Approach to the Real-Time Adaptive Control of Multiple High-Frequency Sonar Survey Systems for Unmanned Underwater Vehicles

Approach to the Real-Time Adaptive Control of Multiple High-Frequency Sonar Survey Systems for Unmanned Underwater Vehicles Edward Thurman, James Riordan, Daniel Toal Mobile & Marine Robotics Research Centre Department of Electronic & Computer Engineering University of Limerick Ireland (e-mail: {edward.thurman, james.riordan, daniel.toal}@ul.ie) Abstract: This paper presents an approach to the real-time adaptive control of multiple high-frequency sonar for Unmanned Underwater Vehicles (UUVs). The central acoustic sensor controller manages the sensors during survey missions, facilitating the operation of co-located, similar-frequency sonar. Multibeam data is processed in real-time to provide a priori bathymetry data to auxiliary acoustic sensors. The automated system is based on the interleaving of the sonar transmission-reception cycles to avoid the saturation of acoustic harmonics, permitting the integration of multiple acoustic sensors operating in parallel. By dynamically adapting the ping rates of the payload sensors, the system optimises the execution of the seabed mapping survey and improves the quality of the resulting data, thereby significantly increasing survey productivity. Keywords: Adaptive control, multibeam sonar, optimisation, seabed mapping, sidescan sonar, Unmanned Underwater Vehicles. 1. INTRODUCTION The subsea environment represents the last major frontier of discovery on Earth. It is envisaged that exploration of the deep-ocean seabed, will present a multitude of potential economic opportunities. Recent interest in the exploration for valuable economic resources, the growing importance of environmental strategies and the mounting pressure to stake territorial claims, has been the main motivation behind the increasing importance of detailed seabed investigations, with vast sums of time and money being spent on producing comprehensive information of the seabed. Recent technological advancements have allowed Unmanned Underwater Vehicles (UUVs) to provide high-resolution survey capabilities for deep-water environments, previously considered uneconomical or technically infeasible (Whitcomb, 2000). Principal UUV survey missions include bathymetric mapping, sidescan imaging, magnetometer surveying and sub-bottom profiling, of which bathymetric mapping and sidescan imaging appear to be of primary commercial interest (Whitcomb, 2000; George, Gee et al., 2002; Desa, Madhan et al., 2006). High-frequency systems (>200kHz) are the desired sensor systems used to obtain high-fidelity maps of the seafloor. Since high-frequencies are quickly attenuated by sea water, it is necessary to deploy these systems as close to the seabed as possible. UUVs, in particular Autonomous Underwater Vehicles (AUVs), are ideal for deep-water surveys, providing the close-proximity capability required for the generation of high-resolution

datasets; leading to detailed datasets and improved data analysis. However, due to acoustic harmonics interfering with adjacent receivers during parallel operation of similar-frequency surveying sonar, the integration of co-located, high-resolution acoustic devices on these deep-water vehicles remains restricted. This paper presents the continuing development of a central acoustic sensor controller for a UUV. The sensor controller processes multibeam data in real-time to provide the a priori bathymetry data to auxiliary acoustic sensors facilitating the concurrent operation of co-located, similar-frequency sonar. The automated system is based on the interleaving of the sonar transmission-reception cycles to avoid the saturation of acoustic harmonics. The controller dynamically adapts the ping rates of the payload acoustic sensors, allowing the optimisation of the execution of a UUV deep-ocean seabed mapping survey. The remainder of the paper is described as follows: Section 2 provides background understanding of the problem and the motivation underlying this paper. Section 3 presents a review of commercial AUV systems and their payload control approaches. Section 4 details the sensors used during system integration. Section 5 describes the implemented system configuration. Section 6 details system testing and Section 7 presents the paper conclusions.

2. BACKGROUND Due to the high attenuation of electromagnetic waves underwater, video and radar are rarely used to map the subsea environment. Instead, acoustic waves are the only practical way to carry energy underwater, with underwater survey operations employing sonar sensors to successfully chart the inshore, coastal, and deep-ocean seafloors. Data gathered by a range of acoustic instruments from multiple sources is a major component of modern marine data acquisition. In particular, multibeam echosounders and sidescan imaging sonar are perceived as being the most effective and efficient seabed mapping instruments available (Kenny, Cato et al., 2003; Shono, Komatsu et al., 2004). The contoured bathymetry maps, generated from multibeam sonar data, and the acoustic reflectivity images, generated from sidescan sonar data, provide complementary information of the seabed (Pouliquen, Zerr et al., 1999; Fanlin, Jingnan et al., 2003; Duxfield, Hughes Clarke et al., 2004; Kirkwood, Caress et al., 2004). Accurately fusing multibeam and sidescan sonar datasets of the same seafloor region allows for an improved level of analysis and the removal of possible dataset ambiguities. Sidescan images enhance the geologic interpretation of bathymetry by providing an acoustic characterisation of the seafloor from which geologic composition can be inferred, while bathymetry information improves the representation of the seafloor relief in sidescan imagery by providing the geometric configuration of the seabed (de Moustier, Lonsdale et al., 1990; Pouliquen, Zerr et al., 1999). The integration of multiple survey sensors onto a UUV platform minimises the relative positional error between features evident in various datasets, as the target region is ensonified by the sensors under the same environmental conditions and geo-referenced by the same navigational data. Simultaneous multi-sonar acquisition also eliminates the need to conduct separate surveys for each instrument, as well as the collection of supporting data required to fully understand the operating environment during each individual survey, thereby significantly reducing the survey duration and consequently the survey costs. The synergies offered by concurrently operating multiple acoustic mapping devices on a UUV underpin the desire for such an operational configuration, facilitating high-resolution surveys of the deep-ocean environment, while enabling the information encoded in one instrument’s dataset to be used to correct artefacts in the other. The reception circuitries of sonar transducers are typically frequency band-limited to prevent harmonic interference from parallel operating acoustic instruments of different frequencies. However, the high-resolution versions of most imaging sonar operate within the same frequency band, with typical working frequencies for high-frequency multibeam echosounders being 200kHz-400kHz, and high-frequency sidescan sonar ranging from 200kHz to 500kHz. While multibeam instruments can be depth gated to filter spurious returns, sidescan sonar records can be severely distorted by sensor cross talk, as they rely on the full temporal trace of the

returned backscatter to construct an intensity image. Consequently, the concurrent operation of multiple similarfrequency sonar is prohibited by the inherent complication of cross-sensor acoustic interference (Ishoy, 2000; Kirkwood, 2007), resulting in a compromise of separating the operating frequency of the payload sonar sufficiently far that such interference is avoided. However, such a compromise results in the deployment of a lower frequency instrument, significantly degrading the quality of the generated datasets, with the result that small-scale features may not be evident, thereby compromising the data interpretation process. 3. COMMERCIAL AUV SYSTEMS As a result of the scientific needs and advantages for multisensor underwater surveys, a number of AUVs have been developed that are capable of performing simultaneous multibeam and sidescan sonar surveys of the deep-ocean environment. One such, well known, UUV is C&C Technology’s HUGIN 3000, a third generation AUV manufactured by Kongsberg Simrad. The AUV is integrated with Simrad EM2000 multibeam echosounder (200kHz), an Edgetech sidescan sonar (120kHz), an Edgetech sub-bottom profiler (2 – 8 kHz), and a CTD sensor. The system is capable of collecting multiple datasets of the seabed at depths of 3000m (Kongsberg-Simrad, 2006). The payload sensors are interfaced and controlled by the HUGIN Payload Processor. The command and control of the sensors are typically preprogrammed and/or operator based (Hagen and Kristensen, 2002) allowing for the simultaneous operation of the sensors integrated into the AUV platform. The Atlas Maridan’s SeaOtter AUV, based on the Maridan 600 AUV (M600), is a well proven vehicle. The system is equipped with a Reson 8125 multibeam echosounder (455kHz), Klein’s 2000 sidescan sonar (100/500kHz) and a GeoAcoustic sub-bottom profiler (ATLAS-Maridan 2005). To enable the synchronised operation of the multiple sensors, the surveyor specifies the repetition rate, delay and duty cycle for each sensor during survey planning. The values are sent over the vehicle network to the Local Trigger Manager (LTM), which generates the signals required for each instrument during deployment (Ishoy, 2000). The Monterey Bay Aquarium Research Institute (MBARI) has developed the DORADO AUV, capable of conducting simultaneous multibeam bathymetry, sidescan sonar and subbottom surveys of an area of interest. The AUV is integrated with Reson’s 7100 multibeam echosounder (200kHz), Edgetech’s 110/410kHz chrip sidescan sonar and an Edgetech 2 – 16kHz chrip sub-bottom profiler. Simultaneous operation of the multiple sensors is managed by the Reson propriety timing algorithm. The multibeam echosounder acts as the master system. The multibeam sonar and other integrated systems are pinged using a fixed 1 pulse per second (PPS) clock, made available by the navigation system (Kirkwood, 2007).

Survey results have shown that the described systems have successfully completed simultaneous multibeam and sidescan sonar surveys of an area of interest (George, Gee et al., 2002; Wernli, 2002; Kirkwood, Caress et al., 2004; Desa, Madhan et al., 2006). However, the integrated systems do not allow for optimal surveys, leading to deficient datasets; the acoustic sensors used are not all of high-frequency and the payload control is pre-programmed and non-adaptable. The close proximity platform offered by a UUV permits the use of high-frequency systems, as the distance between the sensor and the seabed reduces the effect of signal attenuation, and real-time, dynamic control of the survey sensors will enable their optimised use during survey operation, producing the detailed datasets desired.

Fig. 1 System configuration: A multibeam echosounder, sidescan sonar and INS integrated on the MMRRC UUV platform.

4.3 Inertial Navigation System The Ixsea PHINS is used as the Inertial Navigation System (INS), interfaced with multiple navigational aiding sensors, including a RDI 600kHz Doppler Velocity Log (DVL), CDL Microbath depth sensor and a USBL acoustic positioning device. The system outputs the vehicle’s position, orientation (yaw, pitch and roll) and angular and linear velocities. This high accuracy inertial measurement unit is based on Ixsea’s FOG technology coupled with an embedded digital signal processor that runs an advanced Kalman filter (Ixsea, 2005).

Fig. 2 System configuration: A multibeam echosounder and sidescan sonar integrated on the MMRRC UUV platform. 5. SYSTEM DESCRIPTION

4. SYSTEM SENSORS The sensors integrated on the Mobile & Marine Robotics Research Centre (MMRRC) UUV platform provide the highresolution seabed mapping capabilities required to perform accurate and detailed recording and analysis of the seabed (Fig. 1).

Recent advances in computational technology enable realtime processing of multibeam data. The developed system processes multibeam bathymetric data in real-time to enable synchronised multi-sonar operation through the prediction and temporal separation of each of the UUV’s payload sonar’s transmission-reception window.

4.2 Sidescan Sonar

Unlike traditional sensor triggering routines, which operate on fixed timing schedules, the system dynamically adapts the time separation between successive pings according to the range between the sonar and the ensonifed seafloor. The terrain-adaptive timing schedule enables optimal use of the available transmission-reception cycle windows; providing the capability to interleave the pings of multiple acoustic sensors, while still adhering to the resolution requirement of each sensor’s resulting dataset, set down by the International Hydrographic Office (IHO, 1998) in the case of a hydrographic survey.

The sidescan sonar used is the Tritech 325kHz SeaKing sidescan sonar. The system is an ROV/AUV mounted sidescan sonar system operating at a frequency of 325 kHz. The horizontal beamwidth is 1° and the vertical beamwidth is 50°. The system is depth rated to 4,000m (Tritech, 2002).

The integrated acoustic controller system is comprised of the multibeam sonar data acquisition module, the sidescan sonar data acquisition module, and the multi-sonar synchronisation module. The system framework is decomposed as such to ensure that the latency inherent in processing the multibeam

4.1 Multibeam Echosounder The multibeam echosounder employed is the Reson Seabat 7125, operating at a frequency of 400kHz and featuring 512 beams. The system is optimised for measuring within an angular swath of 128°. Product of the transmit and receive beams provide a total effective beam width of 0.5° x 1° (H x V). The system provides bathymetric data at a depth resolution of 3mm (Reson, 2006).

generated data stream does not stall the data acquisition processes. If the system executed sequentially, periodic gaps would occur in data coverage due to the multibeam and sidescan sonar acquisition processes remaining idle while the synchronised multi-sonar timing schedule is being evaluated by the synchronisation module. Consequently, the three modules are executed in parallel, with the three sharedmemory buffers acting as the inter-process communication framework. The multibeam is the master system and provides to the survey controller the raw data that determines the multisensor triggering routine. The developed multibeam sonar data acquisition module reads in the raw data from the Sonar Processing Unit (SPU), filters the data for outliers and extracts each individual beam’s time and angle couplet and deposits the beamformed swath profile into the memory buffer, while auxiliary navigation sensors also deposit concurrently generated high frequency time-stamped Motion Reference Unit (MRU) data into the buffer. Both data streams are combined and a select number of geo-referenced depth points are used to generate and populate a Digital Terrain Model (DTM) of the surveyed region. In calculating the adaptive timing schedule, it is not required to build a full 512 point swath DTM, therefore freeing up the processor and avoiding any bottlenecks. It is also not essential to interrogate the sidescan sonar data to enable the real-time adaptive control of the system; therefore the sidescan sonar data acquisition module accepts the sidescan data and produces a time versus amplitude plot. No calculations are to be performed on the sidescan sonar data during the survey operation.

Fig. 3 Timing diagram for the integrated acoustic controller.

To prevent sensor cross talk, the ping-interleaving timing schedule must be determined in real-time to maintain dynamic terrain-adaptive control of the multi-sensor triggering sequence. This requires synchronous operation of the data acquisition and the multi-sonar synchronisation modules. However, depending on the PC resources available to process the synchronisation module, processing a continuous train of successive pings may create an excessive computational bottleneck. This is avoided by exploiting the fact that as the seafloor topography in general will not change significantly over successive pings, the synchronisation process can thus operate with a lower frequency update rate than the sonar ping rate due to the high degree of redundancy exhibited by successively imaged swaths. The timing schedule of the payload acoustic sensors is determined by processing alternate pings, with the separation between processed pings determined by the capability of the available computational resources to execute without backlog.

The system exploits the fact that, due to the slow forward speed of the survey platform, typically 2 – 4 knots, there occurs a high ping-to-ping coherence between successive multibeam swaths. This permits the duration of the next multibeam transmission-reception window to be accurately predicted. The multibeam transducer is mounted to the fore of the survey platform such that the geometry of the region of the seafloor, to be interrogated by the sidescan sonar, will already have been mapped (Fig. 2), providing the a priori information needed to predict sidescan sonar’s transmissionreception window. The triggering cycle, TC, of the payload sensors is calculated such that the time separation between successive pings is sufficient to prevent the harmonic interference, while still providing datasets that adhere to the IHO resolution requirements. TC is determined by the required accuracy and the velocity of the sensor platform. Within each TC, the transmission-reception windows of the multibeam echosounder, Tmb, and the sidescan sonar, Tss, must be separated by a time, Tgap, that is sufficiently wide enough to minimise interference from scattered spurious returns (Fig. 3). If it is found that TC is unable to sufficiently accommodate Tmb, Tss and Tgap, then the vehicle’s velocity will be reduced, providing a longer TC timing window.

Fig. 4 A simulated image of the concurrent operation of multibeam and sidescan sonar from an AUV platform. 6. SYSTEM TESTING The Mobile & Marine Robotics Research Centre (MMRRC), University of Limerick, Ireland, has, at its disposal, the modern facilities and equipment essential for the development and validation of advanced marine systems. Reson 7k multibeam data and Ixsea PHINS data has been successfully acquired, parsed and geo-referenced during data

simulations, providing diagnosis and initial performance assessment of the timing scheduler and adaptive trigger interleaving protocols. The recently constructed Wet Laboratory, incorporating a 2m x 3m x 1.5m test tank, will aid system integration and initial laboratory based testing, while the MMRRC developed Thrusted Pontoon will provide the platform on which system lake testing will be performed. The Thrusted Pontoon can be operated on the surface as a survey platform, either towed or thrusted by 4 horizontal electric thrusters to allow it to surge, sway and yaw or, using quick release buoyancy modules, the vehicle becomes neutrally buoyant and can be operated as a survey ROV/AUV with control in six degrees of freedom (surge, sway, yaw, heave, pitch and roll). All components integrated on the vehicle, including payload sensors, are depth rated beyond 2,000m. Lake testing will take place at the University of Limerick’s Activity Centre on Lough Derg in early summer 2008. Sea trails are to be carried out onboard the RV Celtic Explorer during a multi-disciplinary research sea cruise, planned for summer 2008. Sea trails will provide authentic survey conditions and datasets, presenting final verification of the developed system.

multibeam sonar, mounted to the fore of the UUV, provides a priori seabed geometry information to auxiliary acoustic imaging and navigation sensors in real-time. The multibeam bathymetry data is processed online and used to construct a Digital Terrain Model (DTM) of the ensonified seabed. The generated multibeam DTM mesh is evaluated to enable the predicted time-of-flight information of payload acoustic sensors to be determined. By dynamically adapting the interping delay, all sonars can be triggered asynchronously according to the range between the sensors and the ensonifed seafloor below. Using the method presented in this paper, multiple acoustic survey sensors are permitted to operate, without interference, concurrently from a UUV platform. Such an operational configuration allows for an improved level of analysis and a significant enhancement in survey productivity.

ACKNOWLEDGEMENTS The authors wish to acknowledge project funding support provided by the Irish Research Council for Science, Engineering and Technology (IRCSET) Embark Initiative and Marine RTDI Measure grant aid under the National Development Plan (NDP) 2000-2006.

7. CONCLUSION Increased interest in the exploration of the deep-water environment has spurred the development and technological advancements of Unmanned Underwater Vehicles (UUVs) that are able to provide the close-proximity capability required for high-resolution surveying of the deep-ocean seafloor. A move towards the UUV integration of multiple acoustic mapping sensors is becoming more and more apparent, with the synergies offered by such an operational configuration outperforming single sensor execution. However, due to acoustic interference suffered by parallel operating receivers, the concurrent operation of multiple similar-frequency acoustic sensors is restricted. As a result of the scientific needs and advantages for multisensor deep-ocean surveys, a number of AUVs have been developed. These systems have successfully integrated and simultaneously operated multiple acoustic survey sensors, including multibeam and sidescan sonar. However, the published systems lack real-time adaptive payload sensor control and optimal sonar configurations, leading to an inferior survey operation. To overcome the effect of acoustic interference and to provide the optimised operation of all sensors during deployment, an adaptive survey sensor controller has been developed to enable each acoustic sensor to follow a dynamic timing schedule such that all sensors can be operated concurrently through the interleaving of their pings in a noninterfering fashion. This paper detailed the ongoing development of a terrain adaptive seabed mapping system whereby a high-resolution

REFERENCES ATLAS-Maridan (2005). ATLAS Maridan SeaOtter Autonomous Underwater Vehicle (AUV), ATLAS Maridan ApS, Horsholm, Denmark. de Moustier, C., P. F. Lonsdale, et al. (1990). Simultaneous operation of the Sea Beam multibeam echo-sounder and the SeaMARC II bathymetric sidescan sonar system. Oceanic Engineering, IEEE Journal of, 15(2), 84-94. Desa, E., R. Madhan, et al. (2006). Potential of autonomous underwater vehicle as new generation ocean data platforms. Current Science, 90(9), 1202-1209. Duxfield, A., J. E. Hughes Clarke, et al. (2004). Combining multiple sensors on a single platform to meet multipurpose nearshore mapping requirements. Canadian Hydrographic Conference, Proceedings of, Ottawa, Canada. Fanlin, Y., L. Jingnan, et al. (2003). Multi-beam Sonar and Side-scan Sonar Image Co-registration and Fusing. Marine Science Bulletin (English Edition), 5(1), 16-23. George, R. A., L. Gee, et al. (2002). High-Resolution AUV Surveys of the Eastern Sigsbee Escarpment. Offshore Technology Conference, Proceedings of, Houston, Texus. Hagen, P. E. and J. Kristensen (2002). The HUGIN AUV "Plug and play" payload system. Oceans '02 MTS/IEEE, Proceedings of, Biloxi, Mississippi.

IHO (1998). IHO Standards for Hydrographic Surveys, Special Publication No. 44, International Hydrographic Bureau, Monaco. Ishoy, A. (2000). How to make survey instruments "AUVfriendly". OCEANS 2000 MTS/IEEE, Proceedings of, Providence, Rhode Island. Ixsea (2005). PHINS User Guide, Ixsea. Kenny, A. J., I. Cato, et al. (2003). An overview of seabedmapping technologies in the context of marine habitat classification. ICES Journal of Marine Science, (60), 411-418. Kirkwood, W. J. (2007). Development of the DORADO Mapping Vehicle for Multibeam, Subbottom, and Sidescan Science Missions. Field Robotics, Journal of, 24(6), 487-495. Kirkwood, W. J., D. W. Caress, et al. (2004). Mapping payload development for MBARI's Dorado-class AUVs. OCEANS '04. MTS/IEEE TECHNO-OCEAN '04, Proceedings of, Kobe, Japan. Kongsberg-Simrad (2006). Hugin 3000 - An Autonomous Underwater Vehicle (AUV) for accurate and effi cient seabed mapping, Kongsberg Simrad A/S, Horten, Norway. Pouliquen, E., B. Zerr, et al. (1999). Seabed segmentation using a combination of high frequency sensors. OCEANS '99 MTS/IEEE, Proceedings of, Seattle, Washington. Reson (2006). SeaBat 7125 High Resolution Multibeam Echosounder System, Reson A/S. Shono, K., T. Komatsu, et al. (2004). Integrated hydroacoustic survey scheme for mapping of sea bottom ecology. OCEANS '04. MTS/IEEE TECHNO-OCEAN '04, Proceedings of, Kobe, Japan. Tritech (2002). Tritech ROV/AUV Mounted Sidescan Sonar Specification, Tritech International Limited. Wernli, R. L. (2002). AUVs - A Technology Whose Time Has Come. Underwater Technology, 2002. Proceedings of the 2002 International Symposium on. Whitcomb, L. L. (2000). Underwater robotics: out of the research laboratory and into the field. IEEE International Conference on Robotics and Automation, 2000, Proceedings of.