MOBILE DOCKING OF REMUS-100 EQUIPPED WITH USBL-APS TO AN UNMANNED SURFACE VEHICLE: A PERFORMANCE FEASIBILITY STUDY.

Transcription

1 MOBILE DOCKING OF REMUS-100 EQUIPPED WITH USBL-APS TO AN UNMANNED SURFACE VEHICLE: A PERFORMANCE FEASIBILITY STUDY by Mario Miranda II A Thesis Submitted to the Faculty of The College of Engineering and Computer Science In Partial Fulfillment of the Requirements for the Degree of Master of Science Florida Atlantic University Boca Raton, Florida May 2014

2 ii

3 ACKNOWLEDGMENTS The author wishes to thank the Technical support office at Kongsberg Hydroid, as well as Tom Austin (WHOI), with help on the REMUS-100 and DUSBL-APS. Also a special thanks to Michael Porter for help with the Bellhop interface. This work is sponsored by the Office of Naval Research (Kelly Cooper, code 33) under ONR contract number N iii

4 ABSTRACT Author: Title: Institution: Advisor: Degree: Mario Miranda Mobile Docking of REMUS-100 Equipped with USBL-APS to an Unmanned Surface Vehicle: A Performance Feasibility Study Florida Atlantic University Pierre-Philippe Beaujean, PhD Master of Science Year: 2014 The overall objective of this work is to evaluate the ability of homing and docking an unmanned underwater vehicle (Hydroid REMUS 100 UUV) to a moving unmanned surface vehicle (Wave-Adaptive Modular Surface Vehicle USV) using a Hydroid Digital Ultra-Short Baseline (DUSBL) acoustic positioning system (APS), as a primary navigation source. An understanding of how the UUV can rendezvous with a stationary USV first is presented, then followed by a moving USV. Inherently, the DUSBL-APS is susceptible to error due to the physical phenomena of the underwater acoustic channel (e.g. ambient noise, attenuation and ray refraction). The development of an APS model has allowed the authors to forecast the UUV s position and the estimated track line of the USV as determined by the DUSBL acoustic sensor. In this model, focus is placed on three main elements: 1) the acoustic channel and sound ray refraction when propagating in an inhomogeneous medium; 2) the detection component of an ideal DUSBL-APS using the iv

5 Neyman-Pearson criterion; 3) the signal-to-noise ratio (SNR) and receiver directivity impact on position estimation. The simulation tool is compared against actual open water homing results in terms of the estimated source position between the simulated and the actual USBL range and bearing information. v

6 MOBILE DOCKING OF REMUS-100 EQUIPPED WITH USBL-APS TO AN UNMANNED SURFACE VEHICLE: A PERFORMANCE FEASIBILITY STUDY List of Tables... vii List of Figures... viii 1 Introduction... xii 1.1 Background Unmanned Marine Vehicles Hydroid REMUS-100 UUV Marine Advanced Research WAM-V Research Objectives Scientific Contribution Literature Review Scientific Background DUSBL Acoustic Positioning System Key Parameters in USBL Positioning System Update Rate Intermittent Communication Dropout Subsurface Obstacles Sound Propagation Simulation Software Overall Architecture Mathematical Derivations Signal Detecton model Generating a Normalized Signal Scaling Signal Spectrum and Estimating SNR Estimating Probability of Detection Position Estimation Model vi

7 Estimating Range to Source Estimating Bearing to Source UUV and USV Navigation Model Simulation Functionality Field Experimentation Stationary Homing Transponder Buoy Homing Test Fixed USV Homing Test Mathematical Derivations Research Results Fixed Transponder Homing Simulation Mobile Transponder Homing Simulation Experimental Results Fixed Buoy Open-water Homing Fixed USV Open-water Homing USV Homing Seatrial USV Homing Seatrial Conclusion Appendices A1. FAU USBL Beam Pattern A2. Acoustic Refraction and FER Tracing A3. Simulator Interface Notes A4. Fixed Transponder Homing Simulation Results A5. Mobile Transponder Homing Simulation Results Bibliography vii

8 LIST OF TABLES Table 1. WHOI/Hydroid Homing Test Results Table 2. Bellhop FER Trace Output Parameters Table 3. Varied Simulation Input Parameters Table 4. Constant Simulation Input Parameters Table 5. Fixed Homing Simulation Results Table 6. Sea-trial #1, Homing Results Table 7. Sea-trial #2, Homing Results Table 8. Sea-trial #3, Homing Results Table 9. Fixed Homing Simulation #1 Analysis Details Table 10. Fixed Homing Simulation #2 Analysis Details Table 11. Fixed Homing Simulation #3 Analysis Details Table 12. Fixed Homing Simulation #4 Analysis Details Table 13. Fixed Homing Simulation #5 Analysis Details Table 14. Mobile Homing Simulation #1 Analysis Details Table 15. Mobile Homing Simulation #2 Analysis Details Table 16. Mobile Homing Simulation #3 Analysis Details Table 17. Mobile Homing Simulation #3 Analysis Details Table 18. Mobile Homing Simulation #4 Analysis Details Table 19. Mobile Homing Simulation #5 Analysis Details viii

9 LIST OF FIGURES Figure 1. Unmanned Marine Vehicles...2 Figure 2. Swarm Vehicle Configuration...3 Figure 3. Conops for Homing Scenario...4 Figure 4. Modified Hydroid REMUS-100 UUV...5 Figure 5. Marine Advanced Research WAM-V USV...6 Figure 6. FAU USBL Mean Position Estimation Error...11 Figure 7. Hydroid Docking Photos for Commercial Products...16 Figure 8. Hydroid Docking Results at Buzzards Bay, MA...16 Figure 9. Hydroid DUSBL System...18 Figure 10. DUSBL Signal Processing Flow Chart...20 Figure 11. Ray Path Free Body Diagram From Source to Receiver...21 Figure 12. Sound Velocity Profile Used in Simulation...26 Figure 13. Simulated Ray Trace for Ft. Lauderdale Area...27 Figure 14. Design Architecture of Simulator...31 Figure 15. FFT of Band-Pass Filtered Message...33 Figure 16. Polar Plot of FAU USBL Beam Pattern...35 Figure 17. Probability Distribution for Signal Detection...38 Figure 18. Free-Body Diagram of REMUS-100 Orientation...42 Figure 19. Closed-loop Control Diagram for Simulated UUV Nav...43 Figure 20. Simulation Functionality Diagram...48 Figure 21. LIDAR Image of Experimental Test Location...49 Figure 22. Hydroid REMUS-100 with Docking Wings (1 st Gen)...51 Figure 23. Sea-trial #1, REMUS-100 with Docking Wings Implemented...52 Figure 24. Sea-trial #1, Observation and Docking Equipment Setup...53 Figure 25. Sea-trial #1, Programmed UUV Track...53 Figure 26. Sea-trial #2, REMUS-100 with Stinger Capturing Device...54 ix

10 Figure 27. Sea-trial #2, Observation and Docking Equipment Setup...55 Figure 28. Simulated Horizontal CT Error Histogram Range=200m,Depth=5m...63 Figure 29. Simulated Vertical CT Error Histogram Range=200m,Depth=5m...63 Figure 30. Cross-track Error YZ Plane Polar Plot Range=200m,Depth=5m...63 Figure 31. Simulated Vertical CT Error Histogram Range=200m,Depth=10m...64 Figure 32. Simulated Horizontal CT Error Histogram Range=200m,Depth=10m...64 Figure 33. Cross-track Error YZ Plane Polar Plot Range=200m,Depth=10m...65 Figure 34. Simulated Horizontal CT Error Histogram Range=200m,Depth=5m...67 Figure 35. Simulated Vertical CT Error Histogram Range=200m,Depth=5m...68 Figure 36. Cross-track Error YZ Plane Polar Plot Range=200m,Depth=5m...68 Figure 37. Simulated Vertical CT Error Histogram Range=200m,Depth=10m...69 Figure 38. Simulated Horizontal CT Error Histogram Range=200m,Depth=10m...69 Figure 39. Cross-track Error YZ Plane Polar Plot Range=200m,Depth=10m...70 Figure 40. Sea-trial #1 Results, Image of Drop Target During Docking Success...72 Figure 41. Sea-trial #1 Results, UUV Track During Docking Success...73 Figure 42. Sea-trial #1 Results, UUV Track Vertical Plane...73 Figure 43. Sea-trial #2 Results, Image Recorded By GoPro Cam Dock Attempt...75 Figure 44. Sea-trial #3 Results, Programmed UUV Track for Runs 1 and Figure 45. Sea-trial #3 Results, Programmed UUV Track for Runs 3 and Figure 46. Sea-trial #3 Results, Run 1 UUV Track Results...79 Figure 47. Sea-trial #3 Results, Run 1 Zoomed-in UUV Track Results...80 Figure 48. Sea-trial #3 Results, Run 1 UUV Estimated Forward Velocity...80 Figure 49. Sea-trial #3 Results, Run 1 UUV Vertical Track...81 Figure 50. Sea-trial #3 Results, Run 2 UUV Track Results...82 Figure 51. Sea-trial #3 Results, Run 3 UUV Track Results...83 Figure 45. Sea-trial #3 Results, Run 3 GoPro Cam Photo of Docking Station During Fly-By...83 Figure 53. Sea-trial #3 Results, Run 3 Zoomed-in UUV Track Results...84 Figure 54. Sea-trial #3 Results, Run 3 UUV Vertical Track...84 Figure 55. Sea-trial #4 Results, Run 3 UUV Vertical Track...85 Figure 56. Simulation functional diagram x

11 Figure 57. Vertical Cross-track Error Results, Sim Figure 58. Vertical CT error PDF estimate, Sim Figure 59. Sea-trial #3 Results, Run 3 UUV Vertical Track...95 Figure 60. Horizontal CT error results, Sim Figure 61. Horizontal CT error PDF estimate, Sim Figure 62. CT error polar plot, Sim Figure 63. Vertical CT error results, Sim Figure 64. Vertical CT error PDF estimate, Sim Figure 65. Horizontal CT error results, Sim Figure 66. Horizontal CT error PDF estimate, Sim Figure 67. CT error polar plot, Sim Figure 68. Vertical CT error results, Sim Figure 69. Vertical CT error PDF estimate, Sim Figure 70. Horizontal CT error results, Sim Figure 71. Horizontal CT error PDF estimate, Sim Figure 72. CT error polar plot, Sim Figure 72. Vertical CT error results, Sim Figure 73. Vertical CT error PDF estimate, Sim Figure 74. Horizontal CT error results, Sim Figure 75. Horizontal CT error PDF estimate, Sim Figure 76. CT error polar plot, Sim Figure 77. Vertical CT error results, Sim Figure 78. VCT error PDF estimate, Sim Figure 79. Horizontal CT error results, Sim Figure 80. Horizontal CT error PDF estimate, Sim Figure 81. CT error polar plot, Sim Figure 82. Vertical CT error results, Sim 1 mobile homing Figure 83. Vertical CT error PDF estimate, Sim 1 mobile homing Figure 84. Horizontal CT error results, Sim 1 mobile homing Figure 85. Horizontal CT error PDF estimate, Sim 1 mobile homing xi

12 Figure 86. CT error polar plot, Sim 1 mobile homing Figure 87. Vertical CT error results, Sim 2 mobile homing Figure 88. Vertical CT error PDF estimate, Sim 2 mobile homing Figure 89. Horizontal CT error results, Sim 2 mobile homing Figure 90. Horizontal CT error PDF estimate, Sim 2 mobile homing Figure 91. CT error polar plot, Sim 2 mobile homing Figure 92. Vertical CT error results, Sim 3 mobile homing Figure 93. Vertical CT error PDF estimate, Sim 3 mobile homing Figure 94. Horizontal CT error PDF estimate, Sim 3 mobile homing Figure 95. Horizontal CT error results, Sim 3 mobile homing Figure 96. CT error polar plot, Sim 3 mobile homing Figure 97. Vertical CT error results, Sim 4 mobile homing Figure 98. Vertical CT error PDF estimate, Sim 4 mobile homing Figure 99. Horizontal CT error results, Sim 4 mobile homing Figure 100. Horizontal CT error PDF estimate, Sim 4 mobile homing Figure 101. CT error polar plot, Sim 4 mobile homing Figure 102. Vertical CT error results, Sim 5 mobile homing Figure 103. Vertical CT error PDF estimate, Sim 5 mobile homing Figure 104. Horizontal CT error results, Sim 5 mobile homing Figure 105. Horizontal PDF estimate, Sim 5 mobile homing Figure 106. CT error polar plot, Sim 5 mobile homing xii

13 1. INTRODUCTION 1.1 Project Background The overall goal of this joint project is to develop and utilize a modular design that allows an unmanned surface vehicle (USV) to operate with multiple USV s, and have the payload capacity and capability to carry, launch, home, dock, recharge, and download data with an unmanned underwater vehicle (UUV). This project is a collaborative research performed by several universities as a part of the Atlantic Center for the Innovative Design & Control of Small Ships (ACCeSS) program. The program consists of faculty, scientists, and graduate students from various universities (Stevens Institute, Webb Institute, University College London, United States Naval Academy, United States Naval Postgraduate School, George Mason University, Florida Atlantic University), with specializations ranging from hydrodynamics to system and sensor modeling, simulation and implementation. The research promotes advancement of integrated hull design, and shipboard automated systems to perform cutting edge tasks for military and industrial applications. Experimental testing for this project is implemented using two unique vehicles. The Wave Adaptive Modular Vehicle (WAM-V, shown in Figure 1) is an unmanned, catamaran style surface vessel with shock dampers designed for stability enhancement when carrying 1

14 payloads (such as a UUV). The underwater vehicle used is the Kongsberg Hydroid REMUS-100 UUV equipped with a Hydroid Digital Ultra-Short Baseline (DUSBL) acoustic positioning system (APS) (also shown in Figure 1). The WAM-V is equipped with a homing transponder that the REMUS-100 s DUSBL transceiver acoustically interrogates to estimate the range and bearing of the vehicle relative to the transponder [7]. The WAM- V is also modified with an autonomous Launch and Recovery (LAR) system that will have the capacity to carry the REMUS-100 [16]. By the end of the project life cycle, a fleet of these unmanned vehicle configurations are expected to follow an asset ship as it approaches and enters a sensitive area. The UUVs are to be launched directly from the USVs, allowing them to perform their necessary surveillance routines. Upon completion of the mission, the UUV enters into a retrieval Figure 1: Left Hydroid REMUS 100 UUV, Right (top) Wave Adaptive Modular Vehicle in stationary position, Right (bottom) Wave Adaptive Modular Vehicle while under way. 2

15 mode, which allows the tracking and homing software to position the UUV under its rendezvous point with the USV for docking. A concept of operation is shown in Figure 2. As seen in Figure 3, the horizontal target range for the proposed mission is less than 2000 [m] away and the UUV is less than 40 [m] in depth during homing. As the WAM-V follows an independent trajectory at slow speeds, the DUSBL tracking system aboard the REMUS- 100 transmits a request message to the appropriate homing transponder and waits for its reply. Using the acoustic arrival time-series, the DUSBL receiver calculates an azimuthal angle, elevation angle and range to the moving transponder in body-fixed frame. This information is utilized in a closed-loop control system to home the UUV to its targeted recovery location in a line of sight (LOS) trajectory. 3

16 The process of successfully homing and docking moving unmanned marine vehicles using underwater acoustics has a level of uncertainty associated due the acoustic positioning system s limited ability to dynamically adapt to the environment. This limits the UUV from accurately homing to the desired moving source, thus brings the proposed need to research the range of the rendezvous error expected in homing the REMUS-100 to an unmanned surface vehicle while underway. Figure 2: Image of swarm configuration of USVs transporting, launching, and recovering UUVs. 4

17 ASV following an independent track < 2000 [m] AUV homing in on ASV in a follow the leader positioning track Figure 3: Image of UUV tracking and homing to rendezvous point using DUSBL APS. The system calculates range, bearing, and elevation to the rendezvous point. 1.2 Unmanned Marine Vehicles The term heterogeneous collaboration is an emerging phrase currently being spoken about in the unmanned marine industry. In particular, it is the use of different types of unmanned marine vehicles (USVs, UUVs, amphibious vehicles, etc.) working cooperatively to accomplish a common mission. For this project, two unique heterogeneous unmanned marine vehicles are used: the Hydroid REMUS-100 UUV and the Wave-Adaptive Modular Surface Vehicle (WAM-V). 5

18 1.2.1 Hydroid REMUS-100 UUV The REMUS-100 UUV is a Kongsberg Hydroid product (Figure 4). The REMUS-100 UUV is a fully integrated package containing an acoustic modem, LBL/USBL/inertial navigation systems, Acoustic Doppler system, CTD sensor, side scan imager, bottom seafloor camera, GPS, and RF communication. The system also uses a unique PC and operating interface program called VIP (Vehicle Interface Program) to setup and track the vehicle (with basic commands such as start and stop mission, loiter, etc. commands). The FAU version of Hydroid REMUS-100 has been enhanced with a software upgrade for underwater multi-vehicle collaboration and secondary missions using specialized Remote External Control (RECon) software. The DUSBL navigation system is the primary tool used for homing the UUV to the WAM-V, however, is not integrated yet at this stage of the project. Figure 4: FAU's Hydroid REMUS-100 with stinger capturing device. To program the vehicle, commands are used within the mission planner to program a specific mission. Along-side, there is a Home Transponder function in which forces the vehicle to use the DUSBL-APS to home towards a specified transponder in docking scenario. The UUV is also equipped with a mechanical docking mechanism that allows it to latch to a capturing mount located below the USV [16]. 6

19 1.2.2 Marine Advanced Research WAM-V The WAM-V, is a lightweight, modular, pontoon style vessel with a sleek design made to stabilize heavy payload when operating in surface wave action (Figure 5). The WAM-V s mechanical design is owned by Marine Advanced Research. The electrical hardware and software for the vehicle is an in-house development by Florida Atlantic University [16]. The main objective of the WAM-V in this project is to transit the UUV payload package to a desired location, as well as launch and recover the UUV while underway. The WAM- V is also planned to have multi-vehicle collaboration capabilities in the event of utilizing several USV/UUV configurations to protect a desired asset. Development of the WAM-V control architecture is still under work, along with the LAR system [16]. Figure 5: Marine Advanced Research Wave-Adaptive Modular Vehicle (WAM-V), Unmanned Surface Vehicle. 7

20 1.3 Research Objectives The primary objective of this research is to maximize the surface and sub-surface vehicle rendezvous accuracy using a USBL acoustic positioning system which will provide reliable estimations for autonomous docking and sub-surface vehicle recovery. One of the major aspects expected to hinder the systems performance and rendezvous accuracy is the ability to provide the UUV with a reliable fix of where the USV is located at any given point in time. For this reason, this issue has become a primary design target for this project. The following is a list of objectives to be completed during this thesis: 1. Characterization of signal specifications: obtain knowledge about the acoustic channel and components that will affect signal bandwidth, pulse length and signal strength. 2. Characterization of the USBL positioning system: obtain knowledge about the communication hardware being used, including the size of the array, the number of elements being used, the directivity of the system and its sensitivity to background noise. 3. Characterization of the acoustic channel: model the acoustic channel that the system is operating in by utilizing the CTD sensor onboard the REMUS 100 and a recorded sound velocity profile. 4. Assuming the WAM-V is tracking a simple trajectory at a slow speed, simulate the homing and rendezvous of the REMUS-100 underwater vehicle and WAM-V surface vehicle while in motion. The algorithm shall calculate the expected error induced by sound refraction and estimate the error circle produced around the rendezvous point. 8

21 5. Validate the simulation results by testing if the homing algorithm on the REMUS-100 works in the field as expected from the simulated mission. 1.4 Scientific Contribution The purpose of this thesis is to evaluate the ability to successfully home to a moving acoustic source using an acoustic positioning system. The causes of such limitations are also investigated. Most importantly, this thesis involves the development of a simulation software that simulates the acoustic channel and evaluates the errors produced by the navigation system. The analysis shall forecast the expected azimuthal bearing, elevation bearing and slant range errors for each fix location, and the overall associated cross-track and along-range homing error. The overall outcome of this thesis will provide an answer to the feasibility of utilizing the Hydroid DUSBL system to successfully home and dock the two moving vehicles in a given hypothetical environment. 9

22 2. LITERATURE REVIEW The following paragraphs in this section shall briefly explain the key literature articles that were utilized for this research project. Some of the articles were written by the designer of the REMUS-1000 and articles chosen were obtained from the originators of the REMUS 100 and DUSBL system. Acoustic Positioning Using a Tetrahedral Ultrashort Baseline Array of an Acoustic Modem Source Transmitting Frequency-Hopped Sequences In the article titled Acoustic Positioning Using a Tetrahedral Ultrashort Baseline Array of an Acoustic Modem Source Transmitting Frequency-Hopped Sequences [3], the authors developed a USBL acoustic positioning system used to track underwater objects such as a UUV. Most importantly, the authors developed a theoretical formulation to evaluate the accuracy of the USBL system and provide results based both on simulation and experimental analysis. This article helped in developing the methodology used to evaluate the DUSBL-APS used in this thesis. The scientific contribution of this project is the development of a USBL system which processes a known pre-amble of a communication message using a tetrahedral receiver array. In such case, the preamble is a frequency-hopped multiple shift keyed (FH-MFSK) signal. According to the authors, this signaling technique is very well suited for reliable bearing estimation in the presence of strong multipath interferences. The transmitted 10

23 message is a short Tukey-weighted pulse shifted to specific frequency bands in a specific frequency-happed sequence. The detection process of the USBL receiver requires that the SNR exceeds a specific threshold for accurate detection. The SNR is the power of the received signal plus noise, divided by the power of the noise. The authors then explored the variance of the system when detecting a given angle of arrival. They showed that the bearing estimate to the source is a function of the true bearing to the source, plus a bias due to underwater sound refraction, plus another bias due to the system uncertainty. The range estimation is dependent on how well that the computer can track the time between transmission from the interrogator and arrival of the acoustic sound to the USBL. For the experimental analysis, a USBL with tetrahedral array was built and tested inside an acoustic testing tank. A modem source was placed in a corner of the tank, and the USBL antenna was mounted in the center of the tank where it could be robotically rotated to change its directionality. The results show that at a fixed SNR, the positioning accuracy increases as the number of pings increases. Figure 6 shows the mean estimation error of the simulated data. One can see in the figure that the percent error is expected to drop as the SNR increases. This results shows that the SNR has direct impact on the overall accuracy of the USBL system. The results also show that the system was able to successfully locate the position of a MFSK modem in which part of the message (such as the preamble) is known. 11

24 Figure 6: Mean position estimation error as a function of SNR [3]. Enabling Technologies for REMUS Docking: An Integral Component of an Autonomous Ocean Sampling Network As seen in Enabling Technologies for REMUS Docking: An Integral Component of an Autonomous Ocean Sampling Network [16], the authors provide a description of the different technologies that are aboard the REMUS-100 UUV; in particular they go over the use of the Hydroid DUSBL-APS and its performance in the field during applications such as static docking systems. For this project, the research group from WHOI utilizes the 12

25 REMUS-100 UUV, equipped with a DUSBL navigation system, to home and dock with a station moored to the seafloor. In particular, they go over the need of autonomous docking stations, and how well they work. The major reason for a docking system is to enable the ability to reduce significant costs of operation in current ship deployment and networking/surveying applications. In this case, a homing and docking system must have an accurate navigation system, a means to fasten the vehicle to the dock, the ability to recharge the AUV underwater, have communications for download/uplink and diagnostics, and have a way to track the vehicle to estimate the mission performance. A couple of known projects similar to the one presented in this thesis have also been successfully demonstrated, including: a system developed by MIT and WHOI, which uses a USBL and scissor latch system to allow the vehicle to dock from all angles [16]. Another project was conducted by Space and Naval Warfare Systems Center Pacific (SPAWAR PAC) with a system called the Flying Plug, and was equipped with an optical system for data transfer [16]. With this idea, the Odyssey vehicle was created and equipped with a USBL system that docks into a conical garage. In this paper, the research group decided to utilize a hybrid form of these two approaches. They docked the REMUS 100 equipped with a USBL into a conical shaped static station. The dock had a 1-meter diameter cone and 0.25-meter garage. The garage was wrapped with copper coils used to measure the location of the vehicle once it entered the garage. According to the authors, a major source of error in the homing accuracy is due to angular measurements. Essentially, they explained that error in the compass heading is a major 13

26 contributor to homing inaccuracy due to bias issues, and also error is due to the noise in the acoustic channel and the resolution of the system (number of fixes per meter). Accuracy of the homing system is said to be +/- 0.5 meters in range and 1-3 degrees in elevation and azimuth. The docking algorithm goes as followed; the vehicle navigates to a point 50 meters in front of the dock and then engages the USBL. It requires REMUS to follow a track line and it will crab if current is present. Once REMUS is within 2 meters of the dock, it straightens its fins and continues to drive itself forward towards the dock for 10 seconds. The vehicle then turns the rudder hard over, while continuing to thrust. If the vehicle sees the compass turns to 90 degrees, then it knows it has missed the dock and will then return to the start point to retry the homing. If it does not turn, it continues to hold its thrust until the latching mechanism is enabled. The at-sea testing took place off of Tuckerton, NJ. The test was conducted in LEO-15 Long Term Environmental Observatory, 15 meters water depth. At LEO-15, the vehicle docked in 21% of the attempts (without any retry). It succeeded, with retries, in 47% of all missions. The same tests were conducted later in the year at WHOI in a controlled environment. The vehicle docked in 87% of the missions. After even further development, the system was tested again, this time at Gulfport. The vehicle docked in 91% of the missions. From a controls standpoint, the vehicle was flown using two different methods: (1) The vehicle maintained a constant altitude using the DVL system to control the vehicles vertical 14

27 cross-track error; (2) The vehicle maintained depth within a pre-defined depth window. According to the reports, the second method was more effective. Table 1 WHOI/Hydroid Homing Test results Results Attempts % Missions % LEO-15 20/97 21% 20/43 47% WHOI 13/19 68% 13/16 87% Gulfport 30/50 60% 30/33 91% Underwater Mobile Docking of Autonomous Underwater Vehicles In the paper, Underwater Mobile Docking of Autonomous Underwater Vehicles [4], another research group, which consists of the Hydroid research and development engineers, performed a similar project to the one proposed for this thesis. For this project, the REMUS 100 equipped with a DUSBL APS docked to a mobile platform dragged by a surface ship (see Figure 7). The docking system was a depressor style system, so that its depth was constant and solely function of the towing speed. The concept of utilizing the DUSBL APS as a primary navigation sensor is not fully accurate. For this project, the DUSBL APS is primarily used a reference beacon to determine slant range to the moving source. The azimuthal bearing and desired elevation is determined by transmitting the moving sources heading (provided by an onboard compass) and its depth as a part of the acoustic message used for interrogation to determine the slant range. 15

28 As explained, the mission consisted of the vehicle moving and docking with the moving platform with no prior knowledge of the location of the dock. The first task is for the vehicle to loiter about an area in which the moving platform is expected to pass. The moving platform s heading is transmitted as part of the acoustic reply and its slant range calculation is determined using time interrogation. If the transponder s signal is detected by the DUSBL, the moving source s track heading becomes an intercept point for the UUV to dead reckon towards. The UUV needs to be within 200 meters behind the dock to track it. Once within range, the DUSBL sensor provides angular azimuthal measurements to the moving source. If the vehicles fails to dock, it will turn around and approach to within 100 meters of the dock and reattempt. To determine that the UUV failed to dock, the Long- Baseline (LBL) transponders can be used. Once the vehicle makes a successful dock attempt, it performs several tests to confirm it is locked in. Testing was performed in Buzzards Bay, Massachusetts. The group tested the UUV homing performance in down, head, and cross current conditions, in which all provided consistent results. The testing results suggest that the UUV homes best when it can follow the docking station as soon as possible, providing an interception point within a few acoustic navigation updates. Unfortunately, the shallow depths of Buzzards Bay prevented the docking system to be towed at deeper depths to evaluate if results would get better or worse. As a testing requirement, the UUV was given 10 attempts for docking per mission, and the UUV was able to dock within the first two attempts in 77% of the 11 conducted tests. Figure 8 shows the results of a demonstration that was given in Weymouth, England. 16

29 Figure 7: The figure above shows the typical equipment used to dock the REMUS 100 to a moving platform. Figure 8: The image above show the test results conducted in England. The vehicle missed the first two docking attempts and then make entry on the third attempt. 17

30 3. SCIENTIFIC BACKGROUND The implementation of unmanned marine vehicles that can cooperatively work together to perform a mission is fairly novel research. Homing and docking a UUV onto a moving USV can be a daunting task due to maintaining vehicle stability in current and wave action, as well as maintaining acoustic communication for a reliable navigation source. The governing equations presented in this thesis incorporate effects due the acoustic channel utilized and the computing uncertainty generated by the system itself (signal processing error). Vehicle pitch and heading rate are included for navigation purposes, however, hydrodynamic effects of the vehicles are neglected. The positioning errors produced have impact on the slant range r s, azimuthal angle θ, and the elevation angle φ ; which are spherical estimates of where the UUV thinks that the moving source is currently located. At the most fundamental level, the acoustic channel directly affects the signal to noise ratio (SNR), which is the acoustic signal quality; it also affects the direction of acoustic arrival (DOA) at the receiver, and the acoustic time of travel (TOT) estimate, all of which are the most important parameters that are needed for underwater acoustic positioning. 18

31 3.1 The DUSBL Acoustic Positioning System The DUSBL-APS is the primary sensor system for navigating the UUV to the mobile USV docking station. There are two main components that are utilized in the DUSBL-APS: 1. Acoustic modem interrogator source 2. USBL receiver array with processing electronics The acoustic interrogator source is the same acoustic source used to communicate with the vehicle via an acoustic modem Ranger [7]. To maximize the data throughput and reduce false alarm signals during acoustic communication, the REMUS-100 s acoustic modem source transmits a pulse train with a unique carrier frequency for each device it deliberately attempts to communicate with [1]. The DUSBL-APS, as seen in Figure 9, has been designed to use the arrival time-series to determine range and bearing to the local transponder s reply message. The expected reply signal is a 4 ms, khz pseudorandom noise (PN) sequence used for Long-Base-Line (LBL) and Ultra-Short-Base-Line (USBL) navigation scenarios. 19

32 Figure 9: Hydroid DUSBL Acoustic Positioning System. Left - DUSBL functional diagram Right - Hydroid DUSBL receiver Courtesy of Hydroid. The DUSBL-APS has a four channel receiver mounted to a processing board that has been potted into an aluminum enclosure and connects to the front of the UUV. Each element in the array has been specifically placed slightly larger than ½ wavelength of the operating frequency apart from each other [7]. Typically the elements are placed no larger than ½ wavelength to avoid aliasing, however, a compromise was made to increase the beam resolution. To avoid aliasing, the system has a virtual window, or listening angle, that is smaller than the actual theoretical beam angle that the receiver can handle. The virtual window only allows messages to be received within ±17.5º from bore site. To allow for tracking wider angles, the host has the ability to steer the center up to an additional ±17.5º, equating for a maximum ±35.0º of listening coverage. The array pattern used is a tetrahedral configuration, in which the vertical baseline measures the elevation DOA, and the horizontal baseline measures the azimuthal DOA. A valid fix is accepted when the received message is within the effective listening cone and above a given detection threshold (DT) [8]. 20

33 In terms of theory of operation, the interrogator transmits a request message to trigger the homing transponder to send a reply message back. When the reply message is sent from the homing transponder, the DUSBL receiver located on the front of the UUV can interrogate the message to generate a temporal and spatial representation of the received ray. The time series provides phase information between the arrivals at each element in the receiver antennae to calculate the azimuthal and elevation DOA. Using the vehicle s state information from the inertial navigation system (INS) to back out the vehicle s attitude, the DUSBL system then calculates the best approximate fix of where the homing transponder is located [8]. The DUSBL receiver s digital processing electronics contain a variable gain pre-amplifier for each channel, band-pass filter, digitizer, match filter, signal detector, and digital signal processor (DSP) [7]. Figure 10 shows the system processing. In order to unmask the desired signal from the background noise, it is amplified and filtered. The amplification has two stages of automatic variable gain (AVG) controlled by the DSP and the signal is band-pass filtered with respect to the desired operating bandwidth B, as derived from the DUSBL manufacturer specifications. The signal in each channel of the hydrophone is stored to a buffer one point at a time. Once filled, the oldest data point is removed first, as in a firstin-first-out (FIFO) buffer scheme. The sampling rate F s is 100 ksps. Each digitized signal coefficient removed from the buffer gets passed through a matched filter for decoding and then detection processing [7]. 21

34 The matched filtering process is a real-time processing of the digitized signal to compare the inputs with a desired reference signal stored in memory. The output of the matched filter maximizes the SNR of the received signal [1]. The signal detector compares the SNR output from the matched filter against a detection threshold DT that is also saved in memory. If the SNR exceeds the detection threshold, then a detection is confirmed; if the SNR is below the detection threshold, then the acquired signal is dismissed and the system refills the receiver buffer in time to process the next set of data [20]. s(t) 2-Channel Auto- VGA Band-pass Filtering x(n) x(n 1) Pattern Matched Filter DT x(n N Template Buffer Figure 10: DUBL-APS signal processing flow diagram. The DUSBL provides three body-fixed components of information: slant range, azimuthal bearing, and the elevation bearing to the homing transponder [7]. From a dynamical modeling standpoint, focus is placed on accurately calculating the DUSBL-APS azimuthal and elevation estimated direction of arrivals. The slant range calculation is directly a function of the average acoustic time of travel from the source to the DUSBL receiver, the estimated sound speed within the local fluid, and accuracy of the system [7]. The azimuthal and elevation bearings are more complex to model due to multivariate uncertainties in the channel estimation. The amount of uncertainty in estimating the true bearing to the homing transponder is coupled to the amount of sound ray refraction in the seawater and self-noise of the system [3]. This occurs in 3D space, consisting of a vertical and horizontal plane that 22

35 contains a body-fixed range and bearing to the moving source. Figure 9 shows a free-body diagram of an acoustic ray traveling from a point source to a point receiver. The image is made to exaggerate the ray refraction, however, strongly depict that the estimated source position by the DUSBL system is directly coupled to the DOA of the refracted ray. Z Acoustic ray estimated by USBL Vertical plane and DOA 3D Positioning Error Homing Transponder DUSBL receiver Horizontal plane and DOA Y θ φ Actual acoustic ray path X θ = estimated azimuthal angle φ = estimated elevation angle Figure 11: Three-dimensional free-body diagram of ray transmission from homing transponder to the DUSBL receiver. Shown is the ray path from the source versus the estimated ray path from the source. The homing transponder has a processing turnaround time required to respond to the DUSBL receiver which is the time required to process the interrogation message, store a new message to the transmit queue, and finally send the message. This time is approximately 800 ms according to manufacturer specifications [7]. This time, however, is deducted by the DUSBL-APS in reality after it has calculated the two-way-time (TWT) of travel, and so it does not needed to be considered in our model. There is, however, an important amount of time considered in the model due to the uncertainty in processing time it takes to detect the signal within the expected time window this is the signal detection 23

36 time jitter. The jitter, according to manufacturer specifications, ranges within -1 ms and 1 ms of the average time it take to detect the signal [7]. 3.2 Key Parameters in USBL Positioning The underwater environment and the system s ability to adapt to it plays a critical role in the homing performance since underwater acoustics is the primary means for navigation. There are three major sources currently known to affect the quality of underwater acoustic positioning: System Update Rate The system update rate is essentially the "heart beat" of the communication link, and in turn affects the accuracy of the UUV s rendezvous. If the update rate is too slow then the UUVs homing performance drops. It is expected, however, for the update rate to increase as the vehicle gets closer to the moving source since the acoustic TOT decreases Intermittent Communication Dropout The acoustic source on the UUV can be modeled as an omnidirectional point source. Theoretically, the acoustic interrogation message should be heard from all directions. Technically, the acoustic source on the UUV that transmits the interrogation message is located on the bottom side of the vehicle, which acts as a baffle and casts an acoustic shadow over the acoustic rays that would normally transmit directly towards the surface. 24

37 The homing transponder located on the USV is also modeled as an omnidirectional system in which can transmit and receive messages from all directions. The USBL receiver equipped on the UUV, however, has a forward-looking directional listening window that must be pointed towards the moving source within its effective spatial sensitivity in order to receive a message. Intermittent directional dropout occurs when the orientation of the array on the UUV is out of scope, or off-axis, of the transmitted sound from the moving source. This affects the number of successfully acquired acoustic positioning fixes that the UUV can generate during a homing attempt, which in turn affects homing spatial and temporal resolution. For example, as the UUV acquires a fix, it may move tens of meters before it receives the next response, which may have caused the vehicle to overshoot its set point before the system even realized it. However, as the system update rate increases (as in the UUV is getting closer), the error circle around the rendezvous point is expected to get smaller, thus the overall cross-track and along-range errors shall decrease. If it takes too long for the vehicle to reestablish communication, this may cause a mission time out or may prevent the vehicle from having the opportunity to recuperate its positioning track Subsurface Obstacles The underwater world is populated with environmental effects that can interfere with underwater sound propagation: 25

38 The presence of a thermocline in the water column has the largest effect on accuracy of the acoustic positioning system. The system s accuracy on range and bearing estimations are highly dependent on knowledge of the true TOT and true DOA of the received acoustic ray. These measures are dependent on knowing the true sound speed of the ray, which is highly dependent on the water temperature and many other non-linear parameters. This in turn, causes the acoustic ray to refract, or bend in the wave guide thus corrupting the true bearing to the source. In areas of the water column with large gradients, significant amounts of refraction in both upwards and downwards direction generate shadow zones in which no sound will be heard if the receiver is located in that area. Self-noise is also of an underlying issue that affects the system. Many electronics aboard emit electrical noise that may exist at or near the operating frequency of the acoustic transmission. Also, acoustic imaging systems that may be running contain frequency harmonics that can also interfere with acoustic navigation system. The self-noise, along with background environmental noise masks the incoming signal with unwanted distortion and creates issues for signal detection. Large schools of fish passing in the front of a transducer while communicating with a UUV will block sound transmission and interrupt the communication link with false alarms or dismissals. Snapping Shrimp pinch their appendages, producing a large amount of the underwater ambient noise found in the Ft. Lauderdale area. Bubbles produced by the UUV, USV, and subsurface bottom will cause sound scatter and disrupt the signal quality. 26

39 Bio-fouling caused by plant and algae growth on the transducers if left submerged for a long period of time can also dramatically affect signal to noise ratio and overall system estimations Sound Propagation To best simulate sound propagation, the source interrogation and transponder reply messages are modeled using a finite element ray (FER) tracing algorithm; Bellhop, as provided by the Ocean Acoustic Toolbox [9,13,14]. Using open-source software allowed research development to be focused on analyzing the signal detection component of the APS, and deriving a set of dynamic equations that can be used to forecast the navigation performance during a typical homing routine. The FER trace output provides necessary propagation parameters acquired from the channel response (CR) given a prescribed sound velocity profile (SVP); it is used to determine the eigenray time series (the most direct ray) and provide arrival information with the presence of ray refraction and ray multipath. In particular, the output provides for each received ray the normalized receive amplitude, phase, TOT, vertical DOA ( φ ), number of surface bounces (SB), and number of bottom bounces (BB). Channel details are also accounted for by providing half-space attenuation properties, and transducer directivity. Details are discussed in the simulation results section. 27

40 Table 2 Bellhop FER-Trace Output Parameters Parameter Symbol Description Unit H(f, r) Normalized channel No units response amplitude τ Phase Degrees TOT Acoustic time of travel Seconds (one-way) φ Vertical Direction of Degrees arrival SB/BB Number of ray bounces Number present/range Simulations are based on operations in shallow water along the coast of South Florida. In 2003, a series of measurements were conducted at the South Florida Ocean Measurement Center (located just south of the homing test site) to monitor the environment along the coast. Data collected included conductivity and temperature measurements using MicroCAT conductivity and temperature sensors placed at various depths. The data were merged to develop the local sound velocity profile as seen in Figure 12. Although not recent, the data were collected near the test field used in this research at a similar time of the year. The image on the left shows the measured data with a corresponding SVP and the image on the right shows the profile that was generated in Matlab and used for the homing simulator. Figure 13 is the ray-trace Eigen plot in which is generated using the prescribed SVP parameters for the Ft. Lauderdale area. The figure shows the refracted Eigen ray (in red) and the refracted multipath ray (in black and blue). A reference line (in green) is also shown to represent a non-refracted direct ray; although not an actual simulated ray, it is used to exploit the amount of curvature in the refracted Eigen rays. 28

41 Figure 12: Left Data collected from the South Florida Ocean Measurement Center using conductivity and temperature sensor. Right Sound velocity profile generated by the data and used in simulation. Figure 13: Ray trace simulation using prescribed sound velocity profile for the Ft. Lauderdale area. Red traces show the Eigen rays (most direct rays), and the black/blue traces represent the multi path rays. The green trace (not actually simulated) represents a direct arrival with no refraction, used as a reference line to exploit the curvature present in the Eigen rays. 29

42 In order to accurately model the detected amplitude of the DUSBL receiver, spreading attenuation is properly accounted for in the transmission loss (TL) equation. The seawater absorption coefficient α is considered during the signal processing section. For accuracy, it was estimated within 10 bins in a 1000 Hz band PSD for carrier frequencies of interrogation message and reply message using the following equation [10], α = F 2 ( F F 2 exp ( Z 6 ) ) (db km) (1) Where, F is the frequency in khz. The magnesium sulfate term, MgSO4 (second to the right), is compensated for depth Z. The received echo level EL is estimated by deducting the expected transmission loss from the target s source level (SL). The total transmission loss is a combination of geometrical spreading losses and seawater absorption. The geometrical spreading loss can be estimated by utilizing the normalized channel response from Bellhop, giving, TL(f, r) = 20 log ( 1 H(f,r) ) db (2) where, H(f, r) is the normalized channel response provided by Bellhop, and f is the carrier frequency in khz. The received echo level (EL, neglecting seawater absorption) as a function of frequency and range is estimated as, EL (f, r) = SL TL(f, r) db (3) where, SL is the known source level of the transmitting device. 30

43 4.1 Architecture 4. SIMULATION SOFTWARE The simulation contains three main components required to demonstrate underwater acoustic homing to a moving source: a UUV transceiver model (as applied on the UUV), the sound propagation model (as provided by Bellhop), and the homing transponder model (as applied on the USV docking system). A functional flow diagram is shown in Figure 14. Each block represents a function in the simulation. Note that the transponder and USBL antenna use the same signal detection function; its use has been optimized for the request and reply message s carrier frequency respectively. The DUSBL transceiver model contains a function to generate a position estimate to the moving source once the reply signal from the homing transponder has been received and detected. The homing simulation software has been designed to emulate the DUSBL system installed on the REMUS-100 UUV. The simulation emulates the system architecture as described in Section 3. The program is written in Matlab and can be interfaced as a sub-system component used to acquire an accurate position estimation of a moving underwater acoustic source in a larger-scale simulation program. The simulation function AquireTargetFix( ) can be polled for single-threaded output or used for asynchronous continuous-stream output at a user-defined update rate. The output of this simulation function is a localized position estimation of the moving target in rectangular and spherical coordinates USV[X, Y, Z, r, θ, φ ].To demonstrate the feasibility of navigating the UUV 31

44 to a moving source, a procnav( ) (process navigation) function was created to simulate the heading and pitch rate of the REMUS-100 with a desired forward speed to provide vehicle motion. The USV s motion follows a simple straight track-line (not limited to any distance) independent of the UUV s navigation. The heading and pitch goal of the UUV is always in the direction towards the position estimate provided by the DUSBL-APS model (towards the sound source). 32

45 Figure 14: Functional architecture of the simulation program. 33

46 4.2 Mathematical Derivations Signal Detection Model The acoustic signals simulated are assumed to be broadband (within khz) and contain additive white Gaussian noise. To generate an accurate spectral model of the signals transmitted between the DUSBL antennae and transponder, the interrogation and reply messages were recorded using an underwater acoustic acquisition system in the Seatech Marina (courtesy of Seaton [17]). The recorded signal is used to create a normalized signal that is scaled to the simulated amplitude expected at the DUSBL antennae and homing transponder, given a particular environment and the device positions. To model the signal detection process, a power-spectral density (PSD) profile of the scaled signal is generated using Welch s periodogram function in Matlab. The output is broken into ten equal bins across the desired bandwidth and the bins located within the operating band are summed to measure the total squared amplitude of the received signal. Finally, this value is used along with the expected ambient noise PSD to determine the SNR. If the SNR exceeds a probability-based detection threshold, then the signal is considered to be detected Generating a Normalized Signal To generate a normalized signal, the recorded underwater raw signal was first band-pass filtered (BPF) using Matlab to remove spectral power outside of the desired frequency band, 22 khz 30 khz. Detailed information about of the filter design can be found in the 34

47 Appendix. To create the normalized signal s (n), the band-pass filtered output signal s(n) is divided by its sample standard deviation, such that, s (n) = s(n) ( 1 (4) N N n=1 (s(n) ) 1/2 Here, n is the discrete time index and N is the maximum number of samples collected in the record. Figure 15 shows a plot of the original signal (in blue) and the band-pass filtered signal (in red). 2.5 FFT of Original vs. BP-Filtered Message 2 Original Signal Bandpass Signal X: 2.865e+04 Y: Y(f) Frequency (Hz) x 10 4 Figure 15: FFT of the original signal versus the band-pass filtered message. 35

48 Scaling the Signal Spectrum and Estimating the SNR This process is the first step towards signal detection the ability to unmask the desired signal from additive noise. The acoustic signal quality between the homing transponder and the DUSBL transceiver is directly measured using the in-band SNR, which is modeled as a ratio of the received signal power to the detected ambient noise power in the frequency band of the received signal. To scale the normalized signal, the echo level term derived in (3) is corrected for the directional attenuation index b r (θ, φ ) and processing gain specific to the Hydroid APS receivers. The processing gain PG represents the gain in SNR produced by the correlator. It can be estimated as [19], PG = 10 log BT db (6) where B is the filter bandwidth in Hz and T is the estimated signal duration. In the case of a carrier wave, PG = 0 db; whereas, PG > 0 db if the received signal is modulated in frequency. The homing transponder has an omni-directional antenna so that b s (θ, φ ) 0 db. Conversely, the DUSBL antenna has a 35 degree listening cone: the ray must arrive within in this cone in order to be detected. The DUSBL theoretical beam pattern is modeled as: b r (θ, φ ) = A θ + A φ db, for W B 2 θ W B 2 W B 2 φ W B 2 (7) {, else where 36

49 Where A θ and A φ are the attenuation coefficients in the direction of arrival given by θ and φ. Figure 16: Approximated DUSBL beam pattern. Courtesy by Seaton [16]. Figure 16 shows a polar plot of the FAU USBL beam pattern. The lookup table categorizes the incoming arrival angle to provide attenuation coefficients A θ and A φ based on calibration provided by the beam pattern. The PSD of the scaled received signal is generated using a periodogram approach (pwelch() function in Matlab), such that, G R (f, r, θ, φ ) = 1 E { F 2T {10(EL (f,r) + b r (θ,φ )+PG)/20 s (n)} 2 } µpa 2 Hz (8) The PSD is distributed over ten frequency bins with equal width. The PSD within each bin is multiplied by the frequency resolution f and the seawater absorption coefficient 37

50 α (as derived in Equation 1, chapter 3.1.1) to determine the acoustic power within that bin. Then the power within each bin is cumulated to finally provide a total power (squared acoustic pressure) received across the frequency interval [f o 5 f, f o + 5 f], such that, p 2 rms (G R, f o, f, k) = G R (f o + k f, r, θ, φ ) f 10 α(f o+k f) K=5 10 k= 5 μpa 2 (9) where f o is the signals carrier frequency and k is the frequency index. The broadband background noise is an isotropic sound distributed fairly evenly within the system bandwidth [19]. According to reference [19], the average noise PSD or noise spectrum level (NSL) for a signal at 25 khz is approximately 40 db re 1 μpa/ Hz in the Ft. Lauderdale area with wind speeds about 7-10 knots. Intermittent noise can be found up to 5-10 db higher. The detected noise level (DNL) received within the operating bandwidth B can be formulated as, DNL(NSL, B, b r (θ, φ )) = NSL +10 log(b) + b r (θ, φ ) db (10) The noise power σ n 2 is then, DNL(NSL,B,br(θ,φ )) σ 2 n (DNL(NSL, B, b r (θ, φ ))) = μpa 2 (11) 38

51 Where NSL is the average noise spectrum level (in db re 1 μpa/ Hz) across the frequency band B, and b r (θ, φ ) is the directivity index for the DUSBL receiver model (recall that directivity is neglected for the homing transponder model). Finally, using the detected signal power and detected noise power, the SNR in db can be formulated as, SNR(r, θ, φ, DNL) = 10 log ( p rms 2 (G R,f o, f,k) + σ 2 n (DNL(NSL,B,b r (θ,φ ))) ) db (12) σ n 2 (DNL(NSL,B,b r (θ,φ ))) Estimating the Probability of Detection The signal detection is simulated using a stochastic, performance-based analysis approach the Neyman-Pearson method [4]. The probability of signal detection is calculated, along with the probability of a false alarm. If the SNR exceeds a specified detection threshold then the signal is detected. To relate the SNR value to a detection threshold, the SONAR equation is utilized to show that the minimum detected echo level EL min must exceed the detected noise level DNL plus the detection threshold DT to satisfy the detection criteria to be satisfied. In such case, EL min > DNL + DT = 10 log(γσ n ) db (13) Where, γ is a multiple of the noise standard deviation σ n. Thus, 39

52 γ = 10(DNL+DT) 20 σ n (15) γ is analogous to the minimum SNR for the detected pressure p rms to satisfy the detection criteria, such that, p rms γσ n = 10 EL min/20 μpa (16) The condition for signal detection is, SNR(r, θ, φ, DNL) > 10 log (γ 2 ) db (17) The gain value γ tunes the sensitivity of the detection model. If set too low, the model is susceptible to false alarms. If set too high, the detection model limits the operating range of the DUSBL [4]. Once the detection gain has been determined, the model performs a statistical analysis to generate the likelihood that a detection has been accurately made. A graphical formulation of this concept is shown in Figure 17. The figure displays the probability density functions of the noise only and of the signal-plus-noise, after matched filtering. This figure also p 1 (β) p 0 (β) < P FA 1 P FA p 1 (β) p 0 (β) P FA 1 P FA p 0 (x) p 1 (x) False Dismissal No Detection Detection False Alarm Average Signal-plus-Noise Figure 17: Statistical probability density function of the signal and signal-plus-noise; detection threshold. 40

53 shows the areas corresponding to the probability of false alarm and false dismissal. These two probabilities are functions of γ. The probability density function (PDF) of the signal being detected can be represented as Gaussian distribution, such that, p 1 (p rms (r, θ, φ, DNL)) = ( p rms(r 1,θ,φ,DNL) prms) 2 σ n 2 e 2σn 2 (18) The probability of detection is formulated as, P D = γ p 1 (β) dβ = erf ( 10 DNL+DT 20 p rms 2 DNL ) (19) Periodic transponder interrogation and replies are tested using (19). When the vehicle interrogates the transponder, the simulator produces a random value, uniformly distributed between 0 and 1. If this value is larger than P D, the simulator interprets the trial as a miss, and the interrogation fails. If the interrogation succeeds, the reply is tested the same way. Successful interrogation and reply are required to generate a navigation update. Note that the simulator does not simulate the false alarms at present. When using the APS as a primary navigation source, the overall homing performance becomes directly coupled to the SONAR performance due to the system update rate. In such case, the navigation performance inherently depends on the signal quality and direction of arrival due to the spatial sensitivity of the antennae. The DUSBL-APS update rate increases as the UUV approaches the homing transponder due to shorter travel-time and larger desired signal spectral power, which leads to increased detection probability and 41

54 thus more valid fixes. As shown in the background section, the increase in valid fixes improves temporal and spatial resolution, as well as reduce unwanted navigation overshoot Position Estimation Model The USBL-APS produces a vector S[r, θ, φ ], which represents the spherical coordinates of the transponder in the fixed-body frame of the UUV Estimating Range to Source Mathematically, the slant range estimate r provided by the APS is the average one-way travel time of the acoustic ray, multiplied by the local sound speed which is preprogrammed during mission planning [7]. Since the sound speed is assumed to be constant and cannot be adjusted in real-time, as in a closed-loop system, there can be an error as much as m/s in the sound speed used for ranging estimation, unless the actual sound profile is isotropic and the user accurately provides the local average. Although possibly negligible, this produces an error which causes the APS to either overestimate or underestimate the true slant range to the source. To match the design of the Hydroid DUSBL-APS, the sound speed value C was held constant in the simulation as well. The time estimate also has a latency due to detection jitter of the order of milliseconds, as provided by the manufacturer specifications. This jitter is due the matched-filter processing time [16]. The simulator models the uncertainty by generating a uniform random value 42

55 within a dynamic range specified within < t 1,2 < [sec]. The slant range estimated by the APS can be modeled as, r s(t 1(r ), t 1, T 2(r ), t 2 ) = C 2 [T 1(r ) + t 1 + T 2(r ) + t 2 ] (20) where C is the local sound speed, T 1 is the acoustic travel time from USBL to moving source, T 2 is the acoustic travel time from moving source to DUSBL receiver, and t 1, t 2 is the system jitter associated with T 1 and T 2, respectively Estimating Bearing to Source The bearing estimate processed by the APS provides the body-fixed azimuthal angle estimate θ and the elevation angle estimate φ to the homing transponder. The errors produced in the bearing estimates are primarily due to refraction caused by the sound speed profile c(z) in the acoustic channel and the error due to system uncertainty (the uncertainty is modeled post Bellhop propagation) [3]. The nominal bearings θ and φ (actual DOA of refracted ray) are modeled as the true bearing to the source θ s and φ s, plus an error angle due to refraction δθ, δφ respectively. The final estimated bearings to the source are modeled as the nominal bearing θ and φ, plus the sensor s uncertainties, θ and φ. As given in the manufacturer specification, the DUSBL-APS has an elevation accuracy such that 2 θ 2 deg and 2 φ 2 deg. This is simulated as a uniformly distributed random value. 43

56 If the pitch and roll are very small and we assume constant sea state, variable depth z, sound velocity profile c(z) and range r, then the estimated bearing angles can be defined as a function of the nominal DOA, range, sound speed profile, and the signal-to-noise ratio: [ θ (θ s, r, c(z), SNR) φ (φ s, r, c(z), SNR) ] = [ θ s(r, c(z)) + δθ(r, c(z)) + θ(θ, SNR(r, θ, φ )) φ s (r, c(z)) + δφ(r, c(z)) + φ(φ, SNR(r, θ, φ )) ] (21) Appropriately applying equation (22) to the homing application, we can define bearing state to the source as, [ θ (θ s, r, c(z)) φ (φ s, r, c(z)) ] = [ θ (θ s, δθ) + θ φ (φ s, δφ) + φ ] (22) Figure 18 below shows a free body diagram of the UUV attitude (heading, pitch, and position) and defines the directions of arrivals from the acoustic source. X (North) Refracted Ray φ UUV Y (East) Ψ UUV φ Acoustic Source Z (Down) θ Y (East) X (North) Z (Down) Figure 18: Free-body diagram of REMUS-100. The image shows the heading, pitch, and vehicle position and also defines the directions of arrival from the acoustic source. 44

57 4.1.3 UUV and USV Navigation Model The purpose of this research is not to study low-level control of the UUV; however, a simple navigation controller model is needed to simulate the homing performance when using a DUSBL-APS. The control scheme could only be assumed since the specifics are proprietary information owned by Hydroid. This section will explain the governing equations used to formulate the UUV navigation model. The output of the APS model provides a spherical position estimate of the transponder s location. These values are used as heading and pitch information for the UUV navigator in the simulation. By combining equations (21) and (23), a final APS output can be represented as, c r (i) (T 1(r (i)) + t 2 1 (i) + T 2(r (i)) + t 2 (i)) [ θ (i) ] = [ θ (i) + θ(i) ] (23) φ (i) φ (i) + θ(i) where, i is the time index in the simulation. The simulator uses a simple closed-loop algorithm to navigate the UUV towards the acoustic source (see Figure 19). In such case, the feedback provided by the APS is proportional to the control set point for the UUV body-fixed heading and pitch adjustments, resulting in a line-of-sight (LOS) trajectory. The goal of the closed-loop system is to force the control error (bearings provided by the APS) to converge towards zero over time. 45

58 θ, φ, r Figure 19: Closed-loop control diagram of the vehicle navigator simulated for the homing process. The simulated controller generates a heading error and a pitch error. These control errors are calculated by backing out the UUV attitude from the simulated APS bearing outputs for each valid fix, such that, [ e Ψ(i) e φ (i) ] = [ θ (i 1) Ψ UUV (i 1) φ (i 1) φ UUV (i 1) ] (24) where, e Ψ is the heading control error, e φ is the pitch control error, Ψ UUV is the UUV heading, and φ UUV is the UUV pitch. The values then become proportional to the vehicle s desired heading and pitch rate so that the simulator can estimate how far the vehicle traveled over a given period of time. If the control error is larger than the maximum heading and pitch rate available by the UUV, then it gets saturated. The saturation points were estimated by analyzing field data of the UUV during basic turning maneuvers and found to be approximately 7 deg sec. In such case, we can show that the UUV s current heading and pitch at time index i can be formulated as, Ψ UUV i ( e Ψ) = { Ψ UUV (i 1) ± 7 if e Ψ (i) Ψ UUVmax (25) Ψ UUV (i 1) + dt e Ψ (i) if e Ψ (i) < Ψ UUVmax 46

59 φ UUV i ( e φ) = { φ UUV (i 1) ± 7 if e φ (i) φ UUVmax (26) φ UUV (i 1) + dt e φ (i) if e φ (i) < φ UUVmax where, dt is a user-defined time resolution of the simulation. To develop the position of each vehicle over time, a displacement (given particular vehicle speed) in the desired vertical and azimuthal bearing is projected onto the horizontal and vertical components of a NED coordinate system. The NED position of the vehicle at time index i, is, UUV N (i) UUV N (i 1) + Vdt cos φ UUV (i) cos Ψ UUV (i) [ UUV E (i)] = [ UUV E (i 1) + Vdt cos φ UUV (i) sin Ψ UUV (i) ] (27) UUV D (i) UUV D (i 1) + Vdt sin φ UUV (i) with initial conditions such that UUV N (0) = 0, UUV E (0) = 0, and UUV D (0) = Z UUV (0). Also, the USV coordinates in the NED frame are, USV N (i) USV N (i 1) + Vdt cos φ USV (i) cos Ψ USV (i) [ USV E (i)] = [ USV E (i 1) + Vdt cos φ USV (i) sin Ψ USV (i)] (28) USV D (i) 0 with initial conditions such that USV N (0) = 0, USV E (0) = 0, and USV D (0) = 0. To produce a more accurate estimate of the homing performance, the cross-track and alongrange errors should be measured using the positions of the DUSBL receiver as mounted on the UUV and the homing transponder as mounted on the USV. These position adjustments are simply an offset from the UUV and USV s center of gravity, respectfully. In such case, the final DUSBL receiver and homing transponder positions can be formulated as, 47

60 USBL N (i) UUV N (i) + dx U cos φ UUV (i) cos Ψ UUV (i) [ USBL E (i)] = [ UUV E (i) + dx U cos φ UUV (i) sin Ψ UUV (i)] (29) USBL D (i) UUV D (i) + dx U sin φ UUV (i) Xponder N (i) USV N (i) + dx T [ Xponder E (i)] = [ USV E (i) + dy T ] (30) Xponder D (i) USV D (i) + dz T where, dx U is the horizontal length from the UUV center of gravity to the center of where the DUSBL receiver; dx T, dy T, dz T are the respective displacement components of where the transponder is mounted relative to the USV s center of gravity. 4.2 Simulation Functionality The simulator requires two input files: (1) simulator configuration file; (2) Bellhop environment file. The simulator configuration file contains typical vehicle and DUSBL system initialization settings including vehicle start positions, vehicle speeds, and DUSBL system specific settings such as system bandwidth and detection threshold information. The Bellhop environment file configures items such as sound velocity profile information, surface and seafloor half-space properties, transducer directivity, and transducer positions needed for simulating sound propagation. A table of the simulated input settings can be found in the Results section in Chapter 6. The start depth of the UUV and range from the USV, as well as transponder start depth and vehicle speeds are initialized at the very beginning of the program. Once the simulation has started, the UUV navigator function requests to acquire the homing transponder 48

61 location. Since no timeouts or distance watchdog features have been implemented in the simulation, if no initial fix is made then the UUV and USV will continue on their independent paths until the initial acoustic fix occurs. The simulation enters a loop in which the DUSBL requests an update every 3 seconds. This update rate of Hz is user defined in the simulation configuration file. For every fix attempt requested, the program uses the DUSBL-APS model to simulate the acoustic handshake between DUSBL transceiver and the homing transponder located on the USV. Referring to Figure 14, an acoustic interrogation message is generated and then propagated using the Bellhop program. The handshake file is read by the receiving end of the homing transponder model. The spectral response is estimated and a signal detection attempt is made if the acoustic arrival is within the cone of sensitivity of the receiver array. If a detection is made, a homing transponder s reply message is generated and propagated back using the Bellhop program. The handshake file is then read by the DUSBL receiver to produce the spectral response, just as the homing transponder model does. If the detected SNR is above the detection threshold, the range and bearing estimator generates a bodyfixed position to the source. The position estimate is then passed into a vehicle navigator that updates the UUV and USV pose (position and orientation) over time. Figure 20 describes the simulation process. As previously explained, the UUV has a narrow cone of sensitivity which the sound ray must be received. The top left portion of Figure 20 shows this cone in more detail. For each successful fix, there is a cone of error associated with the range and bearing estimate. This is seen in the figure as the colored cones. The 49

62 error is due to the system uncertainty and refraction of the acoustic ray, given by a sound velocity profile (bottom right corner of the figure). In order for a fix to be determined valid, the SNR within the desired frequency band must exceed a detection threshold. A fix attempt is generated every 3 seconds, as seen for every fix index. If a fix is not made, the UUV continues on the same track until the next valid fix is acquired (represented by red crossed-out circles). The final cross-track and along-range error is measured at the latest fix. 50

63 Figure 20: Simulation functionality diagram. Image shows the UUV as how it is modeled in the simulation. The UUV acquires a fix every 3 seconds and then adjusts its attitude if a fix is acquired. The UUV stays on the same trackline if no fix is acquired. 51

64 5. FIELD EXPERIMENTATION To validate the simulation results, open-water homing tests were conducted in coastal waters (approximately 1 mile offshore, Figure 21) off the Fort Lauderdale area located in South Florida. The ocean environment of the test site is a flat sandy bottom with maximum depths of 15 meters. Test Location Test Location Figure 21: Satellite image of the test site, Ft. Lauderdale, Florida. Left: Zoomed out view of Ft. Lauderdale relation to Florida. Right: Zoomed in view of test location relative to Ft. Lauderdale and Seatech Research Center. Measuring accurately the shortest distance between the UUV and the transponder is the most challenging aspect of the field experimentation. Error in the results will occur simply due to the watch circle created by the circulating transponder in the swirling ocean current, which worsens the estimation accuracy of the transponder location. Also, deployment of 52

65 the transponder in its precisely planned location is not accurate as it is difficult to stationkeep a deployment vessel in heavy surface current. To minimize the problem at hand, the homing missions were developed in increasingly complex stages. Keeping the problem as simple and controlled as possible allowed to the to gain understanding of the REMUS-100 behavior and maneuverability before attempting more advanced missions. The missions carried out consisted of first homing to a nonmoving or stationary transponder. This stage was broken down into two parts: Phase1: Homing to a stationary transponder fastened to a surface buoy with a drop target assembly moored to the seafloor. Phase 2: Homing to a drop array mounted to a non-moving WAM-V in which was anchored to the seafloor. The final stage of homing the UUV incorporates a moving WAM-V with an attached drop array/tow fish assembly, details of which are fully described in the following sections. The mission plan both in simulation and the field experiment consisted of a 200 meter homing transit with various start depths. During the experimentation a docking tube designed to allow the vehicle to latch on was located just above and below the transponder [16]. 5.1 Stationary Homing Transponder Buoy Homing Test Homing the UUV to a surface buoy with a drop target moored to the seafloor was the first sea-trial. The drop array consisted of a round surface buoy (used for retrieval of the drop 53

66 array) moored to a 3 x3 drop target on the seafloor. The drop target had a checker board pattern easily identifiable in the water from above. The homing transponder was attached at 5 meters depth to the mooring line, along with a plastic docking tube located just above and below the transponder. Located at the top of the docking tube (above the transponder), a GoPro underwater camera was attached looking downwards as to film the incoming UUV. The REMUS 100 was equipped with mechanical latching wings (Figures 22 and 23). The latching system was designed to catch the docking tube as the UUV flew into it [16]. The REMUS-100 was also equipped with a forward-looking GoPro underwater camera (Figure 22, right side) to record the homing approach. The video recorded could also be used to estimate the vertical and horizontal cross-track error in the YZ plane at the nearest point of rendezvous captured. A downward-looking greyscale camera was also equipped and used to capture the drop target located on the seafloor as the UUV flew over it, providing the cross-track error in the XY plane. A diagram of the homing setup can be seen in Figure 24. Figure 22: Hydroid REMUS-100 with docking wing mechanism assembled. Left: UUV in the lab. Middle: UUV in water, Right: Image from GoPro camera of UUV in water. 54

67 Figure 23: Docking wings implemented for sea-trial #1. If the vehicle were to dock, a stop command could be sent to the UUV via the acoustic Ranger to end the mission. The overall mission consisted of five waypoints and the homing process occurred on the last leg. Figure 25 shows the concept of operation for this mission. The UUV was dropped at a set point located just south of waypoint 1. Once given the start command, the UUV transited to waypoint 1, located at 10 meters depth. The vehicle was then given additional time to level and stabilize its pitch this occured between waypoints 2 and 3. On the third leg, the UUV began its 200 meter transit using the DUSBL-APS to locate the homing transponder due north (waypoint 4) at 5 meters depth. 55

68 Figure 24: Test observation setup for sea-trial #1. Figure 25: Programmed UUV track using the VIP interface. 56

69 5.1.2 Fixed USV Homing Test The second stage of sea-trials consisted in homing to a non-moving USV. The WAM-V replaced the surface buoy and the checkerboard drop target was removed so that the surface vehicle could be moored using an anchor. A drop array was attached to the bottom-side of the WAM-V deck to hold the camera, docking tube, and homing transponder. Its total length was 7 meters. The configuration of the GoPro camera, docking tube, and homing transponder stayed the same, however, a 10lbs. mushroom anchor was attached at the end to keep the array vertical in the water column. Figure 26 shows an image of the capturing mechanism, which consisted of a simple stinger mounted on to the UUV. The stinger was used to latch the UUV to the docking station. Figure 26: REMUS-100 with stinger capturing mechanism assembled for sea-trial #2. This test was of interest because the heave and sway motion caused by the wave-action was directly coupled to the drop array, and in turn changed the motion of the underwater 57

70 transponder. Also, to keep the anchor in the seafloor, the anchor line needed to be longer than the mooring in the previous test, which resulted in a much large watch circle. An image of this test setup can be seen in Figure 27. The overall mission plan is the same as explained in the previous section. The overall goal of this test was to verify whether the UUV could home to a nearly static transponder off the programmed location. Figure 27: Observation and docking setup for sea-trial #2. 58

71 5.2 Mathematical Derivations The data utilized for analyzing the experimental homing performance are provided by the State and ADCP files recorded by the REMUS-100. The data are used to generate the horizontal cross-track error and vertical cross-track error of the UUV at the rendezvous point. This information is then used to estimate the probability density function of crosstrack error about the homing transponder. The experimental vertical cross-track error e vct is obtained by using the depth data provided by the vehicle s depth sensor and known depth of the homing transponder. To do so, the homing transponder depth is subtracted from the UUV depth at the point of rendezvous, such that, e vct(k r ) = Z U (k r ) Z T (31) Where, Z U is depth of the REMUS 100, Z T is depth of the homing transponder according to the mission plan, and k r is the k th index of the data set where rendezvous occurs. The experimental horizontal cross-track error e hct can be calculated by referring to the positioning data located in the State file. The data, along with position of the homing transponder, are converted from a geodetic (latitude, longitude, altitude LLA) reference frame to an Earth Center Earth Fixed (north, east, down NED) reference frame that is a convenient value measured in meters. The position of the UUV is subtracted from the position of the homing transponder along the y-axis at the rendezvous point to determine the horizontal cross-track error. The conversion from LLA to NED is performed by 59

72 converting to an earth-centered-earth-fixed (x,y,z ECEF) coordinate frame using the following transformation, X g N cos λ cos φ + h cos λ cos φ [ Y g ] = [ N cos λ sin φ + h cos λ sin φ ]. (32) Z g N( ) sin λ + h sin λ Where, h is the altitude (set to zero assuming close enough to sea-level), λ is the geodetic latitude, and φ is the geodetic longitude. Earth s geometric constant N can be calculated as, N = sin 2 λ. (33) The experimental horizontal cross-track error e hct can be formulated as, e hct(k r ) = Y gu (k r ) Y g T. (34) Where Y gu is the horizontal ECEF position of the UUV, Y g T is the horizontal ECEF position of the homing transponder, and k r is the index in the data set where rendezvous occurs. Further insight of the homing performance can be observed by visually comparing three track lines: (1) Where the UUV was programmed to go; (2) Where the vehicle actually went. It is best to overlay these graphs relative to a coordinate reference, which provides the position in meters relative to the set position of the UUV data. The vehicle s mission track and measured track (using the vehicle s sensors) is obtained from the State file. The variables are labeled as goal_latitude/goal_longitude and 60

73 latitude/longitude, respectfully. Another estimated track is computed by fusing the UUV s surge and sway velocities with the compass heading. This operation produces a rate-based NED position of the vehicle during the mission. The latitude/longitude data sets are not used because the USV is almost never located exactly where it is programmed due to lack of GPS accuracy and sea surface dynamics. As a summary, the mathematical derivations are provided below. In order to display these trajectories in the NED coordinate frame, the start position in ECEF frame is backed out of the data and the unbiased set is multiplied by a coordinate transformation matrix C b g, such that, sin λ cos φ sin λ sin φ cos λ C g b = [ sin φ cos φ 0 ] (35) cos λ cos φ cos λ sin φ sin λ and X b [ Y b ] = C b Z b g [ X g X g (0) Y g Y g (0) ]. (36) Z g Z g (0) Where, λ is the geodetic latitude data vector, φ is the geodetic longitude data vector, C g b is the coordinate transformation matrix, X b is the body-fixed north position, Y b is the bodyfixed east position, and Z b is the downward position data vector relative to (0,0,0) located at the vehicle start position. To verify the sensor data, a DR trajectory was generated to estimate the vehicle track. The State file provides compass heading Ψ data and the ADCP file provides the surge velocity 61

74 u and the sway velocity v data. The data logging rate of the State and ADCP files are logged by separate interrupts in the system and the DR estimation routine used in postprocessing requires that the input vectors be of the same length. To satisfy this requirement, the data sets are interpolated using Matlab s interp1( ) function. Following this, the DR position estimation can be formulated by first calculating the spatial velocity components of the UUV x DR and y DR, such that, [ x DR u cos Ψ v sin Ψ ] = [ y DR u sin Ψ v cos Ψ ] (37) The body-fixed NED spatial components x and y are calculated by multiplying the velocity components by a interpolated time step, dt DR. The spatial position components are calculated as, 0 for k 0 x DR (k) = { K k=1 x(k 1) + x (k))[t(k) t(k 1)] for k > 0, (38) 0 for k 0 y DR (k) = { K k=1 y(k 1) + y (k))[t(k) t(k 1)] for k > 0, (39) Where K is the maximum number of data points in the interpolated time vector t, k is the interpolated time index, x DR is the North position, and y DR is the East position. In the experimental results section, the values X b U_goal, Y bu_goal, X bu, Y, x, y, and the bu UUV depth Z U (the true Down component) are plotted for every mission. 62

75 6. RESEARCH RESULTS 6.1 Fixed Homing Simulation For this simulated test, the USV was programmed to be stationary (USV speed set to zero) and only the UUV had the ability to move. Several sets of simulations were performed using different input parameters such as, UUV start range, UUV start depth, and homing speed to observe how the homing performance responded to the changes. The simulations were conducted in batch processes of 100 runs to build a statistical profile of the results for a given set of input parameters. This section reviews the major observations that were found in the results. The simulations were primarily analyzed for two different start depths and over several ranges. The number of start depths were limited to narrow down the endless number of choices; in particular, they were chosen based on the most probable start depth of the UUV when following the USV in a docking attempt. Typically, the UUV would navigate in midwater to receive as many successful acoustic homing signals as possible. Here, the simulations focused on homing the UUV with a start depth at 5 meters (about mid water column for the proposed area) or a start depth at 10 meters (about 3 meters off the seafloor). The simulation would be most useful when estimating the homing performance in a noiselimited condition, in order to find the maximum homing range in presence of high ambient noise. For such case, the ambient noise was held constant throughout the simulations at 40 db re 1 μpa/ Hz. The optimal detection threshold was found to be about 50 db; anything 63

76 much higher may start to produce rejected signals within close ranges (outside 500 meters). Table 3 and 4 shows a list of the input parameters and their settings that were used over a course of different simulations. Table 3 Varying Simulation Input Parameters Parameter Sim 1 Sim 2 Sim 3 Sim 4 Sim 5 Sim 6 UUV Start Depth 5.0 m 10.0 m 5.0 m 5.0 m 10.0 m 10.0 m UUV Start Range 75 m 75 m 100 m 200 m 200 m 100 m Table 4 Constant Simulation Input Parameters Parameter Transponder Depth USV Speed UUV Speed UUV Max Pitch Rate UUV Max Heading Rate USBL Source Level USBL Carrier Frequency USBL Bandwidth USBL Pulse Length USBL Detection Threshold USBL Bearing Accuracy Value 5.0 m 0.0 m/s 1.50 m/s 7 deg/s 7 deg/s 186 db re 1 1m Hz 6000 Hz 4 msec 50 db ± 2 deg 64

77 USBL Range Accuracy USBL Sample Rate USBL Update Rate Transponder Source Level Transponder Carrier Frequency Transponder Bandwidth Transponder Pulse Length Transponder Detection Threshold Ambient Noise PSD ± 1 ms Hz 0.33 Hz 186 db re 1 1m Hz 6000 Hz 4 msec 50 db 40 db re 1 μpa/ Hz Focus was placed on simulation #4 and #5 since these represent the same scenario tested in the field. Starting with simulation #4, according to the output report, (found in the Appendix) the UUV should make about 55 fix attempts, assuming the vehicle s speed was true (no current disturbance). The maximum vertical cross-track error observed is 9 cm above the transponder and 9 cm below the transponder. The standard deviation over 100 runs was 4 cm. Larger errors were observed in the horizontal plane. The maximum horizontal cross-track error expected is 1.41 meters port of the transponder and 1.33 meters starboard of the transponder. The horizontal cross-track error standard deviation was 0.55 meters. Figure 28 and Figure 29 show the probability density distribution of the horizontal and vertical cross track error for this type of run. The vertical cross-track distribution appears somewhat Gaussian, as the horizontal cross-track error seems to be more uniformly distributed across the error range. Figure 30 shows a polar plot of cross-track error in the 65

78 YZ plane. The red circle represents the expected error circle about the transponder with radius equal to the maximum cross-track standard deviation. Horizontal CT Error PDF estimate - Sample mean = m - Sample variance = m PDF (1/m) Horizontal CT Error (m) Horizontal Figure 28: CT Error Horizontal CDF estimate cross-track - GOF error test histogram passedat5% for simulation significance #4. level Vertical CT Error PDF estimate - Sample mean = m - Sample variance = m 2 15 PDF (1/m) Vertical CT Error (m) Vertical Figure CT 29: Error Vertical CDF estimate cross-track - GOF error test histogram passedat5% for simulation significance #4. level 66

79 UUV Homing Cross-track Error YZ Plane Polar Plot - Start Range = 200m - Start Depth =5m Figure 30: Vertical plane cone error expected for simulation #4. Looking at simulation #5, the results are fairly different. In this scenario, the UUV starts its process from 10 meters depth and has to home to 5 meters depth. The number of attempted fixes stayed the same 55 fix attempts. The maximum vertical cross-track error expected is meters above the transponder and meters below the transponder, with a standard deviation of meters. The average vertical cross-track error is meters. The maximum horizontal cross-track error expected is meters port of the transponder and meters starboard of the transponder, with a standard deviation of meters. The vertical and horizontal cross-track error histograms can be found in Figure 31 and Figure 32. A polar plot of cross-track error in the YZ plane is shown in Figure

80 Vertical CT Error PDF estimate - Sample mean = m - Sample variance = m 2 15 PDF (1/m) Vertical CT Error (m) Vertical CT Error CDF estimate - GOF test passedat5% significance level 100 Figure 31: Vertical cross-track error histogram for simulation #5. %) Horizontal CT Error PDF estimate - Sample mean = m - Sample variance =0.2713m PDF (1/m) Horizontal CT Error (m) Horizontal CT Error CDF estimate - GOF test passedat5% significance level 100 Figure 32: Horizontal cross-track error histogram for simulation #5. ) UUV Homing Cross-track Error YZ Plane Polar Plot - Start Range = 200m - Start Depth =10m Figure 33: Cross-track error YZ plane polar plot for simulation #5. 68

81 The results show that the UUV is certainly capable of limiting rendezvous error to within 1 meter of a homing source. The models, however, are quite conservative and produce better results than expected to obtain in the field. This is due to unknown parameters such current flow disturbance on the UUV body, random drift of the transponder, and true control features of the REMUS-100 software. Further results are displayed in Table 5 where all of the simulations can be compared. Table 5 Fixed Homing Simulation Results Sim 1 Sim 2 Sim 3 Sim 4 Sim 5 Sim 6 Avg. # of Valid Fixes Max vct m m m m m m Above Max vct m m m m m m Below Avg. vct m m m m m m vct Std Deviation vct Variance vct Mean-Square Error Upper CI of Std Dev vct w/ 95% CL Lower CI of Std Dev vct w/ 95% CL Max hct Above Max hct Below m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m Avg. hct m m m m m m 69

82 hct Std Deviation hct Variance hct Mean-Square Error Upper CI of hct Std Dev w/ 95% CL Lower CI of hct Std Dev w/ 95% CL m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m 6.2 Mobile Homing Simulation Mobile homing simulations were conducted to characterize the UUV s ability to dock with a moving target. These simulations were conducted in the same manner as explained in chapter 6.1 for the fixed homing simulations. All parameters were kept the same except the USV speed was set to 0.80 m/s (~ 1.6 knots) and travels in a perfectly straight path (no transit noise such as wave/current interaction involved). Looking at simulation #4, the mobile results are slightly larger than the static results, as one would expect considering the UUV is homing towards a moving target. In this scenario, the UUV starts its process from 5 meters depth and has to home to a 5 meter deep target moving at 0.80 m/s. The number of attempted fixes stayed the same 148 fix attempts. The maximum vertical cross-track error expected is meters above the transponder and meters below the transponder, with a standard deviation of meters. The average vertical cross-track error is meters below the transponder. The 70

83 maximum horizontal cross-track error expected is meters port of the transponder and meters starboard of the transponder, with a standard deviation of meters. The vertical and horizontal cross-track error histograms can be found in Figure 34 and Figure 35. A polar plot of cross-track error in the YZ plane is shown in Figure 36. Horizontal CT Error PDF estimate - Sample mean = m - Sample variance = m PDF (1/m) Horizontal CT Error (m) Horizontal CT Error CDF estimate - GOF test passedat5% significance level Figure 34: Horizontal cross-track error histogram for mobile homing simulation #4. Vertical CT Error PDF estimate - Sample mean = m - Sample variance = m 2 15 PDF (1/m) Vertical CT Error (m) Vertical CT Error CDF estimate - GOF test passedat5% significance level Figure 35: Vertical cross-track error histogram for mobile homing simulation #4. 71

84 UUV Homing Cross-track Error YZ Plane Polar Plot - Start Range = 200m - Start Depth =5m Figure 36: Cross-track error YZ plane polar plot for mobile homing simulation #4. Looking at simulation #5, an average of 82 fixes were made in the mobile homing simulation, as compared to 55 successful fixes in the fixed homing simulation. The maximum vertical cross-track error expected is meters above the transponder and meters below the transponder, with a standard deviation of meters. The average vertical cross-track error is meters above the transponder. The maximum horizontal cross-track error expected is meters port of the transponder and meters starboard of the transponder, with a standard deviation of meters. The vertical and horizontal cross-track error histograms can be found in Figure 37 and Figure 38. A polar plot of cross-track error in the YZ plane is shown in Figure

85 Horizontal CT Error PDF estimate - Sample mean = m - Sample variance = m PDF (1/m) Horizontal CT Error (m) Horizontal CT Error CDF estimate - GOF test passedat5% significance level 150 Figure 37: Horizontal cross-track error histogram for mobile homing simulation #5. Vertical CT Error PDF estimate - Sample mean = m - Sample variance = m 2 15 PDF (1/m) Vertical CT Error (m) Vertical CT Error CDF estimate - GOF test passedat5% significance level 100 Figure 38: Vertical cross-track error histogram for mobile homing simulation #5. 73

86 UUV Homing Cross-track Error YZ Plane Polar Plot - Start Range = 200m - Start Depth =10m Figure 39: Cross-track error YZ plane polar plot for mobile homing simulation # Experimental Results For each run careful measurements were taken of the UUV and transponder position, number of DUSBL fixes attempted, number of DUSBL fixes acquired, signal-to-noise ratio for each fix, sound speed data, bathymetry along the vehicle track, and video of the underwater target. The experimental sea-trials were conducted in two separate phases as explained in chapter 5. The first phase consisted of homing the UUV to a static buoy with a sub-surface docking mechanism attached to a mooring line. The second phase consists of homing the UUV to a nearly stationary USV equipped with a sub-surface docking array Fixed Buoy Homing Results The buoy homing test was performed in calm weather: approximately sea-state 1 wave heights. There was a total of eight runs for this sea-trial, in which two of the runs the vehicle 74

87 was able to attempt to home towards the target. The first four runs had to be aborted because the UUV required a locate nearest transponder at the beginning of the mission. Since the vehicle was located on the surface it could not receive a successful reply from the homing transponder from such far distance where the UUV deployment point was located (over 200 meters), causing a mission timeout. This is due to the fading characteristics of the acoustic channel. Once this feature was disabled, the UUV started the mission immediately; however, difficulty was encountered in descending below the surface since the seas were so calm. This forced the vehicle to abort one of its runs because the UUV was stuck on the surface. Table 6 shows a list of the runs and their overall outcomes. When the UUV was able to home towards the target, only one of the attempts proved to be effective as the UUV was able to successfully dock to the mooring station. The UUV made 47 DUSBL fix attempts, with 10 of them being successful. The overall cross-track error was meters above the transponder. Figure 40 shows an image of the drop target located at the seafloor recorded from a grey-scale camera located on the bottom side of the REMUS-100. The original image s brightness setting was too high and needed to be postprocessed with a low-pass filter in Matlab to reduce some of the white light saturation. The left side of the image shows the drop target in the upward portion of the camera s field of view as the UUV approaches from the south. The image on the right is the very last image recorded when the vehicle docked, just before the mission success flag was set and the VCR was shut off. Figure 41 shows UUV track during run #5 from the horizontal plane. The red line represents the commanded track and the black line represents the actual vehicle track. Figure 42 shows the UUV track during run #5 from the vertical plane. The blue track 75

88 in the actual vehicle position as a function of range. The red circles are the positions in which a fix was made. Table 6 Sea-Trial #1 Homing Results Run Result 5 DR Nav: Success 6 DUSBL Home: Attempted, no success docking # of Fix Attempt # of Fix Success Fix Success Rate hct Error vct Error N/A N/A N/A 9.0 meters meters % > 10.0 meters 0.62 meters Docking Line Drop target Drop Target Approaching Figure 40: Image recorded by REMUS-100 grey-scale camera located on the bottom side of the UUV (Run 5). Left: Shows drop target as UUV approaches transponder. Right: Shows drop target below UUV once docking has occurred. 76

89 Figure 41: UUV track during the successful homing mission conducted in sea-trial #1 (Run 5). Figure 42: UUV track in the vertical plane. The UUV was programmed to maintain altitude for this mission. 77

90 6.3.2 Fixed USV Homing Results USV Homing Sea-trial #1 The static USV homing test was performed over a period of two separate sea-trials. The first sea-trial was performed in with sea-state 2-3. Winds were varying; however, gusts were occasionally expected up to knots from the east-south-east. The increase in wind strength produced a stronger surface current; approximately 2 knots flowing in the same direction as the wind. The WAM-V was deployed approximately feet southwest of the programmed transponder location in hope to allow the current to drift the USV in place as the anchor was set. The anchor line length was rigged for 5 times the water depth, which equates to approximately 75 meters of line for 15 meters of water column. The actual water depth at deployment was 14 meters. There were six runs in total conducted for this sea-trial, as shown in Table 7. The first run was a simple waypoint navigation mission on the surface to ensure the UUV was working properly, which was successful. The subsequent runs used the DUSBL as a navigation source to home to the transponder. Only in run #2 and run #5 was the UUV able to acquire valid DUSBL fixes. In run #2, the UUV made 54 fix attempts, in which only five of them were successful, yielding a 9.25% fix success rate. In run #5, the UUV made 57 fix attempts; 6 were successful, yielding a 6.89% fix success rate. The vehicle was able to fly within 10 meters of the target, as seen by the GoPro image in Figure

91 Figure 43: Image recorded by GoPro camera during a docking attempt fly-by in Seatrial #2. Table 7 Sea-trial #2 Homing Results Run Result 2 DUSBL Home: Attempted, no success docking 5 DUSBL Home: Attempted, no success docking # of Fix Attempt # of Fix Success Fix Throughput hct Error % >10.0 meters % ~7.0 meters vct Error 0.08 meters meters USV Homing Sea-trial #2 The next sea-trial was conducted in much calmer conditions with sea-state 1 wave heights and wind speed less than 8 knots. The current was strong, approximately 2 knots from the 79

92 south-south-east. In order to overcome many of the issues experienced in sea-trial #1, several test plan parameters were changed for sea-trial #2. The most significant change was that the UUV was programmed to now home along a north-west to south-east track such that the vehicle flew into the current, as compared to the south to north track run in the previous mission where the vehicle flew down current. The motivation was to reduce the vehicle speed over ground, thus increasing the number fixes acquired while still being able to produce enough water speed to maintain stability over the vehicle s control surfaces. As compared with the first sea-trial, the configuration was different in the following aspects: 1. UUV is run on a north-west to south-east track. 2. USV anchor line scope is reduced to double the water depth; new length is 90 meters. 3. UUV speed reduced to 3 knots for first 125 meters, then 1.5 knots for last 75 meters 4. USV is purposefully placed in various locations to verify UUV behavior. The second sea-trial was conducted to gain further insight on the UUV s behavior when the homing transponder, although mostly stationary, was not located where it was programmed to be. To first acquire a reference performance, the USV was deployed as accurately to the programmed location as possible. The USV s GPS location was then wirelessly sent to a laptop on the deployment vessel so it could be updated in the mission plan. As seen in Figure 44, the UUV was programmed to home from a starting point at 200 meters from the USV, with a flying speed at 3 knots and then slow down to 1.5 knots when 80

93 within 75 meters in range. The same test was then repeated, however, the UUV was programmed to skip the 200 meter engagement mark and instead begin the homing process from 75 meters range, flying at 1.5 knots. Once completed, the same tests were then repeated such that the commanded UUV track deviated 10º port off the actual transponder location (see Figure 45). This forced the UUV to perform a deliberate maneuver off course from the commanded route if it were to follow the actual sound source and not the preprogrammed location of where the sound source might be. A total of four runs were conducted before the test had to be postponed due to a mechanical malfunction of the docking station on the USV. For the first two runs, the USV was located exactly where it was programmed to be, and the last two runs the commanded track of 10º port. The changes that were made were quite effective, as shown in the following results. Table 6, located at the end of this section, provides a summary of the sea trial results. Mission Plan for Run #1 and Run #2: North Set Position Bearing = SE 150º General Direction of Surface Current Actual USV Position Programmed Homing DR Location Homing Start 1 (WP2) Range = 200 meters Speed = 3.0 knots Homing Start 2 (WP3) Range = 75 meters Speed = 1.5 knots Figure 44: Programmed UUV track for seatrial #3, runs 1 and 2. For these run, the UUV starts homing at WP2 (for run 1) and WP3 (for run 2). The actual transponder location coincides with the programmed transponder location. 81

94 Mission Plan for Run #3 and Run #4: Set Position Bearing = SE 140º General Direction of Surface Current Homing 1 (WP2) Range = 200 meters Speed = 3.0 knots Homing 2 (WP3) Range = 75 meters Speed = 1.5 knots North Actual USV Position 10 Programmed Homing Location Figure 45: Programmed UUV track for seatrial #3, runs 3 and 4. For these run, the UUV starts homing at WP2 (for run 1) and WP3 (for run 2). The programmed transponder location is deviated 10 port of the actual transponder. Run 1: In run #1, the UUV made 100 DUSBL fix attempts, of which 48 of them were successful, yielding a 48% fix success rate. This is significantly higher than the fix success rate in the previous sea trials. Figure 46 shows the UUV s position track during this run. Figure 47 shows a zoomed-in view of the UUV track when within several meters of the target. These figures, show that the UUV starts off on the correct track during the first two waypoints which are dead reckoned (DR) waypoints then the UUV flies near the commanded courses once it engages the homing process at the 200 meter mark. The UUV attempted to 82

95 home towards the transponder even though the transponder s actual location was off the programmed location. Unfortunately, the UUV was just short of docking because the flight speed was set too low during the last 75 meter homing transit; this caused the vehicle to stall and hit the sea bottom. Figure ### shows that the UUV successfully slows down from 3 knots (~ 1.5 m/s) to 1.5 knots (~ 0.75 m/s) when within 75 meters of the target (See Figure 48). Unfortunately, the heavy head-current forced the UUV to slow down so much that it stalled and was no longer able to maintain depth. Figure 49, clearly shows this issue: the UUV ran aground although the stern planes were in full ascent position (as verified in the VIP playback software). The overall horizontal cross-track error was less than 3 meters port, and the along-range error was approximately 6 meters short of the target REMUS 100 Acoustic DUSBL Navigation - Homing Range: 200m, Run #1 WP2 Homing Start Position Range = 200 meters Velocity = 1.25 m/s Set Position North (meters) WP3 UUV Slows Down Range = 75 meters Velocity = 0.75 m/s Transponder Goal Position Waypoint Goal Position Ext. Kalman Filter Position DR Actual Position USV Transponder Position East (meters) Figure 46: Sea-trial #3, run 1 UUV track results. 83

96 REMUS 100 Acoustic DUSBL Navigation - Homing Range: 200m, Run # Horizontal Cross-track Error hct = 6 meters North (meters) m Radius Error Circle Actual Transponder Position Transponder Goal Position Waypoint Goal Position Ext. Kalman Filter Position DR Actual Position East (meters) Figure 47: Sea-trial #3, run 1 zoomed-in UUV track results. Figure 48: Sea-trial #3, run 1 UUV estimated forward velocity during the mission. 84

97 REMUS 100 Acoustic Homing Depth Profile - Homing Range: 200m, Run #1 0 UUV Path -2 Transponder Location -4 Depth (meters) Transit Distance (meters) Figure 49: REMUS-100 vertical track for seatrial #3, run 1. The image shows how the UUV loses vertical stability just after within 75 meters in range. The vehicle target speed at that point is 1.5 knots. Run #2: In run #2, the UUV was programmed to engage its homing when within 75 meters of the target. The results were not quite as effective as in run #1. The vehicle made 42 DUSBL fix attempts, in which 21 of them were valid, yielding a 50% fix success rate. Although the fix success rate was reasonable, the UUV ended its mission off the target. Due to the slow speed during homing engagement, the vehicle lost vertical stability. This time, however, the mission was aborted because the UUV could no longer receive an acoustic reply from the homing transponder. Figure 50 shows the UUV track. The along-error is well above 5 meters, and the UUV mission aborted before reaching the target. 85

98 0 REMUS 100 Acoustic DUSBL Navigation - Homing Range: 200m, Run #2-50 Set Position WP1 Start Transit Velocity = 1.25 m/s North (meters) Transponder Goal Position Waypoint Goal Position Ext. Kalman Filter Position DR Actual Position WP2 Homing Start Position Velocity = 1.25 m/s WP3 UUV Slows Down Velocity = 0.50 m/s Mission Timeout, Abort East (meters) Figure 50: Sea-trial #3, run 2 UUV track results. Run #3 In run #3, the commanded UUV track was deviated 10º portside of the programmed transponder location. The results proved successful as the UUV was able to home towards the actual acoustic source, even though the transponder was not located where it was programmed to be. Figure 51 shows the UUV track for this run. It can be seen that the vehicle clearly starts out on the commanded track as it transits to waypoint 2, where the 200 meter homing process engages. Once the vehicle passes waypoint 2, it goes off the programmed course and adjusts its bearing towards the actual acoustic source. As in the previous runs, the vehicle speed was set too low and it lost vertical stability during the last 75 meter leg. Although still stuck on the seafloor (see Figure 54), Figure 53 indicates that 86

99 the vehicle s horizontal cross-track error was less than 4 meters. Figure 52 shows an image of the UUV as it flies just below the homing transponder. 0 REMUS 100 Acoustic DUSBL Navigation - Homing Range: 200m, Run #3-50 Set Position Homing Start Position Range = 200 meters Velocity = 1.25 m/s North (meters) Transponder Goal Position Waypoint Goal Position Ext. Kalman Filter Position DR Position REMUS State Position GPS Fix Position UUV Slows Down Range = 75 meters Velocity = 0.75 m/s USV Transponder Position East (meters) Figure 51: Sea-trial #3, run 3 UUV track results. REMUS-100 approaching below Figure 52: Sea-trial #3, run 3. Image taken by downward looking GoPro camera of UUV flying under the homing transponder. 87