Qt7c403919.Pdf

UC Irvine UC Irvine Previously Published Works

Title New-age satellite-based navigation - STAN: simultaneous tracking and navigation with LEO satellite signals

Permalink https://escholarship.org/uc/item/7c403919

Journal Inside GNSS Magazine, 14(4)

Authors Kassas, Zak Morales, Joshua Khalife, Joe

Publication Date 2019

Peer reviewed

eScholarship.org Powered by the California Digital Library University of California July/August 2019 | Engineering Solutions from the Global Navigation Satellite System Community | www.insidegnss.com

LEO: Simultaneous Tracking and Navigation

WORKING PAPERS | GNSS Authentication WASHINGTON VIEW | Ligado Request SINGLE POINT POSITIONING | LiDAR-Aided GNSS STAN WITH LEO

System Number of satellites Frequency band New-Age Satellite-Based Navigation Orbcomm 36 VHF Globalstar 48 S and C STAN: Simultaneous Tracking and Iridium 66 L and Ka OneWeb 882 Ku and Ka Navigation with LEO Satellite Signals Boeing 2956 V and C SpaceX 11943 Ku, Ka, and V Today’s vehicular navigation systems couple global navigation satellite (e.g., due to unintentional interference or Samsung 4600 V system (GNSS) receivers with an inertial navigation system (INS). intentional jamming), or untrustworthy Table 1: Existing and future LEO constellations: number of satellite (e.g., due to malicious spoofng attacks and transmission bands. Low Earth orbit (LEO) satellite signals are a particularly attractive INS or system malfunctions), the naviga- FIGURE 2 Snapshot of the Starlink LEO constellation. aiding source in GNSS-challenged environments. Over the next few tion system relies on unaided inertial Table 1 summarizes the number of satellites and the trans- years, LEO satellites will be abundantly available at favorable geometric measurement unit (IMU) data, in which mission band of each constellation. case the errors accumulate and eventu- Figure 2 depicts a snapshot of the upcoming Starlink con- conﬁgurations and will transmit in several frequency bands, making ally diverge, compromising the vehicle’s stellation, while Figure 3 is a heat map of the number of visible them an accurate and robust navigation source. This article presents a efcient and safe operation. Starlink LEO satellites above an elevation mask of 5 degrees. framework that enables a navigating vehicle to aid its INS with pseudorange Signals of opportunity are PNT Figure 5 is a heat map showing the position dilution of pre- sources that could be used in GNSS- cision (PDOP) for the Starlink constellation, while Figure 5 and Doppler measurements drawn from LEO satellite signals when challenged environments (See Merry is a heat map showing the logarithm of the Doppler position GNSS signals become unusable, while simultaneously tracking the LEO et alia, and Kassas, 2013, in Additional dilution of precision (DPDOP). satellites. This simultaneous tracking and navigation (STAN) framework Resources). Tese signals include AM/ Figure 2 through Figure 5 together with Table 1 demon- FM radio, cellular, digital television, and strate the potential of using LEO satellite signals for PNT and is demonstrated in realistic simulation environments and experimentally low Earth orbit (LEO) satellites (several imply that the commercial space industry is inadvertently creat- on a ground vehicle and on an unmanned aerial vehicle (UAV), showing papers listed in Additional Resources pro- ing new PNT sources, which could be utilized by future vehicles FIGURE 3 Heat map showing the number of visible Starlink LEO the potential of achieving meter-level-accurate navigation. vide further details). Signals of opportu- to make the vehicle’s PNT system more resilient and accurate. satellites above a 5-degree elevation mask. nity have been demonstrated to yield a For example, a Tesla connected to Starlink satellites could standalone meter-level-accurate naviga- dually provide a passenger with internet access, as designed, ZAHER (ZAK) M. KASSAS esilient and accurate positioning, tion solution on ground vehicles and a while also enabling the vehicle to navigate in GNSS-challenged JOSHUA J. MORALES navigation, and timing (PNT) is of centimeter-level-accurate navigation environments. JOE J. KHALIFE paramount importance in safety solution on aerial vehicles. Moreover, Tere are several challenges that need to be addressed to AUTONOMOUS SYSTEMS PERCEPTION, Rcritical cyber-physical systems (CPS), such these signals have been used as an aid- exploit LEO satellites for navigation. First, their positions INTELLIGENCE, & NAVIGATION (ASPIN) as aviation and transportation. As these ing source for LiDAR and INS. and velocities must be known. Te position and velocity of LABORATORY CPS evolve towards becoming fully auton- LEO satellites are particularly attrac- any satellite may be parameterized by its Keplerian elements. UNIVERSITY OF CALIFORNIA, IRVINE omous, the requirements on their PNT tive aiding sources for an INS in GNSS- Tese elements are tracked, updated once daily, and made systems become more stringent than ever challenged environments for several publicly available by the North American Aerospace Defense before. With no human in-the-loop, an reasons. First, LEO satellites are around Command (NORAD) [see North American Aerospace Defense inaccurate PNT solution; or more dan- 20 times closer to Earth compared to Command, Additional Resources]. However, these elements are gerously, PNT system failure, could GNSS satellites that reside in medium dynamic and will deviate from their nominally available val- FIGURE 4 Heat map showing PDOP for the Starlink LEO constellation have intolerable consequences. Earth orbit (MEO), making LEO satel- ues due to several sources of perturbing forces, which include above a 5-degree elevation mask. Today’s vehicular navigation lites’ received signals signifcantly more non-uniform Earth gravitational feld, atmospheric drag, solar systems couple GNSS receiv- powerful. Second, LEO satellites orbit the radiation pressure, third-body gravitational forces (e.g., gravity ers with an inertial naviga- Earth at much faster rates compared to of the Moon and Sun), and general relativity (Vetter, Additional tions system (INS). By cou- GNSS satellites, making LEO satellites’ Resources). Tese deviations can cause errors in a propagat- pling both systems, one takes Doppler measurements attractive to ed satellite orbit as high as 3 kilometers if not accounted for advantage of the complemen- exploit. Tird, the recent announcements with corrections. Second, LEO satellites are not necessarily tary properties of the individ- by OneWeb, Boeing, SpaceX (Starlink), equipped with an atomic clock, nor are they precisely synchro- ual subsystems: the short-term Samsung, Kepler, Telesat, and LeoSat to nized. Subsequently, their clock error must be known along- accuracy and high data rates of provide broadband internet to the world side their position and velocities. In contrast to GNSS, where an INS and the long-term stabil- via satellites will collectively bring thou- corrections to the orbital elements and clock errors are peri- ity of a GNSS PNT solution to pro- sands of new LEO satellites into opera- odically transmitted to the receiver in the navigation message, vide periodic corrections. However, in tion, making their signals abundant such orbital element and clock corrections may not be available the inevitable event that GNSS signals and diverse in frequency and direction. for LEO satellites; in which case they must be estimated along FIGURE 5 Heat map showing log [DPDOP] for the Starlink LEO FIGURE 1 Existing and future become unreliable (e.g., in deep urban Figure 1 depicts a subset of existing and with the receiver’s states. Tird, ionospheric delay rates become 10 LEO satellite constellations. constellation above a 5-degree elevation mask. canyons or near dense foliage), unusable future LEO satellite constellations. signifcant for LEO satellites, particularly the ones transmit-

56 InsideGNSS JULY/AUGUST 2019 www.insidegnss.com www.insidegnss.com JULY/AUGUST 2019 InsideGNSS 57 STAN WITH LEO

(1)

where is the speed of light and is the carrier frequency. Te pseudorange from the m-th LEO satellite at time-step , which represents discrete-time at for an initial time and sampling time T, is modeled as FIGURE 6 SGP4 position and velocity errors. (2)

where , represents discrete-time at with being the true time-of-fight of the signal from the m-th FIGURE 9 Ionospheric delay rates (expressed in m/s) for 7 Orbcomm FIGURE 10 Residual errors showing the efect of (i) satellite position LEO satellite; and are the LEO receiver’s and m-th LEO satellites over a 100-minute period. Each color corresponds to a and velocity errors, (ii) clock errors, and (iii) ionospheric delay rates satellite’s 3-D position vectors, respectively; and are diferent Orbcomm LEO satellite. for 2 Orbcomm LEO satellites. the LEO receiver and the m-th LEO satellite transmitter clock biases, respectively; and are the ionospheric and C. CLOCK ERRORS delay rates, the residual error between the measured pseudor- FIGURE 7 Time evolution of 1-σ bounds of (a) clock bias and (b) clock tropospheric delays, respectively, afecting the m-th LEO sat- In contrast to GNSS, LEO satellite clocks are not tightly syn- ange rate and the pseudorange rate estimated from the satellite drift for a typical OCXO and a typical TCXO over a 10-minute period. ellite’s signal; and is the pseudorange measurement noise, chronized and the clock errors (bias and drif) are unknown position and velocity obtained from TLE fles and SGP4 are plot- which is modeled as a white Gaussian random sequence with to the receiver. Moreover, LEO satellites are not necessar- ted in Figure 10 for 2 Orbcomm satellites (FM 108 and FM 116). variance . Te pseudorange rate measurement from the ily equipped with high-quality atomic clocks. From what is m-th LEO satellite is given by known about the existing LEO constellations, LEO satellites STAN Framework are equipped with oven-controlled crystal oscillators (OCXOs). To exploit LEO satellite signals for navigation, their states must (3) Practically, the navigating receiver will be equipped with a be known. Unlike GNSS satellites that periodically transmit lower quality oscillator, e.g., a temperature-compensated crys- accurate information about their positions and clock errors, tal oscillator (TCXO). To visualize the magnitude of the clock such information about LEO satellites may be unavailable. Te errors in the satellite and receiver clocks, Figure 7 depicts the STAN framework addresses this by extracting pseudorange where and are the LEO receiver’s and m-th LEO satel- time evolution of the bound of the clock bias and drif of and Doppler measurements from LEO satellite to aid the vehi- FIGURE 8 (a) Skyplot showing the trajectory of an Orbcomm LEO lite’s 3-D velocity vectors, respectively; and are the a typical OCXO and a typical TCXO, obtained from the so- cle’s INS, while simultaneously tracking the LEO satellites. Te satellite (FM 109) and a GPS MEO satellite (PRN 32) over a 10-minute LEO receiver and the m-th LEO satellite transmitter clock called two-state clock model (Brown and Hwang, Additional STAN framework employs an extended Kalman flter (EKF) to period. (b) The elevation angle rate of FM 109 and PRN 32 over the Figure 7 10-minute trajectory. The elevation angle rate of the Orbcomm LEO drifs, respectively; and are the drifs of the Resources). It can be seen from that the satellite and simultaneously estimate the vehicle’s states with the LEO satel- satellite reaches as high as 60 times that of the GPS MEO satellite. ionospheric and tropospheric delays, respectively, afecting receiver clock bias and drif may become very signifcant; lites’ states. Figure 11 depicts the STAN framework. the m-th LEO satellite’s signal; and is the pseudorange rate therefore, they must be accounted for appropriately. ting in the very high frequency (VHF) band. measurement noise, which is modeled as a white Gaussian ran- Simulation Results Tis article presents a simultaneous tracking and naviga- dom sequence with variance . D. IONOSPHERIC DELAY ERRORS Tis section presents simulation results obtained with a real- tion (STAN) framework that addresses the aforementioned Most broadband LEO constellations reside above the iono- istic simulation environment demonstrating UAVs navigating challenges (for more, see 2 papers from Morales, et alia). Tis B. POSITION AND VELOCITY ERRORS sphere, which in turn will induce delays into their signals. via the LEO-aided INS STAN framework without GNSS sig- framework tracks the states of LEO satellites while simultane- One source of error that should be considered when navigat- Although LEO satellite signals propagate through the tropo- nals. Te frst subsection evaluates the achieved performance ously using pseudorange and Doppler measurements extracted ing with LEO satellite signals arises due to imperfect knowl- sphere, its efect is less signifcant compared to ionospheric from current LEO constellations (Globalstar, Orbcomm, and from their signals to aid the vehicle’s INS. Te performance of edge of the LEO satellites’ position and velocity. Tis is due to propagation. Te magnitude of the ionospheric delay rate is Iridium), while the second subsection evaluates the achieved the STAN framework is demonstrated in realistic simulation time-varying Keplerian elements caused by several perturbing (i) inversely proportional to the square of the carrier frequen- performance with an upcoming LEO constellation: Starlink. environments and experimentally on a ground vehicle and on accelerations acting on the satellite. Mean Keplerian elements cy and (ii) proportional to the rate of change of the obliquity an unmanned aerial vehicle (UAV), showing the potential of and perturbing acceleration parameters are contained in pub- factor, which is determined by the time evolution of the sat- A. UAV SIMULATION WITH THE GLOBALSTAR, ORBCOMM, achieving meter-level-accurate navigation. licly available two-line element (TLE) fle sets. Te informa- ellite’s elevation angle. Note that the ionospheric delay rates tion in these fles may be used to initialize a simplifed general also depend on the rate of change of the total electron con- Pseudorange, Doppler Measurement Model perturbations (SGP) model, which is specifcally designed to tent (TEC) at zenith, denoted by TECV. However, TECV var- Tis section describes the LEO satellite receiver pseudorange propagate a LEO satellite’s orbit. SGP propagators (e.g., SGP4) ies much slower than the satellite’s elevation angle; hence, its and Doppler measurement model and discusses the sources are optimized for speed by replacing complicated perturbing efect may be ignored. Te efect of ionospheric propagation of error in LEO-based positioning: (i) satellite position and acceleration models that require numerical integrations with is signifcant on LEO satellite signals since (i) the high speed velocity errors, (ii) satellite and receiver clock errors, and (iii) analytical expressions to propagate a satellite position from an of LEO satellites translates into very fast changing elevation ionospheric and tropospheric delay rate errors. epoch time to a specifed future time. Te tradeof is in satellite angles, as shown in Figure 8 and (ii) some of the existing LEO position accuracy: the SGP4 propagator has around 3 km in satellites transmit in the VHF band where the signals experi- A. PSEUDORANGE AND DOPPLER MEASUREMENT MODEL position error at epoch and the propagated orbit will continue ence very large delay rates. Te aforementioned factors result Te LEO receiver extracts pseudorange and Doppler frequen- to deviate from its true one until the TLE fles are updated the in large ionospheric delay rates, as shown in Figure 9 for 7 cy measurements from LEO satellite signals. A pseudorange following day. Figure 6 shows the accumulated position and Orbcomm satellites over a 100-minute period. rate measurement can be obtained from velocity error for an Orbcomm LEO satellite (FM 112). In order to visualize the efect of (i) the satellite position and FIGURE 11 LEO-aided INS STAN framework. velocity errors, (ii) the clock drif error, and (iii) the ionospheric

58 InsideGNSS JULY/AUGUST 2019 www.insidegnss.com www.insidegnss.com JULY/AUGUST 2019 InsideGNSS 59 STAN WITH LEO

(1)

58 InsideGNSS JULY/AUGUST 2019 www.insidegnss.com www.insidegnss.com JULY/AUGUST 2019 InsideGNSS 59 STAN WITH LEO

lated. The LEO satellite orbits corre- positions, which were produced by the Unaided INS LEO-aided INS STAN sponded to the Globalstar, Orbcomm, GPS receivers onboard the LEO satel- Final Error (m) 174.7 9.9 and Iridium constellations. The UAV lites. Figure 14 shows the trajectories made pseudorange and pseudorange of the simulated LEO satellites and the RMSE (m) 52.6 10.5 rate measurements to all 10 LEO satel- UAV along with the location at which Table 2: Simulation results with Globalstar, Orbcomm, and Iridium LEO satellites for a UAV lites throughout the entire trajectory. GPS signals were cut of (Ardito et alia). navigating 25 km in 200 seconds (GPS signals were cut off after the first 100 seconds). These results are after GPS cutoff. Te LEO satellites’ positions and veloci- To estimate the UAV’s trajectory, 2 ties were initialized using TLE fles and navigation frameworks were imple- SGP4 propagation. Figure 12 shows the mented to estimate the vehicle’s trajecto- LEO-aided INS STAN with periodically Unaided INS transmitted satellite positions trajectories of the simulated LEO satel- ry: (i) the LEO-aided INS STAN frame- lites and the UAV along with the loca- work and (ii) a traditional GPS-aided Final Error (m) 16,589.0 9.8 tion at which GPS signals were cut of. INS for comparative analysis. Each RMSE (m) 6,864.6 10.1 To estimate the UAV’s trajectory, 2 framework had access to GPS for only Table 3: Simulation results with Starlink LEO satellites for a UAV navigating 82 km in 600 seconds navigation frameworks were imple- the frst 100 seconds. Figure 15(a)-(b) (GPS signals were cut off after the first 100 seconds). These results are after GPS cutoff. mented: (i) the LEO-aided INS STAN illustrate the UAV’s true trajectory and FIGURE 14 UAV simulation environment with framework and (ii) a traditional GPS- those estimated by each of the 2 frame- (GESs), and the network control FIGURE 12 UAV simulation environment with the Starlink LEO constellation. (a) LEO aided INS for comparative analysis. works while Figure 15(c) illustrates the center (NCC). The GCC pro- the Globalstar, Orbcomm, and Iridium satellites’ trajectories. The elevation mask LEO constellations. (a) LEO satellites’ was set to 35 degrees. (b) UAV trajectory and Each framework had access to GPS simulated and estimated trajectories of vides switching capabilities to trajectories. (b) UAV trajectory and GPS GPS cut of location. Map data: Google Earth. for only the frst 100 seconds. Figure one of the LEO satellites, as well as the link mobile SCs with terrestrial- cutof location. Map data: Google Earth. 13(a)-(b) illustrate the UAV’s true tra- fnal 95-th percentile uncertainty ellip- based customer systems via stan- jectory and those estimated by each of soid (the axes denote the radial (ra) and dard communications modes. the 2 frameworks while Figure 13(c) along-track (at) directions). Table 3 GESs link the ground segment illustrates the simulated and estimated summarizes the fnal error and position with the space segment. GESs trajectories of one of the LEO satel- RMSE achieved by each framework afer mainly track and monitor satellites, as well as the fnal 95-th percentile GPS cutof. lites based on orbital informa- uncertainty ellipsoid (the axes denote tion from the GCC and transmit the radial (ra) and along-track (at) direc- EXPERIMENTAL DEMONSTRATIONS to and receive from satellites, the tions). Table 2 summarizes the fnal This section describes the existing GCC, or the NCC. Te NCC is error and position root mean squared Orbcomm LEO constellation and the responsible for managing the FIGURE 16 Orbcomm LEO satellite constellation. error (RMSE) achieved by each frame- LEO receiver. Ten, it demonstrates the work afer GPS cutof. performance of the LEO-aided INS STAN B. UAV SIMULATION WITH THE STARLINK framework on a ground vehicle and a LEO CONSTELLATION WITH PERIODICALLY UAV with real Orbcomm satellite signals. TRANSMITTED LEO SATELLITE POSITIONS A UAV was equipped with (i) a tactical- Orbcomm System Overview grade IMU and (ii) GPS and LEO satellite The Orbcomm system is a wide area receivers. Te UAV navigates over Santa two-way communication system that Monica, California, USA, for about 82 uses a constellation of LEO satellites to kilometers in 10 minutes, during which provide worldwide geographic coverage FIGURE 15 UAV simulation results with the it had access to GPS signals only for the for sending and receiving alphanumer- Starlink LEO constellation. (a)-(b) UAV simulated and estimated trajectories. frst 100 seconds. Afer lif-of, the UAV ic packets (See Orbcomm, Additional (c) Simulated and estimated trajectories makes 10 banking turns. Te simulated Resources). Te Orbcomm system con- and the fnal 95-th percentile uncertainty LEO satellite trajectories corresponded sists of 3 main segments: (i) subscriber FIGURE 17 Snapshot of the Orbcomm spectrum. ellipsoid for one of the simulated LEO satellites. Map data: Google Earth. to the upcoming Starlink constellation. communicators (users), (ii) ground seg- It was assumed that the LEO satellites ment (gateways), and (iii) space segment Orbcomm network elements and the in 7 orbital planes A–G, as illustrated AND IRIDIUM LEO CONSTELLATIONS were equipped with GPS receivers and (constellation of satellites). Tese seg- gateways through telemetry monitor- in Figure 16. Planes A, B, and C are FIGURE 13 UAV simulation results with the A UAV was equipped with (i) a tactical- were periodically transmitting their ments are briefy discussed next. ing, system commanding, and mis- inclined at 45° to the equator and each Globalstar, Orbcomm, and Iridium LEO constellations. (a)-(b) UAV simulated and grade IMU, (ii) GPS and LEO satellite estimated position. Tere was a total of (i) Subscriber Communicators (SCs): Tere sion system analysis. contains 8 satellites in a circular orbit at estimated trajectories. (c) Simulated and receivers, and (iii) a pressure altimeter. 78 LEO SVs that passed within a preset are several types of SCs. Orbcomm’s (iii) Space Segment: Orbcomm satellites an altitude of approximately 815 kilo- estimated trajectories and the fnal 95-th Te UAV navigates over Santa Monica, 35° elevation mask set, with an average SC for fixed data applications uses are used to complete the link between meters. Plane D, also inclined at 45°, percentile uncertainty ellipsoid for one of the simulated LEO satellites. Map data: California, USA, for about 25 kilome- of 27 SVs available at any point in time. low-cost VHF electronics. Te SC for the SCs and the switching capability at contains 7 satellites in a circular orbit at Google Earth. ters in 200 seconds, during which it Te UAV made pseudorange and pseu- mobile two-way messaging is a hand- the NCC or GCC. an altitude of 815 kilometers. Plane E is had access to GPS signals only for the dorange rate measurements to all LEO held, standalone unit. inclined at 0° and contains 7 satellites in frst 100 seconds. Afer lif-of, the UAV satellites. Te LEO satellites’ positions (ii) Ground Segment: The ground seg- Orbcomm LEO Satellite Constellation a circular orbit at an altitude of 975 kilo- makes 4 banking turns. A total of 10 in the STAN framework were initialized ment consists of gateway control cen- The Orbcomm constellation, at maxi- meters. Plane F is inclined at 70°and con- LEO satellite trajectories were simu- using the frst transmitted LEO satellite ters (GCCs), gateway Earth stations mum capacity, has up to 47 satellites tains 2 satellites in a near-polar circular

60 InsideGNSS JULY/AUGUST 2019 www.insidegnss.com www.insidegnss.com JULY/AUGUST 2019 InsideGNSS 61 STAN WITH LEO

orbit at an altitude of 740 kilometers. Plane G is inclined at 108° and contains 2 satellites in a near-polar elliptical orbit at an altitude varying between 785 and 875 kilometers. Te LEO receiver draws pseudorange rate observables from Orbcomm LEO signals on the downlink channel. Satellite radio frequency (RF) downlinks to SCs and GESs are within the 137– 138 MHz VHF band. Te downlink channels include 12 channels for transmitting to the SCs and one gateway channel, which is reserved for transmitting to the GESs. Each satellite trans- mits to the SCs on one of the 12 subscriber downlink channels through a frequency-sharing scheme that provides 4-fold channel reuse. Te Orbcomm satellites have a subscriber transmitter that provides a continuous 4800 bits-per-second (bps) stream of FIGURE 18 Outputs of Orbcomm receiver: (a) estimated Doppler, (b) packet data using symmetric diferential-quadrature phase shif carrier phase error, (c) demodulated QPSK symbols, and (d) QPSK keying (SD-QPSK). Each satellite also has multiple subscriber symbol phase transitions. receivers that receive short bursts from the SCs at 2400 bps. Figure 17 shows a snapshot of the Orbcomm spectrum. Figure 18 shows some of the internal signals of the receiver used to extract Doppler measurement from Orbcomm signals, mainly: (a) an estimate of the Doppler frequency, (b) the carrier phase tracking error, (c) the demodulated QPSK modulation, and FIGURE 22 Hardware and software setup for the UAV experiment. (d) the QPSK symbol phase transitions. Te Orbcomm receiver is part of the Multichannel Adaptive TRansceiver Information Unaided INS LEO-aided INS STAN eXtractor (MATRIX) sofware-defned radio (SDR) developed Final Error (m) 3,729.4 192.3 by the Autonomous Systems Perception, Intelligence, and RMSE (m) 1,419.3 416.5 Navigation (ASPIN) Laboratory (see, http://aspin.eng.uci.edu) Table 4: Experimental results with 2 Orbcomm LEO satellites for a ground vehicle navigating (Autonomous Systems Perception, Intelligence, and Navigation FIGURE 21 Results of the ground vehicle about 7.5 km in 258 seconds (GPS signals were cut off after the first 30 seconds). These results Laboratory, Additional Resources). Te receiver performs car- experiment. (a) Orbcomm satellite are after GPS cutoff. rier synchronization, extracts pseudorange rate observables, and trajectories. (b)-(c) Ground vehicle true decodes Orbcomm ephemeris messages. and estimated trajectories. (d) Estimated navigation frameworks were imple- lowing hardware and sofware setup: trajectory and the fnal 95-th percentile Note that Orbcomm satellites are also equipped with a spe- uncertainty ellipsoid for one of the mented to estimate the ground vehicle’s • A high-end quadriflar helix antenna cially constructed 1-Watt ultra-high frequency (UHF) trans- Orbcomm satellites. Map data: Google Earth. trajectory: (i) the LEO-aided INS STAN • A USRP to sample Orbcomm signals. mitter that is designed to emit a highly stable signal at 400.1 framework and (ii) a traditional GPS- Tese samples were then processed by megahertz. The transmitter is coupled to a UHF antenna diferential GPS base station to obtain a aided INS for comparative analysis. Each the Orbcomm receiver module of the designed to have a peak gain of approximately 2 dB. Te UHF carrier phase-based navigation solution. framework had access to GPS for only the MATRIX SDR. signal is used by the Orbcomm system for SC positioning. Tis integrated GNSS-IMU real-time frst 30 seconds. Figure 21(a) illustrate • A consumer-grade MEMS IMU, which However, experimental data shows that the UHF beacon is kinematic (RTK) system was used to the trajectory the 2 Orbcomm LEO sat- is proprietary hardware of the UAV absent. Moreover, even if the UHF beacon were present, one produce the ground truth results with ellites traversed over the course of the manufacturer and used in its flight would need to be a paying subscriber to beneft from position- which the STAN navigation framework experiment, Figure 21(b)-(c) illustrate controller. Log fles were downloaded FIGURE 19 Hardware and software setup for the ground vehicle ing services. Consequently, in this work, only downlink VHF was compared. the ground vehicle’s true trajectory and from the drone to parse the raw IMU experiment. signals are used in the LEO-aided INS STAN. Te experimental setup is shown in those estimated by each of the 2 frame- data, which were subsequently fed to Figure 19. works, and Figure 21(d) illustrates the INS of the STAN framework. Ground Vehicle Navigation Te ground vehicle was driven along the estimated trajectories of one of the • A pressure altimeter, which is also pro- An experiment was conducted to evaluate the performance of U.S. Interstate 5 near Irvine, California, Orbcomm satellites as well as the fnal prietary hardware of the UAV manu- the LEO-aided INS STAN framework on a ground vehicle tra- USA, for 7,495 meters in 258 seconds, dur- 95-th percentile uncertainty ellipsoid facturer and used in its fight control- versing a long trajectory. To this end, a car was equipped with ing which 2 Orbcomm LEO satellites were (the axes denote the radial (ra) and along- ler. Log fles were downloaded from the the following hardware and sofware setup: available (FM 112 and FM 117). Figure track (at) directions). drone to parse the altitude measure- • A custom-built quadriflar helix VHF antenna 20(a) depicts a skyplot of the satellite Table 4 summarizes the fnal error ments, which were subsequently fed to • A universal sofware radio peripheral (USRP) to sample trajectories over the course of the experi- and position RMSE achieved by each the EKF of the STAN framework. Orbcomm signals. Tese samples were then processed by ment. Figure 20(b) shows the Doppler framework afer GPS cutof. Te ground truth trajectory was taken FIGURE 20 (a) Skyplot of the Orbcomm satellite trajectories. (b) the Orbcomm receiver module of the MATRIX SDR. measured by the MATRIX SDR and the from the UAV’s onboard navigation sys- Doppler frequency measurement produced by the MATRIX SDR • An integrated GNSS-IMU, which is equipped with a dual- estimated Doppler using satellite position A. UAV NAVIGATION tem, which consists of a MEMS IMU, a and the expected Doppler according to an SGP4 propagator for the ground vehicle experiment. antenna, multi-frequency GNSS receiver and a microelectro- and velocity obtained from TLE fles and An experiment was conducted to evalu- multi-constellation GNSS receiver (GPS mechanical system (MEMS) IMU. A post-processing sofware an SGP4 propagator for the 2 Orbcomm ate the performance of the LEO-aided and GLONASS), a pressure altimeter, development kit (PP-SDK) was used to process GPS carrier satellites. INS STAN framework on a UAV. To this and a magnetometer. Te experimental phase observables collected by the GNSS-IMU and by a nearby To estimate the UAV’s trajectory, 2 end, the UAV was equipped with the fol- setup is shown in Figure 22.

62 InsideGNSS JULY/AUGUST 2019 www.insidegnss.com www.insidegnss.com JULY/AUGUST 2019 InsideGNSS 63 STAN WITH LEO

metropolitan-scale environment? A probabilistic approach with radio television synchronization signals,” IEEE Transactions on Broadcasting, measurements analysis,” IEEE Transactions on Broadcasting, vol. 55, no. vol. 51, no. 1, pp. 51–61, March 2005. 3, pp. 577–588, September 2009. (24) Reid, T., A. Neish, T. Walter, and P. Enge, “Broadband LEO constella- (6) Federal Communications Commission, “FCC boosts satellite broad- tions for navigation,” NAVIGATION, Journal of the Institute of Navigation, band connectivity and competition in the united states,” https://www. vol. 65, no. 2, pp. 205–220, 2018. fcc.gov/document/fcc-boosts-satellite-broadband-connectivity-com- (25) Shamaei, K., J. Khalife, and Z. Kassas, “Exploiting LTE signals for petition, November 2018, accessed February 27, 2019. navigation: Theory to implementation,” IEEE Transactions on Wireless (7) Hall, T., C. Counselman III, and P. Misra, “Radiolocation using AM Communications, vol. 17, no. 4, pp. 2173–2189, April 2018. broadcast signals: Positioning performance,” in Proceedings of ION GPS (26) Shamaei K., and Z. Kassas, “LTE receiver design and multipath Conference, September 2002, pp. 921–932. analysis for navigation in urban environments,” NAVIGATION, Journal of FIGURE 23 (a) Skyplot of the Orbcomm satellite trajectories. (b) Doppler frequency (8) Joerger, M., L. Gratton, B. Pervan, and C. Cohen, “Analysis of Iridium- the Institute of Navigation, vol. 65, no. 4, pp. 655–675, December 2018. measurement produced by the MATRIX SDR and the expected Doppler according to an SGP4 augmented GPS for floating carrier phase positioning,” NAVIGATION, propagator for the UAV experiment. (27) Thevenon, P., S. Damien, O. Julien, C. Macabiau, M. Bousquet, L. Journal of the Institute of Navigation, vol. 57, no. 2, pp. 137–160, 2010. Ries, and S. Corazza, “Positioning using mobile TV based on the DVB-SH (9) Kassas, Z., “Collaborative opportunistic navigation,” IEEE Aerospace standard,” NAVIGATION, Journal of the Institute of Navigation, vol. 58, LEO-aided LEO-aided INS STAN with periodically and Electronic Systems Magazine, vol. 28, no. 6, pp. 38–41, 2013. no. 2, pp. 71–90, 2011. Unaided INS INS STAN transmitted satellite positions (10) Kassas, Z., J. Khalife, K. Shamaei, and J. Morales, “I hear, therefore (28) Vetter, J., “Fifty years of orbit determination: Development of mod- Final Error (m) 123.5 29.9 5.7 I know where I am: Compensating for GNSS limitations with cellular ern astrodynamics methods,” Johns Hopkins APL Technical Digest, vol. RMSE (m) 53.7 15.9 5.4 signals,” IEEE Signal Processing Magazine, pp. 111–124, September 2017. 27, no. 3, pp. 239–252, November 2007. Table 5: Experimental results with 2 Orbcomm LEO satellites for a UAV navigating about 1.53 km in (11) Kassas, Z., J. Morales, K. Shamaei, and J. Khalife, “LTE steers UAV,” GPS (29) Yang, C., T. Nguyen, and E. Blasch, “Mobile positioning via fusion of 155 seconds (GPS signals were cut off after the first 125 seconds). These results are after GPS cutoff. World Magazine, vol. 28, no. 4, pp. 18–25, April 2017. mixed signals of opportunity,” IEEE Aerospace and Electronic Systems (12) Khalife J., and Z. Kassas, “Navigation with cellular CDMA signals– Magazine, vol. 29, no. 4, pp. 34–46, April 2014. Te UAV few a commanded trajec- error and position RMSE achieved by part II: Performance analysis and experimental results,” IEEE Transactions on Signal Processing, vol. 66, no. 8, pp. 2204–2218, April 2018. Authors tory in Irvine, California, USA, over each framework afer GPS cutof. Zaher (Zak) M. Kassas is an assistant professor in a 155-second period during which 2 (13) Khalife J., and Z. Kassas, “Precise UAV navigation with cellular carrier the Department of Mechanical & Aerospace Orbcomm LEO satellites were available Manufacturers phase measurements,” in Proceedings of IEEE/ION Position, Location, Engineering and the Department of Electrical (FM 108 and FM 116). Figure 23(a) In the Ground Vehicle Navigation sec- and Navigation Symposium, April 2018, pp. 978–989. Engineering & Computer Science at the University depicts a skyplot of the satellite trajec- tion, the authors’ setup included an Ettus FIGURE 24 Results of the UAV experiment. (14) Lawrence, D., H. Cobb, G. Gutt, M. O’Connor, T. Reid, T. Walter, and of California, Irvine (UCI) and director of the (a) Orbcomm satellite trajectories. (b)-(d) D. Whelan, “Navigation from LEO: Current capability and future prom- Autonomous Systems Perception, Intelligence, and tories over the course of the experi- E312 universal sofware radio peripheral UAV’s true and estimated trajectories. Map ment. Figure 23(b) shows the Doppler (USRP) from Ettus Research (Austin, data: Google Earth. ise,” GPS World Magazine, vol. 28, no. 7, pp. 42–48, July 2017. Navigation (ASPIN) Laboratory. He received a B.E. in Electrical Engineering from the Lebanese American University, an measured by the MATRIX SDR and the Texas, USA) to sample Orbcomm sig- (15) Maaref M., and Z. Kassas, “Ground vehicle navigation in GNSS- challenged environments using signals of opportunity and a closed- M.S. in Electrical and Computer Engineering from The Ohio State estimated Doppler using satellite posi- nals; an AsteRx-I V integrated GNSS- authors would like to thank Christian University, and an M.S.E. in Aerospace Engineering and a Ph.D. in tion and velocity obtained from TLE IMU from Septentrio (Leuven, Belgium Ardito, Linh Nguyen, Ali Abdallah, loop map-matching approach,” IEEE Transactions on Intelligent Transportation Systems, 2019, accepted. Electrical and Computer Engineering from The University of Texas at fles and an SGP4 propagator for the 2 and Torrance, California, USA); a Mohammad Orabi, Kimia Shamaei, Austin. In 2018, he received the National Science Foundation (NSF) (16) Maaref, M., J. Khalife, and Z. Kassas, “Lane-level localization and Orbcomm satellites. VectorNav VN-100 microelectrome- Mahdi Maaref, and Naji Tarabay for Faculty Early Career Development Program (CAREER) award, and in mapping in GNSS-challenged environments by fusing lidar data and To estimate the UAV’s trajectory, 3 chanical systems (MEMS) IMU from their help in data collection. 2019, he received the Office of Naval Research (ONR) Young cellular pseudoranges,” IEEE Transactions on Intelligent Vehicles, vol. 4, frameworks were implemented to esti- VectorNav Technologies (Dallas, Texas, Investigator Program (YIP) award. His research interests include cyber- no. 1, pp. 73–89, March 2019. mate the UAV’s trajectory: (i) the LEO- USA); and Septentrio’s post-processing References physical systems, estimation theory, navigation systems, autonomous aided INS STAN framework initialized software development kit (PP-SDK) (1) Ardito, C., J. Morales, J. Khalife, A. Abdallah, (17) Merry, L., R. Faragher, and S. Schedin, “Comparison of opportu- vehicles, and intelligent transportation systems. nistic signals for localisation,” in Proceedings of IFAC Symposium on using TLE fles, (ii) the LEO-aided INS was used to process GPS carrier phase and Z. Kassas, “Performance evaluation of navigation using LEO satellite signals with Intelligent Autonomous Vehicles, September 2010, pp. 109–114. Joshua J. Morales is a Ph.D. candidate in the STAN framework that used the decod- observables collected. Department of Electrical Engineering and Computer ed periodically transmitted LEO satel- In the experiment conducted to periodically transmitted satellite positions,” in (18) Morales, J., P. Roysdon, and Z. Kassas, “Signals of opportunity aided inertial navigation,” in Proceedings of ION GNSS Conference, Science at UCI and a member of the ASPIN Laboratory. lite positions, which were transmitted evaluate the performance of the LEO- Proceedings of ION International Technical Meeting, 2019, pp. 306-318. September 2016, pp. 1492–1501. He received a B.S. in Electrical Engineering with High by the Orbcomm satellites, and (iii) a aided INS STAN framework on a UAV, Honors from the University of California, Riverside. In traditional GPS-aided INS for com- a DJI Matrice 600 UAV with an A3 (2) Autonomous Systems Perception, (19) Morales, J., J. Khalife, A. Abdallah, C. Ardito, and Z. Kassas, “Inertial 2016, he was accorded an Honorable Mention from navigation system aiding with Orbcomm LEO satellite Doppler mea- parative analysis. Te estimated trajec- fight controller was used (Shenzhen, Intelligence, and Navigation (ASPIN) the National Science Foundation (NSF). His research Laboratory http://aspin.eng.uci.edu surements,” in Proceedings of ION GNSS Conference, September 2018, interests include estimation theory, navigation systems, autonomous tories were compared with the trajec- China); again, the setup included an pp. 2718-2725. tory extracted from the UAV’s onboard Ettus E312 USRP from Ettus Research (3) Brown R., and P. Hwang, Introduction to vehicles, and intelligent transportation systems. (20) Morales, J., J. Khalife, and Z. Kassas, “Simultaneous tracking of navigation system. Each framework (Austin, Texas, USA). Random Signals and Applied Kalman Filtering, 3rd ed. John Wiley & Sons, 2002. Orbcomm LEO satellites and inertial navigation system aiding using Joe J. Khalife is a Ph.D. candidate in the Department had access to GPS for only the first of Electrical Engineering and Computer Science at (4) Driusso, M., C. Marshall, M. Sabathy, F. Doppler measurements,” in Proceedings of IEEE Vehicular Technology 125 seconds. Figure 24(a) shows the Acknowledgements UCI and a member of the ASPIN Laboratory. He Knutti, H. Mathis, and F. Babich, “Vehicular Conference, 2019, pp. 1-6. trajectories that the 2 Orbcomm LEO This work was supported in part by received a B.E. in Electrical Engineering and an M.S. position tracking using LTE signals,” IEEE (21) North American Aerospace Defense Command (NORAD), “Two- satellites traversed over the course of the Office of Naval Research (ONR) in Computer Engineering from the Lebanese Transactions on Vehicular Technology, vol. line element sets,” http://celestrak.com/NO-RAD/elements/. the experiment. Figure 24(b)-(d) under the Young Investigator Program American University. In 2018, he received the IEEE 66, no. 4, pp. 3376–3391, April 2017. (22) Orbcomm, “Networks: Satellite,” https://www.orbcomm.com/en/ illustrate the UAV’s true trajectory and (YIP) award and in part by the National Walter Fried Award for Best Paper at the IEEE/ION (5) Fang, S., J. Chen, H. Huang, and T. Lin, networks/satellite, accessed September 30, 2018. those estimated by each of the 3 frame- Science Foundation (NSF) CAREER Position, Location, and Navigation Symposium (PLANS). His research “Is FM a RF-based positioning solution in a works. Table 5 summarizes the fnal award under Grant 1929965. The (23) Rabinowitz M., and J. Spilker, Jr., “A new positioning system using interests include opportunistic navigation, autonomous vehicles, and

64 InsideGNSS JULY/AUGUST 2019 www.insidegnss.com www.insidegnss.com JULY/AUGUST 2019 InsideGNSS 65 STAN WITH LEO

64 InsideGNSS JULY/AUGUST 2019 www.insidegnss.com www.insidegnss.com JULY/AUGUST 2019 InsideGNSS 65