Calhoun: The NPS Institutional Archive □Space Repository

Theses and Dissertations

1. Thesis and Dissertation Collection, all items

2016-06

Terminal homing for autonomous underwater vehicle docking

Bermudez, Eric B.

Monterey, California: Naval Postgraduate School

http://hdl.handle.net/10945/49385

This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States.

Downloaded from NPS Archive: Calhoun

DUDLEY

KNOX

LIBRARY

CaflxKio is the Naval Postgratluate School's pubfe access {#^1 repository for research (Tiateriate and institutional publications created by the NPS community. Calhoun is named for Professor of Mathematics Guy K. CaWwun, NPS's first appointed and pubftshed scholaily author.

Dudley Knox Library / Naval Postgraduate Schcxil 411 Dyer Road / 1 University Circle Monterey, California USA 93943

htt p://vy ww. n ps. e du/I ib ra ry

NAVAL

POSTGRADUATE

SCHOOL

MONTEREY, CALIFORNIA

THESIS

TERMINAL HOMING FOR AUTONOMOUS UNDERWATER VEHICLE DOCKING

by

Eric B. Bermudez June 2016

Thesis Advisor: Douglas Homer

Second Reader: Noel Du Toil

Approved for public release; distribution is unlimited

THIS PAGE INTENTIONALLY LEFT BLANK

REPORT DOCUMENTATION PAGE

Form Approved 0MB No. 0704-0188

Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instruction, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188) Washington, DC 20503.

1. AGENCY USE ONLY 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED

(Leave blank) _ June 2016 _ Master’s thesis _

4. TITLE AND SUBTITLE

TERMINAL HOMING FOR AUTONOMOUS UNDERWATER VEHICLE DOCKING

6. AUTHOR(S) Eric B. Bermudez

11. SUPPLEMENTARY NOTES The views expressed in this thesis are those of the author and do not reflect the official policy or position of the Department of Defense or the U.S. Government. IRB Protocol number _ ^N/A _

13. ABSTRACT

The use of docking stations for autonomous underwater vehicles (AUV) provides the ability to keep a vehicle on station, conducting missions for extended periods of time, with limited human interaction. However, the use of a docking station brings about challenges associated with terminal homing, position estimation, and vehicle control. A traditional single propeller-driven AUV must dock at a high relative approach velocity to maintain controllability, which can lead to serious damage to the AUV and the docking station. Alternatively, equipping a AUV with forward and aft pairs of horizontal and vertical cross-tunnel thrusters enables a hovering capability and allows for a slower, more deliberate approach that can help reduce potential damage during the terminal homing phase. Additionally, the commonly used ultra-short baseline (USBL) acoustic transponder attached to the docking station, which provides bearing and range measurements, can be asynchronous and sparse. The integration of these measurements into an optimal position estimation filter can potentially produce inaccuracies that are detrimental during docking operations. This thesis discusses the development of a hydrodynamic model and a filtering algorithm for position estimation for a cross tunnel thruster-enabled REMUS 100 AUV. The hydrodynamic model provides the capability of simulating vehicle docking with variable environmental effects. The filtering algorithm looks to provide an integrated solution of inertial navigation measurements and UBSL measurements to provide a more accurate vehicle location during docking operations.

16. PRICE CODE

NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89)

Prescribed by ANSI Std. 239-18

i

20. LIMITATION OF ABSTRACT

uu

15. NUMBER OF PAGES

121

14. SUBJECT TERMS

terminal homing, REMUS 100, USBL, UKF, hydrodynamic model, position estimation filtering

18. SECURITY CLASSIFICATION OF THIS PAGE

Unclassified

19. SECURITY CLASSIFICATION OF ABSTRACT Unclassified

17. SECURITY CLASSIFICATION OF REPORT

Unclassified

12b. DISTRIBUTION CODE

12a. DISTRIBUTION / AVAILABILITY STATEMENT Approved for public release; distribution is unlimited

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) Naval Postgraduate School Monterey, CA 93943-5000

9. SPONSORING /MONITORING AGENCY NAME(S) AND ADDRESS(ES)

N/A

5. FUNDING NUMBERS

8. PERFORMING ORGANIZATION REPORT NUMBER

10. SPONSORING / MONITORING AGENCY REPORT NUMBER

THIS PAGE INTENTIONALLY LEFT BLANK

11

Approved for public release; distribution is unlimited

TERMINAL HOMING FOR AUTONOMOUS UNDERWATER VEHICLE

DOCKING

Eric B. Bermudez Ensign, United States Navy B.S., United States Naval Academy, 2015

Submitted in partial fulfillment of the requirements for the degree of

MASTER OF SCIENCE IN MECHANICAL ENGINEERING

from the

NAVAL POSTGRADUATE SCHOOL June 2016

Approved by: Dr. Douglas Homer

Thesis Advisor

Dr. Noel Du Toit Second Reader

Dr. Garth Hobson

Chair, Department of Mechanical and Aerospace Engineering

THIS PAGE INTENTIONALLY LEFT BLANK

IV

ABSTRACT

The use of docking stations for autonomous underwater vehicles (AUV) provides the ability to keep a vehicle on station, conducting missions for extended periods of time, with limited human interaction. However, the use of a docking station brings about challenges associated with terminal homing, position estimation, and vehicle control. A traditional single propeller-driven AUV must dock at a high relative approach velocity to maintain controllability, which can lead to serious damage to the AUV and the docking station. Alternatively, equipping a AUV with forward and aft pairs of horizontal and vertical cross-tunnel thrusters enables a hovering capability and allows for a slower, more deliberate approach that can help reduce potential damage during the terminal homing phase. Additionally, the commonly used ultra-short baseline (USBL) acoustic transponder attached to the docking station, which provides bearing and range measurements, can be asynchronous and sparse. The integration of these measurements into an optimal position estimation filter can potentially produce inaccuracies that are detrimental during docking operations. This thesis discusses the development of a hydrodynamic model and a filtering algorithm for position estimation for a cross tunnel thruster-enabled REMUS 100 AUV. The hydrodynamic model provides the capability of simulating vehicle docking with variable environmental effects. The filtering algorithm looks to provide an integrated solution of inertial navigation measurements and UBSL measurements to provide a more accurate vehicle location during docking operations.

V

THIS PAGE INTENTIONALLY LEFT BLANK

VI

TABLE OF CONTENTS

I. INTRODUCTION . I

A. MOTIVATION FOR THIS WORK . I

B. PROBLEM DESCRIPTION . 4

C. LITERATURE REVIEW . 7

1. Hydrodynamic Models . 7

2. Docking Station . 8

3. Position Estimation Filtering . 10

D. THESIS ORGANIZATION . 12

II. SYSTEM DESCRIPTION . 13

A. REMUS 100 . 13

B. DOCKING STATION . 18

HI. REMUS 100 HYDRODYNAMIC MODELING WITH FORE AND

AFT TUNNEL THRUSTERS . 21

A. INTRODUCTION . 21

B. MODEL BACKGROUND . 21

1. Assumptions . 21

2. Reference F rame . 21

C. EQUATIONS OF MOTION . 23

D. FORCE AND MOMENT COMPONENTS . 26

1. Hydrostatic Component . 26

2. Added Mass Component . 27

3. Hydrodynamic Damping Component . 27

4. Lift Component . 28

5. Propulsive Component . 29

E. COMBINED HYDRODYNAMIC EQUATIONS OF MOTION . 37

F. DEVELOPMENT OF MODEL COEFFICIENTS . 40

1. Previous Coefficient Derivations . 40

2. Current Coefficient Development . 40

G. MODEL CONTROLLERS AND AUTOPILOTS . 48

IV. POSITION ESTIMATION FILTER . 51

A. KALMAN FILTER . 51

B. EXTENDED KALMAN FILTER . 52

C. UNSCENTED KALMAN FILTER . 54

D. FILTER SYSTEM CONFIGURATION . 56

vii

E. TESTING DATA AND METHOD . 60

V. RESULTS AND ANALYSIS . 65

A. HYDRODYNAMIC MODEL RESULTS . 65

B. POSITION ESTIMATION FILTER RESULTS . 78

VI. CONCLUSION . 89

A. HYDRODYNAMIC MODEL . 89

B. POSITION ESTIMATION FILTER . 90

C. FUTURE WORK . 91

1. Hydrodynamic Model . 91

2. Position Estimation Filter . 91

APPENDIX A. VEHICLE PARAMETERS . 93

APPENDIX B. HYDRODYNAMIC COEFFICIENTS . 95

LIST OF REFERENCES . 97

INITIAL DISTRIBUTION LIST . 101

viii

LIST OF FIGURES

Figure L Position Solutions Using Two Transponders. Source: [5] . 3

Figure 2. Standard REMUS Configuration. Source: [5] . 5

Figure 3. Damage Incurred from Unsuccessful Vehicle Docking . 5

Figure 4. Updated, Non-Standard REMUS Configuration . 6

Figure 5. MBARI Docking Station. Source: [12] . 9

Figure 6. Zhejiang University Docking Station. Source: [13] . 9

Figure 7. Vehicle -231 with D-USBL Nose Attachment . 15

Figure 8. USBL Specifications for Determining Good Fixes. Source: [5] . 16

Figure 9. CAVR Docking Station for REMUS 1 00 AUV . 19

Figure 10. Reference Frames and Degrees of Freedom of AUV. Source: [8] . 23

Figure 11. FUTEK Strain Gauge Used for Thrust Model Development . 30

Figure 12. Experimental Results and Polynomial Fit for RPM to Thrust

Correlation . 31

Figure 13. Forward Lateral Tunnel Thruster . 32

Figure 14. Complete Tunnel Thruster Configuration . 32

Figure 15. Experimental Results and Polynomial Fit for Reverse Thruster

Direction . 34

Figure 16. Experimental Results and Polynomial Fit for Forward Thruster

Direction . 35

Figure 17. Simulated AUV Mission Using Prestero Parameters and Coefficients . 41

Figure 18. Simulated AUV Mission Using Doherty Parameters and Prestero

Coefficients . 42

Figure 19. Simulated Mission Using Doherty Parameters and Coefficients . 43

Figure 20. Correlation between REMUS 100 RPMs and Forward Velocity

from Mission Data . 45

IX

Figure 21. RPM to Forward Velocity Correlation for Mission and Model Data . 47

Figure 22. Federated Form of Unscented Kalman Filtering. Source: [17] . 57

Figure 23. Terminal Homing Mission with Consistent USBL Data . 58

Figure 24. Final Filter Configuration Design . 60

Figure 25. Successful Simulated Mission Using Hydrodynamic Model (3D

space) . 65

Figure 26. Successful Simulated Mission Using Hydrodynamic Mode (Y-X

Plane) . 66

Figure 27. Mission Data from Simulated Mission Using Hydrodynamic Model . 67

Figure 28. Surveying Mission of CAVR REMUS 100 (3D Space) . 68

Figure 29. Surveying Mission of CAVR REMUS 100 (Y-X Plane) . 69

Figure 30. Simulated Mission Utilizing Variable, Increasing Forward Velocity . 70

Figure 31. Vehicle Data from Simulated Mission Utilizing Variable Forward

Velocity . 71

Figure 32. Simulated Mission Utilizing Decreasing Variable Speed . 73

Figure 33. Vehicle Data for Simulated Mission Utilizing Decreasing Variable

Speed . 74

Figure 34. General Error Control Instituted for Tunnel Thrusters . 75

Figure 35. Simulated REMUS Mission with Docking Component . 76

Figure 36. Simulated REMUS Mission with Docking Component (Zoomed) . 77

Figure 37. Tunnel Thruster RPMs for Simulated Mission with Docking

Component . 77

Figure 38. Linear Velocities and Angular Displacement for Simulated Mission

with Docking Component . 78

Figure 39. User Generated Measurement Data . 79

Figure 40. User Generated Data Final Filtering Run . 80

Figure 41. Positional Uncertainty Developed from Covariance Matrices . 81

X

Figure 42. Filtered Solution Weighting INS and USBL Measurements . 84

Figure 43. Trace of Covariance Matrices for Filtered INS-USBL Comparison . 85

Figure 44. Filtering Solution Comparing USBL and Propagated Position

Estimate . 86

Figure 45. Trace of Covariance for USBL and Propagated Estimate Comparison . 87

Figure 46. Final Filtered Solution Compared to Vehicle State Position . 88

XI

THIS PAGE INTENTIONALLY LEFT BLANK

LIST OF TABLES

Table 1. REMUS 100 Vehicle Specifications. Source: [5] . 14

Table 2. D-USBL Performance Specifications. Source: [5] . 17

Table 3. Kearfott SeaDeViL INS System Specifications. Source: [22] . 18

Table 4. 6 Degrees of Freedom Notation Used with REMUS AUV . 22

Table 5. Applicable Force Equation for RPM Command . 36

Table 6. Distances of Each Tunnel Thruster to Center of Buoyancy . 37

Table 7. Vehicle Parameters during Coefficient Development . 40

Table 8. Coefficients Altered for Depth Control Performance . 44

Table 9. Coefficients Altered for Heading Control Performance . 44

Table 10. Model Drag Coefficients Required to Achieve Appropriate Forward

Velocity . 47

xiii

THIS PAGE INTENTIONALLY LEFT BLANK

XIV

LIST OF ACRONYMS AND ABBREVIATIONS

ADCP

Acoustic Doppler Current Profiler

AUV

Autonomous Underwater Vehicle

CAVR

Center for Autonomous Vehicle Research

DOF

Degree of Freedom

DR

Dead Reckoning

DVL

Doppler Velocity Log

EKF

Extended Kalman Filter

GPS

Global Positioning System

INS

Inertial Navigation System

KF

Kalman Filter

LBL

Long Baseline

MBARI

Monterey Bay Aquarium Research Institute

NED

North East Down

NPS

Naval Postgraduate School

ODE

Ordinary Differential Equation

PF

Particle Filter

PUC

Positional Uncertainty

REMUS

Remote Environmental Measuring Units

RPM

Rotations per Minute

SNAME

Society of Naval Architects and Marine Engineers

UKF

Unscented Kalman Filter

USBL

Ultra-short Baseline

USE

Unmanned Systems Lab

uuv

Unmanned Underwater Vehicle

WHOI

Woods Hole Oceanographic Institute

XV

THIS PAGE INTENTIONALLY LEFT BLANK

XVI

ACKNOWLEDGMENTS

First and foremost, I would like to thank my family and friends for supporting me throughout this past year. I could not have completed this thesis without them. Next, I would like to thank and acknowledge my advisor. Dr. Douglas Homer, for consistently giving up his time to help and guide me throughout the entire thesis process. I would also like to thank Sean Kragelund, Noel Du Toit, and Aurelio Monarrez for their assistance throughout my time here at NFS.

xvii

THIS PAGE INTENTIONALLY LEFT BLANK

xviii

L INTRODUCTION

A. MOTIVATION FOR THIS WORK

Unmanned Underwater Vehicles (UUV) have been part of inventories for Navies around the world for several decades [1]. However, the known extent of the usefulness of these vehicles has largely grown within the last 10-15 years. A report to the Chief of Naval Operations in 2001 believed that UUVs may augment or completely replace divers and mammals in the coming future due to their mission duration time, reduced risk- management, and cost effectiveness [2]. A renewed focus was put on UUV development and implementation during the 2000s. This resulted in research and development of new capabilities that included exploring the feasibility of using UUVs for inspecting ship hulls, surveying water columns and bottom type, testing non-GPS reliant navigation techniques, and expanding mine counter measure efforts [3].

This ongoing discovery of potential uses however is consistently plagued by one aspect of the underwater domain, accurate navigation. Above the water, light and electromagnetic signals travel well through air and space, mediums that allow for a variety of localization systems for vehicles, the primary method being the Global Positioning System (GPS). However, the characteristic properties of air and space that make them good mediums for these signals, do not carry over to the undersea domain. Compared to air or space, water inhibits the passage of light and electromagnetic signals, and drastically reduces the effectiveness of electro-optical and electromagnetic sensors. This essentially eliminates the use of GPS navigation while submerged. The only signal that performs adequately in water is sound. As a consequence, many UUV systems rely on acoustic beaconing systems that, due to the oceanographic characteristics, can be less accurate and can negatively impact operational flexibility. For position estimation the external beacon system augments navigational filters that use Inertial Navigation Systems (INS) and Doppler Velocity Logs (DVL).

Dead reckoning is recognized as the first and oldest form of underwater navigation dating back to David Bushnelfs Turtle in 1776 and is simply the estimation of

1

position given a known direction and speed, which for UUV would consist of a compass heading and a correlation between propeller revolutions and forward velocity [4]. Significantly more sophisticated than DR is INS, which integrates the data collected from accelerometers and gyroscopes to accurately estimate a vehicles position given an initial starting position. Similar to surface navigation, an underwater vehicle’s positional uncertainty (PUC) grows in between vehicle fixes, whether it be GPS or acoustic fixes. Unlike surface navigation, an UUV receives only an initial GPS fix at the start of a mission and relies solely on DR or INS for the extent of the mission, or until it resurfaces for another fix. The PUC for INS system on board the REMUS 100 increases at rate of roughly 0.45% distance traveled in benign conditions, which means an UUV conducting a mission for an hour at 3 knots could be over 25 meters off course. For DR models it can be significantly worse as it is difficult to factor in the effects from the environment (i.e., currents and waves), and the direction and speed measurements typical maintain a higher degree of inaccuracies. The best current method of reducing the PUC of an UUV, without surfacing for a GPS fix, is using acoustic baseline systems, specifically a long baseline line (LBL) system. This method involves deploying acoustic transponders at known locations to communicate with the UUV using sound. The vehicle sends an initial sound signal to the transponders after which the transponders send a reply message. Using Equation (1), where c is the speed of sound and t is the time it took to travel between vehicle and transponder, the vehicle can determine the distance to each transponder.

d = St (1)

These distances are then used to determine an estimate of the vehicle’s location via intersecting circles shown in Figure 1, where the two position solutions, represented by the red X’s, are developed from using two transponders, represented by DT4C and DTID. While this process can provide the vehicle with a fairly accurate fix, it comes with its own limitations.

2

DT4C’t Ra»%9® Ring ^ -

Figure 1. Position Solutions Using Two Transponders. Source: [5].

The first limitation of the acoustic baseline system is that it is an active sound system, which means it cannot be used in mission requiring discretion because it is actively putting detectable sound into the water. The second limitation of the system is that fix accuracy is greatly dependent on the geometry associated with the system setup. If the vehicle operates to close to the baseline, the line between two transponders, the vehicle can become “confused” as it is given two possible position solutions, and thus limits the operational area of the UUV given the system setup. Additionally, the acoustic system’s accuracy is a function of range due to signal attenuation, which means as the distance between the vehicle and the transponder grows the accuracy decreases due to the possibility of multi-path between transmitter and receiver. As a result, acoustic baseline systems are rated to specific distances after which fix accuracy becomes no longer acceptable. The last limitation of acoustic systems is that the vehicle must have an understanding of the water and environment characteristics in order to determine the value of sound speed to accurately determine position. The predominant governing factors of sound speed are temperature, salinity, and depth, which all must be predetermined or measured by the vehicle.

3

Once the vehicle determines it has a fix, it must then determine whether it is accurate. This method of determining fix accuracy is another ongoing field of study in UUV navigation. It is difficult because establishing ground truth in the ocean experimentation is difficult. Still accurate vehicle position estimation is incredibly important not only for general vehicle navigation but also for the wide variety of missions UUVs can support. For missions tasks such as terrain mapping, mine detection for mine clearance operations, and vehicle docking, knowing exact positions is critical to ensure mission success.

B. PROBLEM DESCRIPTION

This thesis investigates AUV position estimation for undersea docking. The undersea platform is a Remote Environmental Measuring Units (REMUS) 100 autonomous underwater vehicle (AUV) maintained by the Center for Autonomous Vehicle Research Lab (CAVR) at the Naval Postgraduate School (NPS) in Monterey, California.

The deployment and recovery techniques of AUVs typically varies given the vehicle size and mission objective. In the case of long endurance missions it could make strategic and financial sense to keep the vehicle in the area of operation without human assistance and instead use a docking station to recharge the vehicle, download and transmit mission data, and then reprogram it for its next mission. The use of a docking station could increase mission efficiency by allowing the AUV to remain in the operational area for longer while also reducing necessary manpower and potentially necessary funding.

A difficulty experienced with the standard REMUS 100 while docking is controlling vehicle movement within close proximity to the docking station. While the initial trajectory can be within a few meters of the docking station the final trajectory must be accurate within less than one meter in order to successfully dock. For the standard REMUS 100 configuration, shown in Figure 2, these trajectory alterations are made with the pitch and rudder fins.

4

Magnet Switch External ballaaUtnm weights AOCP Array

(upper and lower)

Figure 2. Standard REMUS Configuration. Source: [5].

However, given the “torpedo-like” configuration of the vehicle the fins must have a sufficient flow of water over them to provide the necessary lifts forces to steer the vehicle. For the REMUS vehicle this sufficient flow roughly equates to 2-3 knots of headway as it attempts to dock. While 2-3 knots is not seemingly a fast speed, it is fast enough to cause damage to the docking station and vehicle if the trajectory is incorrect, as seen through the partially broken guidance rails of the docking station and damaged Ultra-short Baseline (USBL) sensor in Figure 3.

Damage caused by unsuccessful docking attempts at 2-3 knots. Damaged areas are indicated by red circles.

Figure 3. Damage Incurred from Unsuccessful Vehicle Docking

5

A solution to this problem was to reconfigure the REMUS vehicle to include horizontal and vertical cross-body thrusters in the forward and aft sections, seen in Figure 4.

Figure 4. Updated, Non-Standard REMUS Configuration

The cross-body thrusters allow for higher control capability at slower forward velocities, which looks to help reduce potential vehicle and docking station damage and increase docking efficiency. However, given the non-standard vehicle configuration of the REMUS 100, the developed hydrodynamic simulation models are no longer accurate and must be altered to reflect the new configuration. To gain a larger understanding of the capabilities of the new REMUS 100 configuration this thesis looks to develop a new hydrodynamic simulation model that incorporates the larger vehicle parameters as well as the tunnel thrusters.

As discussed previously, an AUVs PUC increases after the initial vehicle fix. Unless the PUC is reduced through additional vehicle fixes, the vehicle will begin the docking task with a potentially inaccurate positional estimate. Therefore, the initial trajectory the vehicle uses to approach the docking station will be inaccurate and needs to be corrected in order to successfully dock. The current method used by the REMUS 100 is an on board filtering algorithm that determines, what it believes to be, the most accurate position estimate by incorporating all available sensor infonnation. However, past experience conducting docking operations has shown that this on board position estimation filter is not necessarily accurate. These inaccuracies also contribute to the vehicle damage shown in Figure 3. Due to the proprietary nature of the REMUS software and therefore the filtering algorithm it is not fully known how the vehicle determines a positional estimate. In order to gain greater clarity and ultimately gain greater accuracy.

6

the position estimate using the REMUS 100 sensor systems was also investigated in this thesis. The specific REMUS 100 sensor systems incorporated into this filtering process were the INS, USBL, and the Acoustic Doppler Current Profiler/Doppler Velocity Log (ADCP/DVL).

Given the growing error in the INS solution and the potential errors associated with the USBL solution, the problem arises on how best to filter the two systems into providing an accurate estimate of position as the vehicle approaches the docking station. The combination of an INS and USBL system generate an additional difficulty, which is they operate at different frequencies causing asynchronous data generation. Additionally, data collected from previous REMUS missions show frequent intermittent data generation performance with the USBL system, which adds complexity to the asynchronous behavior. While a variety of filtering algorithms exist that can combine several sets of data to develop a combined solution, much less have been developed that can handle several sensors with asynchronous and intermittent data generation. While it was initially believed that the INS velocities would be more accurate than the velocities measured by the integrated ADCP/DVL sensor system, the ADCP/DVL velocities were tested in the filtering process nonetheless.

C. LITERATURE REVIEW

The following sections seeks to outline the current status of the specific areas to be investigated in this thesis. A primary intention of this thesis is to not completely reinvent docking stations, hydrodynamic models, or position filters but rather work to improve the current methods. Keeping this primary intention in mind, the organization of this literature review consists of three discussion areas: the current status of UUV hydrodynamic models, the current docking stations in use or being tested, and the current methods of data or sensor filtering, specifically related to position estimation.

1 . Hydrodynamic Models

A large area of study within undersea environment is the study and development

of motion models for underwater vehicles. A leader in the field of underwater vehicle

motion is Fossen, whose 6 degree of freedom (DOF) motion model is widely accepted

7

and used a starting point for maritime vehicle models [6]. Fossen’s general model requires further development and calculation of a variety of coefficients, for specific platforms. These coefficients have been developed for the REMUS 100 AUV by Prestero [7] and further tested and presented by Sgarioto [8]. A limitation with these models is that they are developed around the standard REMUS 100 configuration, and the configuration of the CAVR REMUS used in this thesis is approximately double the length and mass. In order to successfully apply these developed models, the vehicle coefficients must be altered to fit the larger vehicle parameters A thesis recently presented by Doherty [9]. redeveloped Fossen’s coefficients using Prestero’s method for the CAVR REMUS in nearly the exact same configuration. The only difference between Doherty’s configuration and the configuration presented in this thesis is the nose end cap for Doherty was a forward looking sonar instead of an USBL. Therefore, Sgarioto’s model and MATLAB code in conjunction with Doherty’s coefficients serves as the most accurate baseline to use in the development of a motion model to the current REMUS configuration.

Another limitation of the Fossen model is the limited discussion on the thruster force developed by the vehicle’s propulsion system. Several models have been presented to fill this void but the three-state thruster model developed by Blanke et al. [10] and then further applied specifically to the REMUS 100 by Sgarioto [8] was determined to provide the most thorough solution and was used as a baseline thruster model for the aft thruster for this thesis.

2. Docking Station

The concept of docking a AUV in a docking station to recharge, download data, and reprogram has been widely studied over in recent years. The original producers of the REMUS 100 vehicle used in this thesis developed and tested their own docking station [11], which was subsequently put into production and is used in this thesis as the docking station. The docking station design created by Wood Hole Oceanographic Institution (WHOI), shown later in Figure 11, is a fairly small and compact design and easily deployable and functional in a variety of locations. The basic design and shape of

8

the docking station is also fairly common as similar docking stations developed by the Monterey Bay Aquarium Research Institute (MBARI) [12] and by researchers at Zhejiang University in China [13], which are shown in Figures 5 and 6, respectively, are seen to resemble and use similar systems.

Figure 5. MBARI Docking Station. Source: [12].

Figure 6. Zhejiang University Docking Station. Source: [13].

MBARI research results cover several similar aspects that are investigated in this

thesis, namely the use of USBL for terminal homing and the development of a control

algorithm for cross track error control during the terminal docking approach. The

approach differs from the WHOI docking station in that the MBARI docking station

9

maintains thrusters that allow the docking station to pivot to compensate for error in the AUV terminal approach. Additionally, as seen in Figure 6, the researchers at Zhejiang University are operating their vehicle at the surface and therefore have high video quality as well as access to GPS.

3. Position Estimation Filtering

There is little research with respect to undersea position estimation during terminal homing. As a result, the literature research was expanded to look for filtering approaches to specific aspects of the terminal homing problem, specifically handling asynchronous data measurements and combining the outputs of several sensor systems. Additionally, research was conducted to determine what specific type of filters would serve the positional estimation problem for terminal homing best.

As previously discussed, a large difficulty with the position estimation for terminal homing of AUV sensor systems, specifically those available on the CAVR REMUS 100, is the asynchronous and intermittent behavior of the sensor data. The sensor systems on the CAVR REMUS 100 vehicle operate at different frequencies and do not always deliver usable data. A prime example is the USBL system whose performance is effected by a variety of factors including the range to the acoustic transponder and surrounding environmental conditions. As a result, USBL measurements can be sporadic and all of the vehicle’s sensors’ data cannot always be combined at every time step. This can result in a degraded position estimate.

Recent work pertaining to this problem has produced several potential solutions. Armesto et al. [14] developed a multi-rate fusion algorithm combining vision and inertial sensor systems for surface robot tracking with 6 DOF. The model accounts for differing sampling times by altering the measurement and output steps to reflect whether measurement data is available, and also utilizes an input hold mechanism to maintain the same input vector. The paper also compares the effectiveness of the implementing this algorithm with an EKF and UKF, and finds the UKF to be slightly more accurate but at seven times the processing cost. A similar process was conducted by Geng et al. [15] as they developed a hybrid derivative-free EKF filter for USBL and INS tightly coupled.

10

The derivative-free EKF combines the linear time propagation technique utilized by the standard EKF and the non-linear measurement propagation, similar to the technique utilized by the UKF. The results showed that the derivative-free EKF operated as well as the UKF for integration navigation [15]. However, the assumptions made with tightly coupled USBL and INS are not necessarily reflective for the REMUS vehicle and thus the results are not entirely applicable to this thesis.

Additional methods considered for dealing with asynchronous data as well as weighting or combining measurement estimates were presented in the works of both Syahroni [16] and Ali et al. [17]. Both papers use decentralized UKFs in a federated configuration to handle multi-sensor data fusion. The process consisted of individual UKFs that are associated with specific sensor systems that all feed into a master UKF that weighs the supporting sensor systems based on user defined coefficients. This decentralized model prevents the possibility of data overload by dispersing the data across several UKFs. The federated configuration also allows the filter to easily account for asynchronous or intermittent sensor behavior by simply adjusting the weighting coefficient for each UKF. This capability allows the master filter to operate largely uninterrupted, regardless of the sensors’ operating frequencies or data quality.

The final works considered in this thesis dealt with determining what specific types of filters would best suit the position estimation problems associated with terminal homing. The first of these works was conducted in the CAVR lab research by Dillard [18]. Dillard’s work focused on the positional estimation a quadcopter using bearing and range estimates to surrounding beacons. The filtering methods explored by Dillard were the extended Kalman filter (EKF) and the unscented Kalman filter (UKF) and ultimately used the UKF over the EKF. While Dillard’s work showed the EKF to required less processing power, the UKF, similar was found to be able to handle the non¬ linearity associated with the bearing measurement better than the EKF [18]. This capability of handling non-linear measurement is believed to be a critical aspect to this thesis as the USBL measurements are non-linear in their calculation process and are sporadic in practice that results in greater non-linearity. Further works by van der Merwe et al. [19], Sarkka [20], and Allotta et al. [21] also presented an improved capability of

11

the UKF over the EKF while handling non-linear data. The deterministic sampling approach utilized in these works by the UKF was shown to be more realistic and less complex than the linearization used by the EKF. However, the EKF is still considered as a potential solution due to its lower required processing power as well as its capability to easily represent growing error or uncertainty with the sensor measurements.

D. THESIS ORGANIZATION

This thesis is organized as follows: Chapter II provides a description of the specific equipment and sensors used in this thesis, Chapter III delves into the governing equations and development process of the hydrodynamic model. Chapter IV delves into the governing filter equations and filter development process. Chapter V presents the results of hydrodynamic model and filtering process, and Chapter VI provides conclusions as well as recommendations for future work.

12

IL SYSTEM DESCRIPTION

A. REMUS 100

The standard REMUS 100, shown in Figure 2, was developed by WHOI and produced by the Kongsberg Company and is a man-portable, shallow water vehicle designed for a variety of operations that include

hydrographic surveys,

mine counter measure operations,

harbor security operations,

environmental monitoring,

debris field mapping,

search and salvage operations, and

scientific sampling and mapping. [5]

The AUVs are rated to operate in depths up to 100 meters, range between 2 and 2.5 meters in length, and weigh between 50 and 65 kilograms. Table 1 further explains the REMUS 100 specifications.

13

Table 1. REMUS 100 Vehicle Specifications. Source: [5],

Physical/Functional Area

Characteristic

Metric

US

Vehicle Diameter

19 cm

7.5 in

Vehicle Length

2.22 m

7.28 ft

Vehicle Length with Forward looking

Sonar

2.72 m

8.94 ft

Vehicle Length with Large USBL

Attachment

2.48 m

8.14 ft

Vehicle Length with Small USBL

Attachment

2.26 m

7.41 ft

Weight in air

52 kg

114.6 lbs

Weight in air with Forward looking Sonar

64.5 kg

142.2 lbs

Weight in Air with Large USBL Attachment

58.4 kg

128.8 lbs

Weight in Air with Small USBL Attachment

52.5 kg

115.6 lbs

External Ballast Weight

1 kg

2.2 lbs

Operating Depth Range

3 m to 100 m

10 ft to 328 ft

Typical Search Area

1200 m X 1000 m

1312 yds X 1083 yds

Rated Transponder Range

2000 m

2187 yds

Rated Acoustic Communications Range

2000 m

2187 yds

Operational Temperature

In Air:

In Water:

-24 to +43 deg C -2 to +43 deg C

-11.2 to +109 degF +28 to +109 deg F

Speed Range

0.25 to 2.57 m/s

1.0 to 4.0 knots

Maximum Operating Water Current

1.0 m/s

2 knots

Max Recommended Operating Sea State

Sea State 3

Battery

1 kWh internally-rechargeable Lithium-ion

Typical Endurance

20 hrs at 2 knots;9 hours at 4 knots

Propulsion

Direct drive DC brushless motor directly connected to open three bladed propeller

Control

2 coupled yaw and pitch fins

Navigation Modes

Long baseline, ultra short baseline, dead reckoning, GPS

The specific REMUS 100 vehicles used throughout this thesis, shown in Figure 8, are modified versions of the REMUS 100 AUV, and supplied by CAVR. The two REMUS vehicles used in this thesis are referred to by their vehicle numbers 231 and 359. The two vehicles are nearly identical except that vehicle 359 is Wi-Fi enabled while vehicle 23 1 is not.

14

Figure 7. Vehicle -23 1 with D-USBL Nose Attachment

Each vehicle operates with the following sensors and systems, [5]:

downward looking acoustic Doppler current profiler (ADCP) and Doppler velocity log (DVL)

acoustic modem

magnetic compass

NMEA 183 GPS/ Iridium antenna

forward and aft, horizontal and vertical cross-body tunnel thrusters,

YSI-600 conductivity, temperature, and depth sensor

Marine Sonic Technology Limited dual frequency sidescan sonar

Kearfott INS

modular end cap with optional Digital USBL (D-USBL), video camera recorder, and forward looking sonar attachments

The three sensor systems critical to this thesis are the D-USBL, Kearfott INS, ADCP/DVL. However, the primary sensor system investigated in this thesis is an acoustic baseline system known the USBL. The USBL system is attached to the nose of the AUV, shown in Figure 7, and communicates with a transponder attached to the docking station. The USBL receives the docking station signal and through a four- channel planar hydrophone array determines the bearing and range to the docking station, thus obtaining a vehicle fix. However, the accuracy of the USBL tends to degrade faster than a LBL system as the range from the transponder increases. The setup of the hydrophone array also requires the docking station signal to be within a 35 degree swath

15

in front of the vehicle in order to achieve a successful fix. As a result signals received outside of this swath can be fairly inaccurate. Figure 8 shows the specific setup necessary for successful USBL fixes between the docking station and the AUV. Further sensor specifications are shown in Table 2.

Transponder within ± 17.5"* from Expected bearing of -10^ from dead ahead.

Example 1)

Good D-VSBi Solution

r Transponder

Transponder is at ±