๐๐๐ผ + ๐๐๐๐๐ฃ๐๐ก๐ฆ Attitude • The attitude error propagation equation is (with simplifications) ๐ฟ ๐ = ๐ ๐2๐ ๐ฟ๐ + ๐ × ๐ฟ๐ ๐ September 2014 3. GPS/IRS Hybridization - 90 - Example of GPS/IRS tight coupling using KF (3/7) Corrected code measurement general model – Corrected code pseudorange measurement general model (satellite ๐, time ๐) is ๐๐ ๐ = ๐๐๐ ๐ ๐ − ๐๐ข ๐ ๐๐ ๐ = where, ๐ฅ๐ ๐ − ๐ฅ ๐ 2 ๐๐ข ๐ = ๐ฅ ๐ , ๐ฆ ๐ , ๐ง ๐ + ๐๐ข ๐ + ๐ ๐ ๐ + ๐ฆ๐ ๐ − ๐ฆ ๐ 2 + ๐ง๐ ๐ − ๐ง ๐ 2 + ๐๐ข ๐ + ๐ ๐ ๐ is the 3D user location to estimate ๐๐ข ๐ is the receiver clock bias to estimate, as well ๐๐๐ ๐ ๐ = ๐ฅ ๐ ๐ , ๐ฆ ๐ ๐ , ๐ง ๐ ๐ is the 3D SV ๐ location ๐ ๐ ๐ stands for the pseudorange residual error – Let ๐ ๐ = ๐๐ข ๐ , ๐๐ข ๐ ๐ be the 4D user location ๏ The true pseudorange mathematical model is โ๐ ๐ ๐ = ๐ฅ๐ ๐ − ๐ฅ ๐ 2 + ๐ฆ๐ ๐ − ๐ฆ ๐ 2 + ๐ง๐ ๐ − ๐ง ๐ 2 + ๐๐ข ๐ ๐ ๐ ๐๐ข ๐ – Let ๐ ๐ be the user location estimate ๏ The SV ๐ predicted pseudorange measurement around ๐ ๐ is ๐๐ ๐ = โ๐ ๐ ๐ September 2014 3. GPS/IRS Hybridization - 91 - Example of GPS/IRS tight coupling using KF (4/7) General principle ๐ฅ ๐ก = ๐ ๐ฅ, ๐ก + ๐ข ๐ก ๐ฅ ๐ =๐น ๐ ๐ฅ ๐−1 +๐ข ๐ โน ๐ฆ ๐ =๐ป ๐ ๐ฅ ๐ +๐ฃ ๐ ๐ฆ ๐ก = โ ๐ฅ, ๐ก + ๐ฃ ๐ก Solved with linearized Kalman Filtering (eg., see [Farrell, Barth, 1999] for details) – The state vector ๐ฅ๐ components are the errors - inertial errors (+ GPS Rx clock bias). • The propagation model is defined from the inertial errors propagation models. – The measurement vector ๐ฆ๐ consists of the difference between the measured pseudoranges at the output of the GPS receiver and, the range between the satellites location and the user location as estimated by the IRS. • The measurement must be an observation of the quantitities to estimate: here, the inertial errors. • The observation model is defined from the GPS measurements model. ๏ Using this measurement vector, the integration process calculates corrections to the inertial solutions, makes sure of the integrity of the corrected solution and computes the integrity level. September 2014 3. GPS/IRS Hybridization - 92 - Example of GPS/IRS tight coupling using KF (5/7) Implementation ๐ท๐ ๐ Observation model ๐ ๐ ๐ = ๐ท ๐ − ๐ท๐ ๐ ๐ GPS Receiver ๐ ๐ ๐ฏ ๐ ๐น ๐ Kalman filter implementation Prediction ๐ ๐|๐ − ๐ = ๐ญ ๐ − ๐ ๐ ๐ − ๐|๐ − ๐ ๐ฟ๐บ๐ฝ ๐ ๐ Compute ๐บ ๐|๐ − ๐ = ๐ญ ๐ − ๐ ๐บ ๐ − ๐|๐ − ๐ ๐ญ๐ป ๐ − ๐ + ๐ธ ๐ ๐ท๐ ๐ = ๐ ๐ ๐๐ฐ๐น๐บ ๐ Innovation ๐๐ฐ๐น๐บ ๐ ๐๐ฎ๐ท๐ฐ๐น๐บ ๐ = ๐ฐ ๐ = ๐ ๐ − ๐ฏ ๐ ๐ ๐|๐ − ๐ ๐๐ฐ๐น๐บ ๐ + ๐ ๐|๐ Update IRS output Inertial Reference System State model Error propagation model ๐ ๐|๐ ๐บ ๐|๐ ๐ฝ ๐ = ๐ฏ ๐ ๐บ ๐|๐ − ๐ ๐ฏ๐ป ๐ − ๐ + ๐น ๐ ๐ ๐ − ๐|๐ − ๐ ๐บ ๐ − ๐|๐ − ๐ ๐ญ ๐−๐ ๐ธ ๐ Estimation ๐ฒ ๐ = ๐บ ๐|๐ − ๐ ๐ฏ๐ป ๐ − ๐ โ ๐ฝ ๐ −๐ ๐ ๐|๐ = ๐ ๐|๐ − ๐ + ๐ฒ ๐ ๐ฐ ๐ ๐บ ๐|๐ = ๐ − ๐ฒ ๐ ๐ฏ ๐ โ ๐บ ๐|๐ − ๐ September 2014 3. GPS/IRS Hybridization - 93 - Example of GPS/IRS tight coupling using KF (6/7) Simulation hypotheses Baro-IRS/GPS hybridization with C/A performance assumptions for the inertial sensors, based on the following simulated flight path. Roulis 0.4 0.9 0.2 [rad] Latitude [rad] Trajectoire de reference dans le plan horizontal 0.95 0 0.85 -0.2 0.8 0 1000 2000 3000 4000 5000 6000 4000 5000 6000 4000 5000 6000 Tangage 0.5 0 0.005 0.015 0.02 0.025 0.03 Longitude [rad] Trajectoire de reference sur l'axe vertical 4 6 x 10 0.01 0.035 0.04 [rad] 0.75 -0.005 0 0 1000 2000 3000 Cap 5 [rad] Altitude [m] -0.5 4 2 0 0 1000 2000 3000 Temps [s] 4000 5000 6000 0 -5 0 1000 2000 3000 Temps GPS measurements assumptions • Time-correlated L1 C/A code pseudorange measurements (Appendix R to D0 229D) • All GPS observations are independent each one from the other September 2014 3. GPS/IRS Hybridization - 94 - Example of GPS/IRS tight coupling using KF (7/7) Simulation results -3 ๏ค ๏ฌ [rad] 2 Horizontal path 0.95 -2 -4 0 1000 2000 3000 4000 5000 6000 5000 6000 5000 6000 5000 6000 5000 6000 Longitude estimate error reference baro-IRS GPIRS 0.85 0.8 0.02 ๏ค ๏ฆ [rad] Latitude [rad] baro-IRS GPIRS 0 0.9 0.01 0 -0.01 -0.01 0 4 x 10 0.01 Longitude [rad] Vertical path 0.02 0.03 0.04 [m] 1000 2000 3000 4000 20 0 -20 reference baro-IRS GPIRS 4 0 Altitude error estimate ๏ค h [m] 0.75 -0.02 6 Latitude estimate error x 10 0 1000 -7 5 2 2000 3000 Time [s] 4000 GPIRS latitude estimate error x 10 0 0 1000 2000 3000 Time [s] 4000 5000 6000 ๏ค ๏ฌ [rad] 0 -5 -10 -15 The important drift formerly observed in the horizontal plane has been corrected. ๏ On the vertical axis, the position error uncertainty is reduced as well as its frequency 1000 -7 5 2000 3000 4000 GPIRS longitude estimate error x 10 0 ๏ค ๏ฆ [rad] ๏ At the end of the path the error is only of few meters 0 -5 -10 -15 0 1000 2000 3000 Time [s] 4000 September 2014 3. GPS/IRS Hybridization Integrity Monitoring • Integrity monitoring consists of 2 functions: – Detection and mitigation of faults • Fault Detection • Fault Detection and Isolation • Fault Detection and Exclusion : alert to the user : … + uncontaminated nav solution : … + the recovered nav solution is fault-free – Solution protection: determination of whether a navigation solution is safe to use (๏จxPL) • Hybridation brings advantages to both functions. - 95 - September 2014 3. GPS/IRS Hybridization Integrity Monitoring – INS Faults • Equipment failure: not more data, wrong biases, etc… • Threat model less known – GNSS • Satellite faults: – wrong orbit, onboard clock anomaly, wrong emitted power, etc. – In principle, the GPS Control Segment provides monitoring function, but with performances incompatible with ICAO requirements (eg time to alert = 8s, Alert Limit may be several km) • Unusual atmospheric propagation – High Ionosphere gradient, Ionospheric scintillation, etc. • Local tracking channel failure – NLOS reception, strong multipath, low C/No, cycle slip, etc. • General equipment failure – Antenna problem, local oscillator problem, jamming – In hybridized systems, we try to detect the GNSS failures assuming the INS is not affected by any. - 96 - September 2014 3. GPS/IRS Hybridization Integrity Monitoring Fault Detection Examples • 2 examples of FD – Kalman Filter Innovation monitoring • Innovation = difference between predicted observations and actual observations • Allow to check the consistency between the KF states and the measurements – Direct Consistency Checks (Solution Separation) • Use of redundancy of measurements to identify a faulty measurement – M measurements required ๏จ M+1 for FD ๏จ M+2 for FDE – M=4 for GNSS, M=3+3 for INS (accelero + gyro) • Comparison of consistency between solutions computed with different of sets of measurements. – Principle: A fault can be detected by hypothesis testing (detection theory) • A chosen test statistic is supposed to have a known distribution in fault-free mode • Using a threshold on the test statistic, a fault can be detected - 97 - September 2014 3. GPS/IRS Hybridization - 98 - Integrity Monitoring Fault Detection and Isolation example – Direct Consistency Checks for GPS only • N positions are computed using N-1 measurements • For each position, the pseudo-ranges between the estimated position and the satellite are computed. • Then, the pseudo-range residual is computed to compare the computed pseudo-ranges to the observed pseudo-ranges • A hypothesis test is applied to the residual to detect the presence of a faulty measurement. • FDI is directly implemented by choosing the fault-free combination for the position solution 1,2,4,5 1,3,4,5 1,2,4,5 True 1,2,3,5 1,3,4,5 2,3,4,5 2,3,4,5 True 1,2,3,5 1,2,3,4 1,2,3,4 No fault Fault on sat 1 September 2014 3. GPS/IRS Hybridization Integrity Monitoring Fault Detection and Isolation example – Parallel-filter INS/GNSS integrity monitoring • Does not monitor INS integrity (no redundant inertial sensor) • N Kalman Filter are run in parallel, each excluding one GNSS measurement. • A fault detection algorithm is applied to each position solution (range check, KF innovation monitoring, residual method) (tight architecture) OR • An hypothesis test is done in the position-domain (solution-separation method) (loose or tight architecture) • FDI is directly implemented by choosing the fault-free combination for the position solution - 99 - September 2014 3. GPS/IRS Hybridization - 100 - INTEGRITY MONITORING PARALLEL-FILTER INS/GNSS INTEGRITY MONITORING EXAMPLE IMU Inertial navigation equations Hybridation (tight) Main integration algorithm - Ch 1 Antenna and frontend Ch 1 - Ch 2 Ch 2 - Ch 3 Ch 3 - Ch 4 Ch 4 Integrity Monitoring max separation solution Sensor & measurements solution-separation method: – Each sub-filter solution is compared to the main integration solution. – A position residual is computed and used as test statistic for hypothesis testing output September 2014 3. GPS/IRS Hybridization Conclusion – Civil Aviation has its specifities • ICAO Requirements: accuracy, integrity monitoring • Equipment: – GNSS: GPS L1 C/A, in the future: multi-constellation, multi-frequency? – Inertial sensor: navigation grade – Inertial Navigation can provide position • with an error drift < 2 NM/h • To obtain this, navigation-grade sensors are required (~100 k$) – Hybridization algorithms are based on Kalman filtering • with different possible architecture: loose, tight, (deep) – Hybridization between GNSS and INS allows: • • • • Increased accuracy Increased continuity Increased integrity performances Increased availability - 101 - September 2014 Additional slides - 102 - Reference documents [Farrell, Barth, 1999] J. A. Farrell, M. Barth, The global positioning system and inertial navigation, Mc Graw Hill, 1999 [Kayton and Fried, 1996] M. Kayton, W.R. Fried, Avionics navigation Systems [Kubrak, 2007] D. Kubrak, Etude de l’hybridation d’un récepteur GPS avec des capteurs bas coûts pour la navigation personnelle en milieu urbain, Ph.D. Thesis report, ENST, 2007 [Palmqvist, 1997] J. Palmqvist, On integrity monitoring of integrated navigation systems, Thesis No. 600, Linköping Studies in Science and Technology, 1997 [Titterton, Weston, 2004] D. Titterton, J. Weston, Strapdown Inertial navigation Technology, 2nd edition IEE Radar, Sonar and navigation series 17 September 2014 Additional slides - 103 - GPS IRS Kalman Filter Fix point at Toulouse location with navigation-grade IRS and σGPS=12.5m GPS/IRS September 2014 Additional slides - 104 - How to get Rm2N from the gyro. measurements? m m ๏ท Nm ๏ฝ ๏ท NI ๏ญ – In the direct method, we solve ๐ ๐2๐ = ๐ ๐2๐ . ๐๐๐ • ๐ m ๐๐๐ ๐ is the skew-symetric matrix associated to ๐๐๐ ๐ , WNm We have to solve 9 equations with 9 unknows – Using a quaternion vector, we solve ๏ฉ 0 ๏ช m WNm ๏ฝ ๏ช ๏ทz ๏ช๏ญ ๏ท y ๏ซ ๏ญ ๏ทz 0 ๏ ๏ is the quaternion vector associated to ๏ท • ๏ท N Nm ๏ทx ๏ทy ๏น ๏บ ๏ญ ๏ทx ๏บ 0 ๏บ๏ป m Nm ๏ ๏ ๏ m ๏ท mI ๏ป from gyrometer ๏ฉ 0 ๏ญ ๏ทz ๏ท y ๏น ๏ช ๏บ ๏ฝ ๏ช ๏ทz 0 ๏ญ ๏ทx ๏บ ๏ช๏ญ ๏ท y ๏ท x 0 ๏บ๏ป ๏ซ m , ๏ท Nm ๏ฝ 0 ๏ท x ๏ท y ๏ท z ๏T • qis the unknown quaternion vector associated to the single rotation describing the rotation of (m) in (N) ๏ฉa ๏น ๏ฉ ๏น ๏ผ ๏ก ๏ฆ ๏ถ ๏ช b ๏บ ๏ช cos๏ง ๏ท ๏ฝreal part ๏บ ๏ฒ 2๏ธ ๏พ ๏จ ๏ช ๏บ ๏ช ๏บ, E rotation axis guiding unit vector, ๏ก rotation angle q๏ฝ ๏ฝ ๏ชc ๏บ ๏ช ๏ฆ ๏ก ๏ถ ๏ฒ ๏ผ ๏บ ๏ช ๏บ ๏ชsin ๏ง ๏ท ๏ E ๏ฝimaginary part ๏บ ๏พ ๏ป ๏ซd ๏ป ๏ซ ๏จ 2 ๏ธ • Rm 2 N 2๏จbc ๏ญ ad ๏ฉ 2๏จbd ๏ซ ac ๏ฉ ๏น ๏ฉa ² ๏ซ b ² ๏ญ c ² ๏ญ d ² ๏ฝ ๏ช๏ช 2๏จbc ๏ซ ad ๏ฉ a ² ๏ญ b² ๏ซ c² ๏ญ d ² 2๏จcd ๏ญ ab ๏ฉ ๏บ๏บ ๏ช๏ซ 2๏จbd ๏ญ ac ๏ฉ 2๏จcd ๏ซ ab ๏ฉ a ² ๏ญ b ² ๏ญ c ² ๏ซ d ² ๏บ๏ป We have to solve 4 equations with 4 unknows; more stable. September 2014 Additional slides - 105 - Further on frames definition Frame Inertial Terrestrial Appellation s Spacestable or Spacefixed frame ECEF Conventiona l Terrestrial System Frame origin Earth center of mass Ox Oy Pointed along the vernal equinox Oz Mean position 90° East / Ox of the spinning axis on the equatorial plane of the Earth Intersection of the Mean position equatorial plane 90° East / Ox of the spinning axis and the Greenwich on the equatorial plane of the Earth reference plane Local-level Geographic Navigation frame Body Body frame Platform Platform axes platform Axes of Computer Computing frame Center of gravity Supposed axes of the platform True True axes True axes of the platform vehicle center of mass Geographic North direction Roll axis, pointed to forward East direction Pitch axis, pointed to the right wing the accelerometers Normal to the reference ellipsoid Yaw axis, pointed to down 1, 2 and 3 Observations - Newton laws - Computing burden - Stabilized INS - Earth -fixed - Computing burden - horizontal navigation - Stabilized on the north - Vehicle fixed - Strapdown INS / - Frame providing the solution / September 2014 Additional slides - 106 - Inertial sensors error sources and classical models Error source Error type Gyroscope Gravity Observations due to the accelerometers mounting on the platform Bias Specific force error supposed as linear in the range of Scale factor measured specific forces Other (nonlinearities...) / due to the gyroscope mounting on Alignment error the platform One of the most g-insensitive drift Random walk significant errors Random proportional to the specific force g-sensitive drift constant (unbalance-type error) Random induced by g²- sensitive drift constant anisoelastic torques Random supposed as linear in the range of Scale factor constant accelero measured velocities The gravitation potential depends Deflections 1st order on tides and gravity anomaly Markov process (moon and sun effects ...) Alignment error Accelerometer Adopted model Random constant Random walk Random constant ? Random constant