Test and Evaluation of Mitigating Technologies for UAS in GPS Degraded and Denied Environments

Test and Evaluation of Mitigating Technologies for UAS in GPS Degraded and Denied Environments Timothy Pitt, US Army AMRDEC Greg Reynolds, US Army AMRDEC Will Barnwell, US Army PM UAS Jonathan Jones, Navigation Technology Associates, Inc. (NTA) Laura McCrain, Navigation Technology Associates, Inc. (NTA) Adam Simmons, Navigation Technology Associates, Inc. (NTA) BIOGRAPHIES Timothy Pitt received Bachelor of Science of Electrical Engineering and Master of Science of Systems Engineering degrees from the University of Alabama Huntsville in 2006 and 2009, respectively. He is currently supporting UAS programs for the US Army. Greg Reynolds received a Bachelor of Science in Electrical Engineering from the Missouri University of Science and Technology in 2004 and his Masters in Electrical Engineering from the University of Alabama in Huntsville in 2009. He has worked 14+ years with the U.S. Army Aviation and Missile Research, Development, and Engineering Center (AMRDEC) in the areas of navigation and simulation. Will Barnwell received Bachelor of Science and Master of Science degrees in Aerospace Engineering from NC State University in 2001 and 2003, respectively. He is currently the lead aircraft engineer for an Army UAS platform. Jonathan Jones received a Bachelor of Science in Electrical Engineering from the University of Alabama at Huntsville in 2004. He has over a decade of experience in navigation simulation, testing, and system integration. Laura McCrain received Bachelor of Science and Master of Science degrees in Aerospace Engineering from NC State University in 2001 and 2003, respectively. She has supported several UAS programs within the US DoD over the past 13 years. Adam Simmons received a Bachelor of Science in Electrical Engineering from Mercer University in 2000 and his Masters and PhD in Electrical Engineering from Auburn University in 2006. He has worked 11+ years with US DoD and Army testing GNSS integration and developing hardware-in-the-loop capabilities. ABSTRACT Unmanned Aircraft Systems (UAS) have been proven as an effective technology in performing Intelligence, Surveillance, and Reconnaissance (ISR) missions for military applications. As a consequence, the mission flight hours of UAS over the previous decade have grown significantly. With the predominance of this technology, there is a corresponding need to address potential threats against UAS. One such threat is a degradation or denial of Global Positioning System (GPS) signals, which have a significant role in the complex navigation solutions used by UAS. The Navigation Function within the Aviation and Missile Research, Development, and Engineering Center (AMRDEC) has partnered with the Army s Project Office for Unmanned Aircraft Systems (PM UAS) to test, evaluate, and mitigate the effects of these navigation threats on Army UAS platforms. The Army has developed a five-year road map to address these threats with two complementary and parallel approaches. The first approach uses new and emerging technologies to increase the

robustness of the GPS-based navigation solution in the presence of threats. The second approach uses novel concepts to lessen the system s dependency on GPS for a navigation solution. The AMRDEC Navigation Function supports PM UAS in testing and evaluating these approaches. INTRODUCTION The Navigation Function at AMRDEC has a long history of developing, testing, and evaluating navigation solutions for US Army platforms. An important portion of that history is a foundational understanding of Inertial Navigation Systems (INS). A historical repository of in-house, measured data from years of testing Inertial Measurement Units (IMU) is used to develop models for simulation purposes. These models are the cornerstone of the Position, Navigation, and Timing (PNT) Test Bed at AMRDEC. The PNT Test Bed was developed under a Missile Science and Technology program with the objective of providing a capability to test and evaluate integrated navigation systems that utilize Global Positioning System (GPS), inertial sensors, and emerging navigation sensors. The PNT Test Bed is adaptable and not platform- or hardware-specific. It has simulated multiple missile and aviation systems and has been validated through flight test data. The PNT Test Bed is used to test future capabilities that include a wide range of alternative navigation solutions. The goal of the Navigation Function is to develop robust holistic navigation solutions that mitigate GPS threats and lessen the dependency on GPS. In order to assess state-of-the-art technologies, as well as new and emerging threats, the PNT Test Bed is continually being expanded and improved to meet the navigation testing needs of the US Army. More recent improvements include integrating a GPS Simulator with seven Radio Frequency (RF) elements and developing an Electro-Optical / Infrared (EO/IR) Payload Model. These improvements were critical for integrating and testing seven element controlled radiation pattern antennas (CRPA) and accompanying electronics, as well as evaluating vision-based navigation solutions. PNT TEST BED The PNT Test Bed is composed of the following subsystems: a Simulation Host, Global Navigation Satellite System (GNSS) Simulators, Simulation Display (SIMDIS), and the Wavefront Integrated Threat Simulator (WITS). Each of these subsystems is discussed in more detail below. A block diagram of the PNT Test Bed architecture is shown in Figure 1. In addition to integrating with the Unit Under Test s (UUT) 6-Degree of Freedom (6-DOF) model (for closed loop simulations), the PNT Test Bed simulates in real-time the inertial inputs to the UUT s navigators, the Satellite Constellation that would be observed by the UUT, and the navigation threats to the UUT for multiple independent, asynchronous navigators. The PNT Test Bed has integrated as many as three concurrent navigators (that utilize inertial sensors with GPS aiding), with independent inertial update rates, into a single simulation scenario. The modularity of the PNT Test Bed allows for additional, concurrently simulated navigators to be integrated in the future.

Figure 1: Block diagram of PNT Test Bed Simulation Host Multiple navigators within the UUT can be included in an open or closed loop simulation. The Simulation Host injects simulated inertial sensor measurements into the UUT s navigators and updates the GNSS Simulator in order to represent the vehicle s dynamics in coordinated real-time. The inertial simulation host software generates inertial data based on representative inertial sensor error parameters and trajectory dynamics. The inertial parameters are defined by an IMU error model representing 84 error sources. Error parameters can be tailored to a specific IMU, or can be generated based on a theoretical IMU for evaluation. The software can also be used to generate perfect sensor data or a Monte Carlo sensor data set using defined error parameters. Global Navigation Satellite System (GNSS) Simulators Multiple GNSS Simulators are available for use with the PNT Test Bed: L3 Communications (Model 2450), Spirent (Model 9000), and Computing Applications Software Technology (CAST) (Model 5000B and Model 5000C). All of these simulators are capable of operating with real-time updates. The Spirent and L3 GNSS simulators have the ability to simulate two simultaneous vehicles or a two-element receiver simultaneously. The CAST GNSS Simulators can simulate seven individual vehicles; or the seven outputs can be configured to represent wavefront effects on a multi-element (up to seven elements) antenna. Combining the outputs from the two CAST GNSS Simulators along with the precision timing reference and Simulation Host allows for generation of 14 independent phase center locations. The Spirent and CAST simulators are fully upgradeable to support Military-Code (M-Code) user equipment. Simulation Display (SIMDIS) The PNT Test Bed uses SIMDIS for visualization of on-going test events. With SIMDIS, multiple navigation solutions can be simultaneously displayed in real time with the truth reference. For example, the effects of integrating an Anti-Jam (AJ) unit on one of the UUT s navigators can be observed in real time to an unprotected navigator (Figure 2). Additionally, if the UUT has a payload simulation integrated, SIMDIS can interface with the payload simulation as well to display sensor pointing angles. Additionally, SIMDIS provides a means to graphically display a threat s area of influence. Post-test analysis is supported by SIMDIS s playback features.

Figure 2: SIMDIS visualization of multiple navigator solutions for Unit Under Test Wavefront Integrated Threat Simulator (WITS) The WITS is an in-house, Government-developed simulation that provides a suite of threat simulations with varying numbers of jammers and associated noise floors. The WITS calculates the phase delay, Doppler, and attenuation values over the course of the UUT s trajectory and properly applies these delay and attenuation values when simulating either fixed radiation pattern antennas (FRPA) or controlled radiation pattern antenna (CRPA). For CRPAs, the threat signal reaches the individual antenna elements at different times, as shown in Figure 3. The time difference of the signal is proportionally related to the phase difference of the signal. The t in Figure 3 is a depiction of this time difference between the signals reaching element 1 relative to element 2. Δt 1 3 2 4 Θ Δt = [ (λ/2) / c ] cos (Θ) Figure 3: Phase delay and attenuation at individual elements of 4-element CRPA The PNT Test Bed provides a modular, hard real-time simulation platform for development, test, evaluation, and integration of navigation technologies and mitigations to threats against these technologies. Development of navigation algorithms and evaluation of navigation system architectures are also supported by the PNT Test Bed. Key features for the simulation framework are a real time operating system, a modular architecture, and precision timing. Incorporating these three key features simultaneously, and having the ability to drive multiple navigators asynchronously, make the PNT Test Bed a unique capability within the Navigation community. The PNT Test Bed simulates multiple, independent inertial sensors, while updating the integrated GNSS Simulators to deliver the corresponding measurements to represent the vehicle dynamics in coordinated real time. The test bed is expanded with the integration of a decommissioned unmanned aircraft and its 6-Degree of Freedom (6- DOF) model provided by PM UAS and the prime contractor, respectively.

Real-Time Operating System (RTOS) The RTOS used in the PNT Test Bed is Xenomai Real-Time Framework for Linux. This RTOS supports development of hard real-time, deterministic applications using an industry standard POSIX interface. Xenomai additionally enables prioritization of threads, including the capability to preempt normal system tasks with higher priority tasks. Modular Architecture Modularity of the PNT Test Bed architecture allows for flexibility and accommodates operation as a closed loop or an open loop simulation. The closed loop configuration requires the integration of the UUT s 6-DOF model, which injects the UUT s control surface commands into the simulation. Incorporation of a 6-DOF results in a higher fidelity test environment, as the UUT s control surface commands are coupled with the PNT Test Bed s inertial simulation. Additionally, incorporating the 6- DOF model allows for visibility into flight control behavior during test and evaluation events. 6-DOF models from various industry partners have been successfully integrated into the PNT Test Bed. If a 6-DOF model is unavailable, open loop testing can be performed with pre-determined trajectories to determine navigation accuracy requirements and perform end game analyses. The open loop mode also allows for Monte Carlo analyses with hardware in the loop. Additionally, the ability to perform all digital Monte Carlo analyses, incorporating the same models used in the Hardware-in-the-Loop (HWIL) variant of the PNT Test Bed, is a cost and time effective way to generate a statistically significant data set. Figure 4 shows results from a Monte Carlo analysis to determine the horizontal position error of a navigator over multiple hours in a GPS denied environment. For each Monte Carlo run, the 84 inertial sensor error parameters included in the inertial model are varied within prescribed bounds for the modeled sensor. Figure 4: Monte Carlo analysis of horizontal position error versus time Precision Timing The hard real-time system used in conjunction with custom developed FPGA hardware and a precision timing reference synchronizes all components within the PNT Test Bed to a common clock for the coordinated, deterministic delivery of inertial sensor measurements, RF signals, and the sensor payload simulation output. This implementation coordinates the necessary, precise delivery of all signals used in the simulation, while allowing the user to collect and store simulation scenario data with time tags in real time. Verification and Validation of PNT Test Bed The PNT Test Bed leverages over 50 years of test and evaluation experience. Past experience in test, evaluation, and integration of both inertial sensors and GPS receivers (using the same PNT Test Bed technologies and techniques discussed herein) provides an excellent foundation for the system-level navigation evaluations being conducted for Army UAS. Flight test data is available for many of the past hardware evaluations, which is a means to verify the PNT Test Bed s simulation capabilities. Figure 5, Figure 6, and Figure 7 show simulated results from the PNT Test Bed (both nominal and Monte Carlo (MC) runs) in comparison to flight test data for average Carrier-to-Noise Density rations (C/N 0), altitude, and pitch versus flight time. This sampling of comparison data, and the compendium of other flight test data, verifies the accuracy of the models used in the PNT Test Bed simulations.

Figure 5: Comparison of simulated and flight test data for average C/N0 versus flight time Figure 6: Comparison of simulated and flight test data for absolute altitude versus flight time

Figure 7: Comparison of simulated and flight test data for pitch versus flight time MITIGATIONS TO GPS DENIED AND DEGRADED ENVIRONMENTS To mitigate the effects of GPS denied and degraded environments on UAS, PM UAS is pursuing two parallel technical solutions. The first approach is to make the existing GPS usage on-board the UAS more robust. The second tactic is to evaluate alternate methods of navigation that do not rely on GPS. For both solutions, which are discussed in the following sections, the Navigation Function serves as an independent evaluator and provides an extensible test bed that can be adapted to perform quantitative evaluations of proposed technologies. Evaluations are conducted in conjunction with industry partners; this is a mutually beneficial arrangement in that the manufacturers can incorporate product improvements based on results from testing in the PNT Test Bed and can actively support optimal integration of their hardware into the PNT Test Bed during the evaluation cycles. Anti-Jam The PNT Test Bed provides a means to independently evaluate currently available anti-jam (AJ) systems. Assessment matrices for each AJ system are provided to PM UAS; the metrics for evaluation are collaboratively generated among PM UAS, AMRDEC, and the prime contractor for the UAS under evaluation. Currently available anti-jam (AJ) systems that meet the envelope of Size, Weight and Power (SWAP) requirements for a particular UUT are considered for evaluation in the PNT Test Bed. In addition to the technical performance of the AJ system, i.e. its ability to enable GPS receivers to maintain higher C/N 0 values in the presence of threats, other system-level characteristics of the AJ unit are evaluated. The SWAP characteristics of the AJ unit are also components of the assessment matrices, along with production readiness, reliability, and extensibility of the system to incorporate future technologies that improve GPS robustness through software and firmware updates. Five AJ systems are evaluated. A primary goal of this testing is to quantify the effectiveness of the AJ systems against various injected noise signals. This evaluation is parametrically evaluated for multiple Jammer-to-Signal (J/S) levels over a range of dynamic angle-of-arrival motion profiles relative to the threat transmitter (>120 db power calibrated J/S and discrete geometric on/off threat transmissions). The initial testing of an AJ system begins with live sky testing where the AJ system is allowed to receive live GPS signals with no noise signals. Two RF data streams are received by commercial-off-the-shelf (COTS) receivers: one path that incorporates the AJ system and CRPA along with a reference path that has no AJ system and uses a FRPA. This initial testing builds familiarity with the AJ system performance, e.g. expected C/N 0 values, and experience in integrating the hardware. The live sky testing is next expanded to include a jamming signal, which is injected downstream of the GPS antennas and prior to the AJ system through a series of RF cables and components as shown in Figure 8.

Figure 8: Schematic of live sky test of AJ hardware with jamming signal Figure 9 shows an approximate 52 db J/S performance improvement in a jamming environment with the inclusion of an AJ system (bottom) in comparison to receiver with no AJ system (top). The C/N 0 values for the visible satellites are shown as a function of time for both the CRPA and FRPA configurations. Qualitatively, as the jammer power increases, the receiver equipped with an AJ system is able to maintain higher C/N 0 values for a longer period of time before losing GPS signal.

Increasing Jamming Power Figure 9: C/N0 for receiver without AJ (top) and with AJ (bottom) in live sky jamming test Candidate AJ systems are then integrated into the PNT Test Bed and undergo a suite of baseline testing, followed by testing with jamming signals provided by the WITS (Figure 1). A navigator with no AJ system an unprotected AJ system is tested simultaneously in the PNT Test Bed as a reference. Results from testing an AJ system with a 4-element CRPA in the presence of a jammer transmitting from beneath the air vehicle s simulated loiter pattern at 45 degrees off zenith are presented below. Figure 10 shows the C/N 0 values of the satellite RF signals from the GNSS simulators, along with an average of those signals, for two cases: one with an AJ system integrated (top) and one without an AJ system integrated (bottom). During the test, the jammer was incrementally increased by the amounts indicated at the top of Figure 10. With an AJ system, the C/N 0 values remain around 20 db*hz in the presence of a jamming signal with a J/S of 78 db. In contrast, without an AJ system, the C/N 0 values degrade below 20 db*hz once the jamming signal reaches a J/S of 43 db. The J/S of the jamming signal cannot exceed 78 db in this particular test set-up because of the test bed s noise floor; RF plumbing and simulation hardware reduce the available power by 30 db.

Figure 10: C/N0 for navigator with AJ (top) and without AJ (bottom), as measured in the PNT Test Bed in the presence of a stepped jamming signal Figure 11 shows the pseudorange residuals for a navigator with an AJ (top) and without an AJ (bottom). Pseudorange residuals are a metric by which to evaluate the accuracy of a tightly-integrated navigator s position solution and pseudorange consistency. As the J/S approaches 43 db, the navigator with no AJ system sees an increase in the pseudorange residual and eventually loses all of its tracked satellites (pseudorange residuals are held constant when lost). The AJ-aided navigator's pseudorange residuals are not only effected by the observed signal strength but by switching behavior within the AJ unit itself. As visible during earlier J/S portions of the scenario, the AJ-aided pseudorange residuals grow at 13dB J/S where the jamming signal is beginning to add more energy to the perceived noise floor. Between 10 to 20dB J/S is typically where an AJ will gain observability into the jamming signal s angle-of-arrival and augment its weights to reduce the signal s unwanted influence. Figure 11: Pseudorange Residuals for a navigator with AJ (top) and without AJ (bottom), as measured in the PNT Test Bed in the presence of a stepped jamming signal The AJ system s influence on the UAS at an integrated, system level is also evaluated. Integrating the AJ system into the RF path between the GPS antenna and the navigator s GPS receiver creates a quiescent time delay, i.e. the introduction of the additional hardware inherently creates a delay in the receipt of GPS data by the receiver. The quiescent time delay is measured in three currently available AJ systems; the measured quiescent time delay varied from milliseconds to microseconds. This metric is sensitive to the particular AJ hardware under evaluation. Additionally, open air jamming test will be performed at a ground test event. The ground test results serve to further validate the PNT Test Bed s hardware-in-the-loop simulation (HILSIM). Alternative Navigation The PNT Test Bed is used to assess potential vision navigation solutions that can be used by Army UAS platforms in GPS denied environments. An electro-optical / infrared (EO/IR) payload simulation has been developed and integrated into the PNT Test Bed. With the integration of the payload simulation, the PNT Test Bed is capable of simulating payload video feeds that are coordinated in real time with the UAS s simulated flight dynamics and geographic location. Candidate vision navigation solutions utilize this simulated payload video feed in the PNT Test Bed and provide navigation updates to the UUT s navigation system (Figure 1). The vision navigation solution s effectiveness in providing corrections to the navigation system during GPS denied and degraded environments is measured. Figure 12 displays initial results from a Vision Based Navigation (VBN) algorithm applied to video captured from the payload simulation. The 2-dimensional (2D) error of the position solution from the VBN algorithm is compared to the position solution from the UUT s navigator in the absence of GPS. During the

video capture, the payload simulation was pointed mostly nadir with a wide field of view, i.e. a best case configuration for VBN algorithms. Additional testing is needed to determine the effects of field of view settings and pointing angles on position accuracy. A design objective is to avoid requiring the payload operator to alter operational procedures when vision based navigation is in use; the payload can be operated as required to execute the mission with no requirement to change field of view and pointing angle to get an updated position solution. 1.00 0.85 2D Error (non-dimensional units) 0.70 0.60 0.45 0.30 0.15 Figure 12: 2-Dimensional (2D) error of Vision Based Navigation (VBN) in comparison to UUT s Navigator The VBN solution is also integrated into the PNT Test Bed with a candidate AJ system. Figure 13 shows the magnitude of position error (top left) and magnitude of the velocity error (bottom left) of a navigator protected with both an AJ system and VBN solution in comparison to an unprotected navigator. This test case has the same J/S profile as that discussed in the Anti- Jam section; the J/S values are incrementally increased, and the unprotected navigator cannot resolve a navigation solution at 43 db J/S. As expected, the unprotected navigator s position error increases more rapidly than the position error of the navigator with an AJ system and VBN solution. The periodic peaks in the position and velocity errors for the VBN solution are the result of the initial VBN algorithms not calculating a solution during the air vehicle s turns; future implementations will not have this limitation. At 385 seconds, the AJ is powered off and the now unprotected navigator s position and velocity errors begin to increase at a rate comparable to the reference unprotected navigator. Once the AJ is powered off, the VBN solution continues to compute and has position errors that are an order of magnitude less than the unprotected AJ system. The velocity error of the VBN solution is more comparable to the unprotected navigator, but the velocity portion of the navigation solution from the VBN is less critical in providing a position correction to the navigator in a GPS denied/degraded environment. The VBN position error shows the advantages of an alternative navigation techniques in GPS denied and degraded environments.

Figure 13: Magnitudes of position and velocity errors for navigators experiencing a stepped jamming signal: with no AJ, with AJ, and with AJ and VBN CONCLUSION The PNT Test Bed is used to characterize both near-term and emerging capabilities that mitigate GPS threats and lessen US Army Systems dependency on GPS. The Army recognizes that both the solutions and the threats themselves will continually evolve and has invested in a PNT Test Bed that is modular and extensible. In addition to evaluating technologies to mitigate the effects of operating in GPS denied and degraded environments, results from the PNT Test Bed also inform an operational envelope for UAS missions. This envelope delineates the operational conditions (including threat environment) in which the various mitigations to GPS threats are applicable. Knowledge about the performance and limitations of each mitigation informs the mission objectives and constraints a tremendous benefit to the Warfighter. ACKNOWLEDGMENTS The authors would like thank the following individuals for their technical contributions to this paper: Aubrey Adams, Ryan Cowart, Aaron Finch, Michael Payne, Jonathan Ryan, and Christopher Yeager. REFERENCES Ding, W. et al., Time Synchronization Error and Calibration in Integrated GPS/INS Systems, ETRI Journal, Volume 30, Number 1, February 2008, pp. 59-67. Khaghani, M. and Skaloud, J., Autonomous Vehicle Dynamic Model-Based Navigation for Small UAVs, NAVIGATION: Journal of the Institute of Navigation, Volume 63, Number 3, Fall 2016, pp. 345-358. Moussa, A., Ali, A., El-Sheimy, N., The Effect of Time Synchronization on Real Time Implementation of Integrated GPS/INS Systems, ION 2010 International Technical Meeting, San Diego, CA, 25-27 January 2010. Reynolds, G., Simmons, A., and Joiner, L., Validated Low-Cost Wavefront Integrated Threat Simulator, Proceedings of the Pacific Positioning, Navigation, and Timing Conference, Honolulu, Hawaii, May 2017. Sheta, B., Vision Based Navigation (VBN) of Unmanned Aerial Vehicles (UAV), PhD Thesis, University of Calgary, 2012.