Design of an Acoustic-Homing Autonomous Surface Vessel Lauren Cooney, Michael J. Stanway, Peteris Augenbergs, Heather Brundage, Bridget Downey, Timothy Pennington, Thaddeus Stefanov-Wagner and David Tobias Department of Mechanical Engineering Massachusetts Institute of Technology Cambridge, Ma 02139 USA Email:
[email protected] Abstract-We designed, built, and conducted trials on an acoustichoming autonomous surface vessel. After characterizing the performance of a commercially available kayak hull, we installed power, propulsion, and control systems for autonomous operation. The acoustic tracking payload was built up from basic components. We overcame hardware limitations in the microcontrollers and demonstrated acceptable average accuracy in locating the acoustic target. This project demonstrated the feasibility of the overall goal of using a small surface vessel to follow a submerged acoustic beacon.
Figure 1. ASV operating on the Charles River.
I. INTRODUCTION
III. PLATFORM E VALUATION
The goal of this project was to create a new capability in ocean observation, by constructing a small, autonomous surface vessel system capable of tracking a subsurface acoustic source. This challenge represents an important problem in ocean engineering. Current surveying techniques using autonomous underwater vehicles (AUVs) lack feedback from the AUV, and provide no real-time data to check the status and progress of the robot while it is underway on a mission. This type of vessel would serve as a mobile communications link between an autonomous underwater vehicle (AUV) and a shipboard or onshore command station. It would use an acoustic modem to communicate with the AUV and relay those communications by radio or satellite to the operator. It is believed that by maintaining position near the AUV, this vessel could provide sufficient communications bandwidth for real-time data upload and navigation feedback to the AUV.
A Wilderness Pungo 120 3.7m kayak serves as the platform for this project. This multi-chine hull features a full-length keel line, providing a good mix of stability and maneuverability. Several different experiments and theoretical analysis were performed to evaluate its seaworthiness and performance characteristics. A. Inclining Experiment A variation on the standard inclining experiment was performed in order to develop righting moment curves for the kayak hull under several possible payload conditions. The results, plotted in Fig. 2, show a strong righting moment, and suitable initial stability for our application. From this data, the metacentric height of the kayak hull was found to be 0.23m with a payload of 100kg. B. Initial Condition Response A two-axis accelerometer and a TattleTale Model 8 microcontroller (TT8) were used to measure and record the roll angle of the hull in an initial condition response test. The
II. OBJECTIVES The autonomous surface vessel (ASV) for this project was designed to have sufficient static and dynamic stability to function in sea state 3 conditions. This required adequate payload capacity for the acoustic tracking system and all propulsion, power, and control systems. The ability to move faster than 1m/s was desired to track a moving target. In order to demonstrate acoustic tracking capability, an acoustic system needed to be developed, determining range and bearing to the target, with all acoustic sampling and processing performed onboard.
Figure 2. Righting moment curves for Pungo 120 hull.
tunnel [3]. The foil-shaped struts were chosen for better drag characteristics and also to enhance the directional stability of the vessel. Additional resistance due to waves was investigated [4], and determined to be negligible for the system in the expected operating conditions.
Figure 3. Damped natural frequency and damping ratio vs. payload mass for Pungo 120 hull.
averaged natural frequency of damped oscillations and the damping ratio as found using the logarithmic decrement for each payload mass measurement are presented in Fig. 3. At the maximum response for both quantities of 100kg payload, the damped resonant frequency was approximately 1.2 Hz, well above the expected sea state 3 wave frequency [1] of approximately 0.13 Hz. C. Projected Wave Response A simple second order model of the Pungo 120 roll response was constructed from the inclining and initial condition roll response test results:
A transfer function was generated using this model, and input sea slope spectra were determined from the Bretschneider Spectrum. From these simulations, the expected roll is 9 degrees in average sea state 3 conditions, with a significant wave height of 0.88m and wave period of 7.5sec. This shows that the Pungo 120 requires no external appendages for roll stability in the operating conditions. D. Resistance & Powering The frictional resistance of the kayak hull was estimated according to the International Towing Tank Conference (ITTC) 1957 model-ship correlation line [2]. Fig. 4 shows a comparison between expected performance with circular struts and streamlined, foil-shaped struts. Performance of the Riptide motor was estimated through a nondimensional comparison to a similar motor that had been thoroughly tested in a water
Figure 4. Total vessel resistance & thrust available.
E. Directional Stability A simple push test was performed during tow tank tests, and it was found that the bare kayak hull did not always track straight (i.e., it was directionally unstable). This meant that its hydrodynamic center was near its center of mass. The streamlined hydrophone struts act as lifting surfaces that pull the hydrodynamic center aft, behind the center of mass. The coefficient of lift for the struts was calculated using JavaFoil [5]. Using a moment balance, the new hydrodynamic center was found 1m aft of the center of mass (Fig. 5). Propulsion from any point forward of the hydrodynamic center (as in our design) will provide a stabilizing term in the steering response of the vessel. IV. ACOUSTIC TRACKING SYSTEM A. Acoustic Components & Layout The following acoustic components are utilized in the acoustic tracking system: an International Transducer Corp. ITC-1001 spherical, omnidirectional transducer; a Sonardyne Type 7656 transponder; and three Sensor Technology Limited SQ03 hydrophones. The components are arranged in a short-baseline (SBL) sonar tracking configuration as shown in Fig. 6. The transducer generates an interrogation signal from the vessel at a frequency of 20.492 kHz. The transponder (target) detects the interrogation signal and transmits a reply signal at 29.7kHz
Figure 5. Moving the hydrodynamic center by using foil-shaped struts.
Figure 6. Layout of a short-baseline acoustic tracking system.
after a 30msec internal delay. The hydrophones, mounted on the vessel, then detect the reply signal which is recorded for later processing. The transducer is mounted through the hull in the central cockpit. One hydrophone is attached forward of the cockpit to the support shaft of the trolling motor. The two hydrophones at the stern are mounted on struts to provide the separation required for a short-baseline acoustic system. B. Control & Sampling Circuitry Custom circuitry was designed and built in order to interface between the TT8 and acoustic components. A flow chart is shown in Fig. 7. A ping generator was developed to provide the link between the TT8 and the audio amplifier/transducer interrogation system. When sent a square wave pulse from the TT8, a circuit containing an Exar XR2206 monolithic function generator produces the 20.4 kHz zero-mean sinusoid to be output by the audio amplifier/transducer. From initial experiments, it was determined that the 2 kHz, three-channel sampling frequency of the TT8 analog-to-digital converters aliased the 30 kHz transponder reply signal. The solution to the problem of aliasing was to modulate the incoming signal from each hydrophone with a carrier sinusoid of a known frequency, using an Analog Devices AD630 balanced modulator/demodulator and a second XR-2206 function generator to create the carrier sinusoid of 29 kHz. Other experiments showed the need for variable amplification of the hydrophone signals depending on range to the source, and narrow bandpass filtering of the hydrophone signal. Solutions to these problems were provided by Frequency Devices’ D68BP filter and D83S programmable amplifier respectively. C. Software Processing & Control Processing algorithms were developed in order to demonstrate acoustic homing while accommodating the limitations inherent in the TT8, namely, budgeting available memory and CPU load. An autocorrelation calculation is performed on each signal in order to determine time delays between signals. This method is demonstrated in Fig. 8, essentially a summation of two signals such that that the index
Figure 8. Autocorrelation of two hydrophone signals to find time delay.
of the calculation peak occurs at the phase lag between signals. This time-and-memory-expensive calculation created an issue in the TT8’s limited memory allocation. The solution was a streamlined acoustic processing program which utilized storage space to full capacity. A flow chart of the program follows in Fig. 9. The acoustic processing program is initialized when the main TT8 enters Acoustic Mode. The programmable amplifiers on the sense/modulation board are set to the corresponding signal amplification level for the desired range. The program then sends a pulse to the ping generator, interrogating the transponder, which responds back to the ASV. The program delays for the two-way travel time of the transducer/transponder pinging, and then records the three modulated and filtered signals on the TT8’s A/D channels for approximately 250msec of sampling time. After the analog-to-digital conversion, the signals must be renormalized. To confirm that the transponder response was received within the sampled time period, the energy as the sum of the squares for each signal vector is calculated and required to be within a predefined threshold limit to continue. Signals are shifted within each of the three vectors to allocate space, and autocorrelation is then done to determine time delays between signals. The triangulation matrix described in the next paragraph then determines bearing to the transponder.
Set Variable Amp
Ping Transducer
Delay
Record Processed Signals
Acoustic . Mode ON
Subtract Mean
Main TT8 No. Adjust Amplifier Level/Delay Range & Bearing Triangulation
Figure 7. Acoustic signal flow path.
Autocorrelation
Shift Window
Figure 9. Acoustic sampling flow chart.
Energy Calc. Signal Received? Yes.
D. Triangulation In Fig. 10, the transponder, shown by the blue dot, is at a range, r, and bearing, relative to the vessel. The polar axis, or the 0° bearing, is the line drawn from the transducer through the bow hydrophone. The radial distance and polar angle to each of the hydrophones is represented by the case of hydrophone 3, or H3 as it is labeled in the figure. R3 is the radial distance and θ3 is the polar angle. The geometry of the onboard system must be controlled in order to resolve the range and bearing to the transponder. The struts used provide a 2.4m separation between the stern hydrophones. Using the geometry shown in Fig. 10 it is possible to construct a series of trigonometric relationships relating the distances and angles to the varying times at which the reply signal reaches each hydrophone. By assuming that r is much larger than any Ri, the two-way distance for the travel of the sounds waves from the transducer (origin) to the transponder and back to the hydrophone is equal to the distance from the origin to the transponder and to the transducer while subtracting the projections of the hydrophone’s distance from the origin onto the range vector. Then, dividing the distances by the speed of sound in water, c, the following generic equation is used to represent the signal travel time:
These equations can be transformed using trigonometric identities and as a series of three linear equations with three unknowns: 2r, cosθ, and sinθ. They can be represented as a single matrix equation:
By inverting the matrix populated with hydrophone geometries, we then solve for the two-way range, the cosine of the bearing, and the sine of the bearing. Since the matrix to be inverted is based entirely on the geometry of the onboard acoustic system, the matrix only needs to be inverted once ahead of the mission, and then can be stored on the TT8 to save crucial computation time and memory space.
Figure 10. Geometry of SBL acoustic tracking system.
Figure 11. High level control system signal flow path.
V. HIGH LEVEL CONTROLS In order to accomplish the mission tasks, a control system was designed to enable the boat to operate autonomously. Three TT8s were used on-board which communicate with each other using serial TPU lines and RS-232. One TT8 controls the acoustic system, another the motor controller, and the third acts as the main controller, interfacing wirelessly with the user through an RF modem connected to a laptop. The main TT8 also reads in data from a PNI TCM2 compass and handheld eTrex Legend GPS that uses the WASS GPS correction system. Three different control modes were designed to meet mission objectives, as shown in Fig. 11. The first mode was a User Mode, developed to aid in testing. This mode allows the user to input the desired heading and speed of the kayak, which is then continually compared to the actual heading as received from the compass and adjusted accordingly. If the User selects Acoustics Mode, the main TT8 continuously looks for new data from the acoustic system, compares that data to the current heading and sends the appropriate commands to the motor controller TT8. In GPS Mode, the kayak’s motion is determined by the GPS and a set of waypoints that the boat navigates. Due to time constraints and the group’s prioritized objectives, the third mode was never completely implemented. VI. TOTAL SYSTEM INTEGRATION A. Power System The kayak is powered by three lead-acid batteries. Two large 79 amp-hour marine deep cycle 12 V batteries provide power for the electronics, trolling motor, and acoustics driver, while a smaller, 8 amp-hour battery provides -12V for the acoustics system audio filters. B. Propulsion Unlike a conventional ship with a stern mounted propeller and rudder, propulsion is provided by a bow-mounted MinnKota Riptide trolling motor. An 80lb bollard thrust, 24V Riptide trolling motor was chosen. This unconventional design has the advantage of increased heading stability of the kayak and keeps the motor and its associated electrical, magnetic and acoustic noise a safe distance away from other
Figure 13. Azimuth motor and Riptide motor assembly installed on ASV.
sensitive systems. The trolling motor has an aluminum sleeve that goes down to the motor and surrounds a smaller fiberglass tube. Steering is accomplished by turning the fiberglass tube, which is directly attached to the motor, while the outer, aluminum sleeve stays stationary. The tube is mounted to the kayak using a series of aluminum brackets which are connected to clamp-on collars around the aluminum pipe. A DC servo motor provides azimuth steering for the Riptide trolling motor. Rate-limited proportional control was found to provide the best overall response for the system. A custom motor control board was designed, including pulse-width modulation (PWM) capabilities through MOSFETS configured as an hbridge, an Agilent HCTL-2016 quadrature decoder, and a Halleffect sensor and magnet to provide self-homing capability to steering.
C. Layout The main batteries, main controller and motor controller are mounted to a removable deck within the cockpit of the kayak and up out of the bilge (Fig. 14). The electronics enclosures mounted to the deck are splash-proof and all connections are made with Switchcraft splash-proof connectors. The main power distribution and emergency stop system as well as the trolling motor controller are mounted high on the side of the hull to decrease chance of exposure to water. All through-hull penetrations are made with waterproof cable grips.
ASV, tests were conducted in the Charles River to show that the completed system could meet the challenges as designed. We were successfully able to demonstrate the individual systems operating independently. During testing, the water was generally flat, sometimes with high frequency waves varying from mere ripples to approximately 10 cm in height. Significant winds were present at times, up to 11 knots. The boat response was tested while operating in User Mode. The ASV’s ability to stay on a desired heading was confirmed by comparing the desired heading with the actual heading. As can be seen in Fig. 15, the kayak initially overshoots the desired heading during high-speed turns, but after some oscillations is able to maintain the desired heading. There was not time to tune the control system to reduce the overshoot and oscillations, though there was a rate limit incorporated, and for a first pass, these results were satisfactory. The top speed of the kayak was observed to be 1.6m/s. Wave resistance also appeared to be non-negligible with the 2.4m struts, which were directly in bow wave of the kayak at full speed. The struts basically amplified the wave at that point, dumping more energy into the wake of the vessel. Since we could not test the kayak in sea state 3, we tried simulating larger, lower frequency waves by creating wake with a motorboat. The kayak’s responses to following and head on waves (Fig. 16) were both tested to our satisfaction, as the boat remained stable and never appeared to be approaching a significant roll angle. B. Acoustic System Due to last minute hardware issues, the boat was not able to be tested in full Acoustic Mode, though the acoustic system was demonstrated to be operating correctly. By hanging the transponder on the sailing pavilion dock and interrogating it
VII. TRIALS & SYSTEM PERFORMANCE A. Boat Response After integrating the acoustics and control system into the Figure 15. Heading control during high-speed turns. Bilge Pump
Main Controller
Motor Controller
Transducer Mount
Figure 14. Component layout in the ASV cockpit.
Figure 16. ASV in following waves.
using a piezo-electric element (instead of using the transducer), the kayak was towed with the on-board acoustic tracking system recording bearing relative to the current heading of the kayak. The heading of the kayak was approximated from position data recorded by an on-board GPS unit. Using the bearing to the transponder calculated by the acoustic system and the calculated kayak heading, the corresponding global heading to the transponder was found, as determined by the acoustic system. This was compared to the actual global heading to the transponder to evaluate the acoustic system’s accuracy. Fig. 17 demonstrates the acoustically-computed heading to the transponder for the traveled AHASV track, while Fig. 18 presents the error between the acoustically-computed heading data and actual heading to transponder. There is a large margin of error in the heading calculations from the acoustic system. The trial data gives reason to be optimistic, however, as the runs with the most accurate GPS heading approximations (ex. Data samples 10-20) coincided
with the most accurate bearing/heading calculations. Because the motor control system implements rate-limiting, we can speculate that as long as the acoustic system determines heading over a period of time with a mean error of zero, a mission will be successful. VIII. CONCLUSION This paper presents the development of an acoustic homing surface vessel. The ASV is shown to maintain desired heading during high-speed turns, as well as successful operation in simulated sea state 3 conditions. Encouraging results are presented for an acoustic tracking system in locating a subsurface beacon. Several major changes can and should be made before the system is put into production, both to improve the current system and to provide extended capabilities. Implementing a different microprocessor for the acoustic system would greatly simplify and accelerate the software processing program. An integrated motor control solution would provide a more robust and replicable system. Lithium ion batteries would allow longer mission times, however, at greater cost and complexity. To achieve full functionality, a directional acoustic modem would also have to be added to the payload. This system is specific to the equivalent system onboard the submerged AUV, and would have to be designed with both vehicles in mind. ACKNOWLEDGMENT We would like to acknowledge the contributions made to this project by the following individuals and organizations: Dr. Franz Hover, Professor Michael Triantafyllou, Professor Chryssostomos Chryssostomidis, the MIT Department of Mechanical Engineering, the MIT Center for Ocean Engineering, and the MIT Sea Grant College Program. REFERENCES
Figure 17. Acoustically-computed heading to transponder.
[1] [2] [3] [4]
[5]
Figure 18. Error between actual and acoustically-computed heading to transponder.
LEWIS, E. V. Principles of Naval Architecture. Jersey City: Society of Naval Architects and Marine Engineers, 1988. GILLMER, T. C. and B. JOHNSON, Introduction to Naval Architecture. Annapolis: United States Naval Institute, 1982. pg. 220 STETTLER, J. W. “Steady and Unsteady Dynamics of an Azimuthing Podded Propulsor Related to Vehicle Maneuvering.” Ph. D. Thesis, 2004. Massachusetts Institute of Technology. COHEN, S. B. “Experimental Results of Non-linear Seakeeping Motions, Wetted Surface and Sectional Force Tests on a Ship with Large Bow Flare.” 1995. University of Michigan, Department of Naval Architecture and Marine Engineering. HEPPERLE, M. “JavaFoil.” 2005.
