Autonomous underwater vehicles for scientific and naval operations

Autonomous underwater vehicles for scientific and naval operations

Annual Reviews in Control 30 (2006) 117–130 www.elsevier.com/locate/arcontrol Autonomous underwater vehicles for scientific and naval operations E. B...

3MB Sizes 13 Downloads 224 Views

Annual Reviews in Control 30 (2006) 117–130 www.elsevier.com/locate/arcontrol

Autonomous underwater vehicles for scientific and naval operations E. Bovio a,*, D. Cecchi b, F. Baralli a b

a NATO Undersea Research Centre, Viale San Bartolomeo 400, 19138 La Spezia, SP, Italy ISME, Interuniversity Centre of Integrated System for Marine Environment, c/o DSEA University of Pisa, Via Diotisalvi 2, 56126 Pisa, Italy

Received 10 June 2005; accepted 6 August 2006 Available online 2 November 2006

Abstract Recognizing the potential of autonomous underwater vehicles for scientific and military applications, in 1997 MIT and the NATO Undersea Research Centre initiated a Joint Research Project (GOATS), for the development of environmentally adaptive robotic technology applicable to mine counter measures (MCM) and rapid environmental assessment (REA) in coastal environments. The August 2001 GOATS Conference marked the end of this 5 years project, but did not mark the end of the work. The Centre initiated in 2002 a new long-term programme to explore and demonstrate the operational benefits and performances of AUV for covert preparation of the battlespace. Recently the work addressed the evaluation of commercial off-the-shelf (COTS) AUV technology for MCM operations in response to terrorist mining of port. The paper summarizes the work performed and refers to the scientific publications derived from the AUV programme at the NATO Undersea Research Centre. # 2006 Published by Elsevier Ltd. Keywords: Autonomous vehicles; Marine systems; Guidance; Navigation; Control; Accuracy

1. Introduction The NATO Undersea Research Centre,1 located in La Spezia, Italy, performs basic and applied research and development to fulfill NATO’s operational requirements in undersea warfare. The results of the Centre’s research, which can be seen at sea in many ships and submarines of the alliance, have contributed to NATO’s military capabilities over the past 41 years. Unique in its international makeup, the Centre functions as the ‘‘hub’’ in a virtual laboratory which brings great synergy to the research process and shortens timelines between research and development (R&D) and military applications. The Centre’s own resources are therefore multiplied by collaboration and Joint Research Projects (JRP). In response to NATO advanced planning that anticipates significant use of autonomous underwater vehicles (AUVs) for mine counter measures (MCM) and rapid environmental assessment (REA), the Centre and the Massachusetts Institute of Technology (MIT) initiated in 1997 a 5-year Joint Research

* Corresponding author. Tel.: +39 0187 527321; fax: +39 0187 527354. E-mail address: [email protected] (E. Bovio). 1 Following the change in NATO command structure, the SACLANT Undersea Research Centre has been recently renamed NATO Undersea Research Centre. 1367-5788/$ – see front matter # 2006 Published by Elsevier Ltd. doi:10.1016/j.arcontrol.2006.08.003

Project, designated generic oceanographic array technology systems (GOATS), for the development of environmentally adaptive AUV technology applicable to MCM and REA in coastal environments. The GOATS JRP grew in membership and scope and it was joined by an international host of collaborators who shared the notion that AUVs were ready to graduate from their role as research objects to a new supporting role for advanced ocean monitoring and maritime military tactics. Between 1997 and 2001 the GOATS JRP explored and expanded the state of the art for networks of robotic ocean observers, supporting new approaches to battlespace preparation and mine hunting. The programme included a sequence of three field experiments with the participation of 14 institutions. The August 2001 GOATS Conference (Bovio, Tyce, & Schmidt, 2001) marked the end of this JRP, but not the end of the work. Building on the success of the GOATS JRP, the Centre initiated in 2002 a new long-term programme called Battlespace Preparation (BP) with AUVs to explore and demonstrate the operational benefits and lperformances of AUVs for military battlespace preparation. Similar to the GOATS series of experiments, the programme organizes multi-national, multi-disciplinary sea trials addressing the utilization of AUVs in coastal waters. The first experiment of the BP series took

118

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

place in May–June 2002 in the Tyrrhenian and Ligurian seas. The results of this experiment have been reported during the Maritime Recognized Environmental Picture (MREP) Conference (Bovio, Coelho, & Tyce, 2003) held in La Spezia in May 2003. This paper summarizes the work performed and refers to the scientific publications derived from the AUV programme. Section 2 reviews the GOATS project. Following that, Section 3 provides an introduction to the Battlespace Preparation project. Section 4 explores the possibility to use AUVs in ports/harbors safety operations. Finally, in Section 5 is shown the importance of control and navigation systems to perform successfull missions. 2. The GOATS JRP The GOATS JRP combined theory and modelling of the 3D environmental acoustics with three experiments (1998, 1999, 2000) involving AUV and sensor technology. The objective of the 1998 sea trial was to use acoustic arrays deployed on the seafloor or mounted on an AUV to characterize the spatial and temporal characteristics of the 3D scattering from seabed targets and the associated reverberation, including the effects of multipath. This effort was aimed at establishing the environmental acoustics foundation for future sonar concepts exploring 3D acoustic signatures for combined detection and classification of proud and buried targets in very shallow water. The GOATS 98 experiment provided a unique data set of 3D scattering and reverberation in shallow water, which has been essential for model validation and identification of features of the 3D acoustic field. In addition, the experiment showed that small and inexpensive AUVs such as the MIT Odyssey can be reliably deployed, operated and recovered in shallow water from a surface vessel. It was also demonstrated that AUVs are an excellent acoustic platform for new sonar concepts for littoral MCM (Schmidt, Maguer, & Bovio, 1998; Schmidt & Bovio, 2000; Moran, 1999). The comprehensive acoustic and environmental datasets acquired during the GOATS 98 experiment have generated several scientific publications. The first papers resulting from the experiment described the physics underlying seabed penetration at sub-critical angles (Maguer, Fox, Schmidt, Pouliquen, & Bovio, 2000; Maguer, Bovio, Fox, & Schmidt, 2000). Several papers deal with the processing of the bistatic synthetic aperture data acquired by the AUV, demonstrating the concept of bistatic SAS autofocusing and imaging (LePage & Schmidt, 2002a; Edwards, Schmidt, & LePage, 2001; Schmidt, Edwards, & LePage, 2000). The JRP has lead to several new developments in regard to modelling of acoustic interaction with the seabed. Specifically a unique modelling capability has been developed, providing a consistent prediction of 3D scattering from seabed roughness and volume inhomogeneities, validated by the GOATS datasets (Veljkovic & Schmidt, 2000; LePage & Schmidt, 2000a,b,c, 2002b). Following the success of the GOATS 98 experiment, the Centre organized a workshop in January 1999 to extend the

scope of the JRP to REA applications. In addition, it was also decided to assess the performance of non-traditional AUV navigation algorithms based on a priori knowledge of the bottom topography. This required a thorough survey of Procchio bay in the Island of Elba, the site of the GOATS 2000 experiment. The GOATS 99 experiment provided a rich data set, collected by traditional instruments (side scan sonar, sub bottom profiler, multibeam echo sounder, underwater video camera, expandable penetrometer) deployed from the NURC coastal boat Manning, that characterizes the bathymetry and the composition of the seafloor of the area and forms the ground truth reference for comparison with data collected subsequently by AUVs. The GOATS 2000 experiment demonstrated the capabilities of AUVs as REA platforms in shallow and very shallow water. The ocean explorer (OEX) vehicle from Florida Atlantic University (FAU), equipped with a colour video camera and the Edgetech dual frequency DF-1000 side scan sonar and the Taipan, from LIRMM France, equipped with the Applied Microsystem CTD were launched from R/V Alliance, to transect the bays to the east of Procchio to acquire side scan sonar data and to measure water mass properties such as current, salinity, density and temperature, for use by the nested oceanographic models. The side scan sonar data were used to generate geo-referenced acoustic images for comparison with ground truth data collected in the same area during previous experiments. The environmental information measured by the AUVs was fused in the Centre Geographic Information Systems (GIS). database. The tiled side scan sonar images were processed with unsupervised segmentation algorithms that demonstrated the capability to distinguish in a quantitative way between different types of seabeds. Fig. 1 shows the Poseidonia oceanica (a seaweed) to sand boundary detected by side scan sonar survey plotted over an aerial picture of the same area (Spina, Bovio, & Canepa, 2001). The video images collected by the OEX were organized in a geographical database using the SeeTrack software. A field of proud and buried targets at the main test site in Biodola Bay was insonified by the topographic parametric sonar (TOPAS). parametric sound source at a variety of incident and aspect angles. The MIT Odyssey AUV sampled 3D reverberation and target echoes obtaining data for validation of numerical models of mono- and bi-static seabed reverberation. A second field of proud targets was imaged at different aspects by the OEX instrumented with the 390 kHz Edgetech side scan sonar and video camera. The experiment demonstrated the potential of high frequency side scan sonar at multiple aspects for classification of proud targets. The results of the GOATS 2000 experiment are reported in Bovio et al. (2001). 3. Battlespace preparation with AUVs Building on the success of the GOATS JRP, the Centre initiated in 2002 a new long-term programme called Battlespace Preparation (BP) with AUVs to explore and demonstrate

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

119

coastal waters for military applications. Each year the experiments are prepared at a planning meeting in January and take place in the Mediterranean sea in May–June. Initial results are discussed in the fall and every second year the Centre organizes a scientific conference to report the findings. The first experiment of the BP series took place in May–June 2002 in the Tyrrhenian and Ligurian seas with three broad objectives: Oceanography, REA, MCM. The results have been reported at the MREP conference (Bovio et al., 2003) and are summarized below. 3.1. Oceanography

Fig. 1. Boundary between sand and Poseidonia oceanica (aerial image and unsupervised seafloor segmentation).

the operational benefits and performances of AUVs for battlespace preparation. Similar to the GOATS series of experiments, the programme organizes multi-national, multidisciplinary sea trials addressing the utilization of AUVs in

Ocean forecasting is essential for effective and efficient use of AUVs in the littoral environment. The first part of the BP02 sea trial was dedicated to carry out and quantitatively evaluate a multiscale real-time forecasting experiment in support of long range AUV missions. A two-way nested Harvard ocean prediction system (HOPS). model was run at Harvard University to predict oceanographic parameters in local (such as around Elba) and far field (eastern Ligurian sea) regions. Adaptive sampling patterns have been determined on short notice based on forecast results for both R/V Alliance and the Remus AUV. The AUV run missions up to 8 h at a speed of 3–4 kts. As shown in Fig. 2, acoustic communication was maintained at rendez-vous points by the Woods Hole Oceanographic Institution (WHOI). Utility Modem deployed from Alliance. The acoustic link allowed real-time data retrieval at a reduced rate and on-line programming of the vehicle mission. Remus executed CTD sampling in yo–yo mode and ADCP sampling in a depth range of 0–100 m. Vehicle navigation was

Fig. 2. Oceanographic experiment in deep water. The Remus track is shown in grey (red in the online version of this paper) and the circular marks (green in the online version) indicate the locations where control information, vehicle status and oceanographic data were exchanged acoustically between vehicle and R/V Alliance. Typical modem range was 800–1500 m, depending on vehicle depth.

120

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

accomplished by dead reckoning, with GPS updates when surfaced. Heading information was obtained with a magnetic compass. Velocity information was obtained from the ADCP when in bottom lock range. When the vehicle was not in range of the bottom, velocity was based on an estimated speed derived from its propeller’s rotation rate. The characteristics of the Remus vehicles used in BP02 are shown in Fig. 3. Oceanographic data acquired by Remus were fused with with Expandable bathythermography (XBTs). and CTDs collected by R/V Alliance and available meteorological information. The fused data set was transmitted via Internet to the modelling team. The oceanographic team onboard Alliance coordinated the assimilation of oceanographic data collected by the various sources. The modelling code was run at Harvard. The model

output was made available via Internet to R/VAlliance to perform adaptive sampling and optimize long range AUV missions. 3.2. REA The REA experiment demonstrated the capabilities of AUVs as REA platforms in shallow water. The information collected included acoustic and video images, bathymetry and water mass properties such as current, salinity, density and temperature. The AUVs were launched from R/V Alliance, and surveyed the bays of Levanto, Bonassola and Framura, Italy, to acquire environmental information in preparation for the MCMFORMED’s percentage clearance (PC) trial which took place in February 2003.

Fig. 3. The Remus AUVs used in BP02. Remus#1 has been used for Oceanographic, REA, MCM, and communication studies. Remus #3 has been used for target ID, Remus #2 has been kept as hot spare.

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

The OEX operated the dual frequency (150/600 kHz) Marine Sonic and the video camera, the Remus operated the 900 kHz Marine Sonic and the DIDSON acoustic lens. All vehicles performed several missions a day, controlled from Alliance by acoustic and/or radio frequency communication. The sonar and video images were downloaded at the end of each mission and stored in the Centre GIS database. Unsupervised segmentation software developed at the Centre divided the seafloor into areas of similar characteristic (Spina & Grasso, 2003). Objects with dimensions similar to a mine were automatically extracted and marked on the GIS map. All data contributed to the production of seabed classification maps according to ATP 24 standards that were provided to the NATO mine hunters participating to the February 03 PC trial with the objective to measure the value of a priori environmental information in planning and conducting a PC trial. The Groupe d’Etude Sous-Marine de l’Atlantique (GESMA) ship R/V Thetis equipped with the interferometric Klein 5400 side scan sonar conducted an independent survey of the area. The ship acquired co-registered imagery and bathymetry that has been compared with that acquired by Manning equipped with DF-1000 side scan and EM-3000 multi beam sonar during previous experiments. Work is in progress to assess the bathymetric capability of the sonar as an alternative to a multibeam (Zerr, Bovio, & Spina, 2003).

121

3.3. MCM The Italian Navy laid a field of exercise targets in two lanes selected to include a variety of different bottom types (rocks, P. oceanica, sand, mud) and portions of highly cluttered areas (wrecks, man made objects laid to protect cables and sewage pipes). In order to compare the performance of the experimental systems with that of the Italian Navy mine hunter Numana, none of the teams knew the position of the targets. Numana surveyed Framura and Levanto lanes and performed visual identification with the Pluto ROV. Remus operated only in Framura due to weather. The AUV surveyed the area with ‘‘lawn mower’’ tracks using the 900 kHz Marine Sonic side scan sonar. Sonar images were downloaded from the vehicle upon return and analysed by the WHOI team using the Marine Sonic software. Interesting objects were successively identified with the Remus vehicle equipped with the DIDSON acoustic camera that provides high resolution video images and with the Centre vehicle ocean explorer equipped with sonar and video. The OEX that was deployed for the first time is shown in Fig. 4. Thetis surveyed both areas with redundant tracks providing multiple aspect of targets. The ship did not cover the north-west corner of the Framura area, which was too shallow for safe towing of the Klein. Sonar images were received and processed in real time. Targets were detected and classified using multiple views of the objects.

Fig. 4. The ocean explorer (OEX) is designed to accommodate various sonar, camera, and oceanographic systems in modular sections. The length and weight of the vehicle depend on the payload configuration.

122

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

Fig. 5. Results of PC trial (Framura). In the surveyed area, there were seven objects to detect. Due to water depth limitations, Thetis did not survey the top corner of the lane where three targets were laid. Two objects were concealed by vegetation or masked by rocks and cannot be detected.

Figs. 5 and 6 show the performance of the three systems. Numana detected, classified and identified three targets in Framura and two in Levanto. Two objects in Framura and one in Levanto were undetectable because they were concealed by the vegetation or masked by rocks. Remus detected, classified and identified five targets in Framura and did not operate in Levanto. Thetis detected and

classified three targets in Framura and two in Levanto. Due to water depth limitations, Thetis did not survey the northern corner of Framura area where three objects were present. Remus and OEX showed great potential for MCM operations and the Klein demonstrated the good performance of a state-ofthe-art commercial sonar.

Fig. 6. Results of PC trial (Levanto). In this area, three exercise targets were laid. One target was concealed by vegetation or masked by rocks and was undetectable.

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

123

4. Ports/harbors safety Recently, responding to a request by SACLANT, a new project was started to evaluate the applicability of commercial off-the-shelf (COTS) AUV technology to MCM operations in response to terrorist mining of ports. Four demonstrations have been successfully conducted in La Spezia and Stranraer and one more is planned in Rotterdam. Current AUV technology is sufficiently mature to complement existing MCM assets (mine hunters and explosive ordinance disposal (EOD). divers) and improve their performances. Of particular interest is the capability to ship overnight small AUVs anywhere a crisis might occur and to place the appropriate sensors (sonar, optical, magnetic) in close proximity of mines without risking human lives. The limited cost of COTS AUV (compared with traditional MCM assets) allows to deploy fleets of specialized vehicles to achieve large area coverage. During exercise Northern Light 03, the OEX launched from Stranraer pier, and the Remus, operated by Royal Navy Fleet Diving Unit 2 (FDU2), surveyed the final part of Loch Ryan, where four exercise targets had been deployed. The AUVs covered the 3000 m by 300 m area with three detection missions followed by a number of classification missions. Remus navigated in a network of acoustic transponders and imaged the seafloor with 900 kHz Marine Sonic side scan sonar. OEX navigated with a GPS tow float and imaged the seafloor with 600 kHz Marine Sonic side scan sonar and with a digital video camera. The side scan sonar data acquired by the vehicles were analysed to determine the nature of the contacts and to provide their location to FDU2 divers for identification and disposal. The purpose of the sea trial in La Spezia harbor (March 2004) was to demonstrate the effectiveness of AUVs and ROVs in support to mine hunters and EOD divers, to counter terrorist mining of an harbor. The exercise measured the effectiveness of AUVs in detecting, classifying and correctly geo-referencing targets for further prosecution by EOD divers. Remus, provided by Hydroid, was configured with Marine Sonic 900 kHz side scan sonar. The vehicle, launched and retrieved from a rib boat, navigated within a network of acoustic long base line (LBL) transponders deployed by Leonardo (see Fig. 7). The Remus covered the assigned channel and the anchorage areas with orthogonal lines in order to obtain multiple aspect insonification of all targets. Line spacings were designed to ensure full bottom coverage of the side scan sonar. All targets were detected and properly localized. At present Remus communicates in real time via acoustic modem only status information. In the near future the vehicle will be able to transmit side scan sonar images of targets to a communication buoy that will relay the information to a ship or a shore installation. 5. Control and navigation: basic tools for successful missions The success of the AUV’s missions showed in previous sections depends strongly on the good performance of the control and navigation systems. Seafloor classification and

Fig. 7. Remus vehicle communicating with the LBL.

target detection are only possible when side scan sonar acquires good quality images. This requires that the vehicles are able to follow pre-progammed paths, maintain a constant heading, speed altitude or depth especially in very shallow water. The AUV position during the missions is required with the highest precision, in order to geo-reference all the detected targets with minimal error. 5.1. Control system Typical requirements for the control system in AUV missions are:  course keeping;  constant depth;  constant altitude;  noise and disturbances rejection. The first three items are satisfied when the vehicle heading, pitch, depth and altitude control loops are stable and well tuned. Moreover the controller should to be insensitive to sensors noise (high frequency noise) and system parameters variations and robust to external disturbances. Example of external disturbances is the presence of P. oceanica on the sea bottom that causes wrong altitude measurements and affects the altitude controller’s behavior. The problem is more evident when the AUV passes the border between clean seafloor and P. oceanica navigating in constant altitude mode. In this case sudden variations in altitude are measured and the controller reacts rapidly running the risk to touch the bottom depending on vehicle length and the current vehicle altitude. Considering a torpedo shaped AUV having the bottom rudder in the longitudinal plane of the tail section, the vehicle rotation in the longitudinal plane causes the bottom rudder to be the AUV part closest to the seafloor. There are several methods for designing the autopilot. It is possible to identify two large classes: model-based or modelindependent control systems (Fossen, 2002). The first approach requires the knowledge of, at least, the reduced vehicle model and allows for the controller design, modification and first

124

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

approximation tuning, by simulation. A drawback of this approach is that vehicle description could change with different payloads, so more vehicle models are needed and, perhaps, different tuning of the controllers is necessary when using different payloads. It is often a hard task to obtain an accurate vehicle model with tests and identification work. Modelindependent controllers are, in general, not so simple to tune but have the advantage of being robust to payloads changes. The autopilots design can be approached in various methods. A standard decoupled PID controller is proposed in Jalving (1994), multivariable sliding mode controllers are presented in Healey and Lienard (1993), optimal controllers for AUVs are shown, for instance, in Juul, McDermott, Nelson, Barnett, and Williams (1994), and Feng and Allen (2002)(LQG/LTR and H 1 methods); a self-tuning autopilot is proposed in Goheen and Jeffreys (1990); fuzzy logic based controllers can be found in Craven, Sutton, and Kwiesielewicz (1998), and Song and Smith (2000). The control system of the OEX-C AUV available at the NATO Undersea Research Centre belongs to the modelindependent class. It is a fuzzy sliding mode controller (FSMC) (Song & Smith, 2000). The vehicle dynamic was estimated through open loop sea tests and, consequently, a nonlinear controller robust to system parameters variations, has been designed. The switching curve of the sliding mode controller can be obtained by at sea measurements and then approximated by fuzzy logic. When the controllers are well tuned, the vehicle is able to track desired paths with minimum oscillations in heading (highly desirable for side scan sonar acquisitions), in pitch and in depth/altitude. The consequences of ineffective controllers can be seen in Figs. 8 and 9 that show oscillations in heading and pitch. The problem is more evident looking at side scan sonar images (Figs. 10 and 11): in Fig. 10 two targets are visible (highlighted with circles) and the image is good quality; in Fig. 11 only one of the targets is recognizable and the quality of the image is less than the previous one. The analysis of several generic mission

Fig. 8. Example of large vehicle oscillations in heading (AUV speed: 3 kts).

Fig. 9. Example of large vehicle oscillations in pitch (AUV speed: 3 kts).

data and the execution of dedicated open loop tests, have allowed to the identification of the vehicle’s maneuverability and a proper tuning of the autopilot. The tuning work was focused on finding a particular trade-off between performance and the stress placed on the actuators. When the controllers are working around the desired set points, the fins change their deflection continuously inside a certain band, approximately centered around the zero. Choosing different values for the controllers parameters, it is possible to vary the width of the mentioned above band. Excessive fin movements often do not improve vehicle’s performances and, at the same time, could damage mechanical parts of the actuators. Several tests have been performed in an area where the presence of weak sea currents, has allowed neglecting the sideslip angle, so that the vehicle’s heading was coincident with

Fig. 10. View of target (good control).

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

125

Fig. 13. Detail of the heading angle in a constant heading test (AUV speed: 3 kts).

Fig. 11. View of target (vehicle oscillating).

the vehicle motion direction. Fig. 12 shows the heading angle along a ‘U’-shaped (path consisting of two parallel legs with a perpendicular one in between) trajectory mission in a constant heading test. Fig. 13 shows the heading angle along a straight portion of the test. Note that the desired value is followed with small oscillations of 1 around the set point. The basic autopilot tuning process has been made assuming that longitudinal and steering plane motions were decoupled. The assumption is true for low values of the pitch angle. Considering the longitudinal plane, the results obtained tuning the depth and altitude autopilots are shown in Figs. 14 and 15. In the test area, the seafloor is sandy and quite regular. The sea was calm so that surface effects were negligible even though the depth was in the range 6–10 m. The constant altitude autopilot tuning has been validated using a long test (about 1 h) in open sea. As Fig. 16

Fig. 14. Depth, altitude and bathymetry in a constant depth test.

Fig. 12. Heading angle in a constant heading test (AUV speed: 3 kts).

Fig. 15. Depth, altitude and bathymetry in a constant altitude test.

126

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

Fig. 16. Depth, altitude and bathymetry in a constant altitude test.

indicates, the desired altitude is tracked with an error of about  25 cm. An example of a survey mission is reported in Fig. 17. The vehicle has followed properly the desired trajectory in all the legs except the last (cross leg). The difference was the guidance system used: in the parallel legs the AUV was forced to track the line joining two subsequent waypoints within a desired corridor centered around this line. The last leg was run with a line-of-sight guidance system without trajectory constraints. The trajectory is curved because of a presence of sea current. 5.2. Navigation System Navigation accuracy is a key factor in the use of the AUV. As part of the global guidance, navigation and control (GNC) system (Fig. 18), the choice of the navigation subsystem should be based on the global performances achieved by all the other vehicle subsystems. Like the control system, navigation

Fig. 18. Block diagram of a typical GNC system.

accuracy has a significant influence on the payload sensor performance, because the vehicle position is used to georeference the collected data as well as the attitude and velocity can be used to process and compensate sensor data. For multi-purpose AUVs the best approach to the navigation problem is the use of an aided inertial navigation system (AINS), taking advantage of the reliability and high bandwidth of an inertial measurement unit (IMU), using an external (aiding) sensor to reduce its typical low-frequency errors. Navigation sensor data are fused using an error state Kalman filter rather than estimating the desired quantities (velocity, position and attitude), this filter estimates errors in measures and computed quantities. Fig. 19 shows the typical scheme of an AINS, where position, velocity and attitude are calculated from IMU data (navigation equation) and then compensated for the errors estimated by the Kalman filter, comparing them with aiding sensor measurements. The basic sensor set for an AINS system includes:    

inertial measurement unit (IMU); speed sensor, typically a Doppler velocity log (DVL); depth/pressure sensor; independent position sensor, typically a GPS for initialization and sporadic error resets.

The accuracy of the speed sensor and the availability of frequent position updates together with IMU characteristics are key factors on the overall system performances.

Fig. 17. AUV trajectory in a survey-type test.

5.2.1. Navigation sensors and mission design The use of an inertial navigation system into an autonomous underwater vehicle has impacts on most of the features and

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

127

Fig. 19. Aided inertial navigation system structure.

specifications of the system. Not only the navigation accuracy is improved, the availability of an estimate of the navigation error extends the possibility of merging the navigation solutions together with the payload data. The knowledge of the system and of the navigation sensor characteristics is important to design optimal missions, from the navigation point of view, accomplishing the tasks required. It is important to note that extremely high accuracy, though desirable, is not required by all the possible missions. The navigation accuracy required for an AUV collecting oceanographic data (CTD) would be much smaller than for an AUV used for MCM or REA missions. Another important aspect of the of the INS is that the total navigation accuracy is strictly related to the mission profile (Jalving et al., 2003). This is because some navigation errors are more observable in certain conditions (during turns for example) and their effects are reduced/canceled in other situations. The effect of the acrosstrack errors is canceled when the vehicle goes back and forth surveying a specific area. For these reasons the navigation error of a certain INS is not a meaningful parameters by itself, if the type of the mission where the system operated is not provided. For an INS with a standard set of aiding sensors the navigation accuracy is dramatically influenced by the availability of the velocity of the vehicle with respect to the seafloor, as measured by the DVL sensor. The maximum distance from

the seafloor to get this measurement is a function of the frequency of the pulse sent by the sensor so that if the mission is performed in an area with deep water, the vehicle will operate without DVL measurements for most of the time. This is a common problem for AUVs and it can be circumvented following different strategies: (1) While the DVL measurements are not available the INS is basically running as pure inertial navigation system. In this case the navigation accuracy is governed by the IMU precision, using a better IMU the system will reduce its error. This approach has several drawbacks: the costs for an upgrade of the IMU can be extremely relevant; moreover, the error will be reduced but it will keep growing in time. (2) Using an additional underwater positioning system it is possible for the INS to receive measurement updates while the vehicle is operating at high altitude. Position and heading errors become observable for the Kalman filter with position updates, producing good navigation accuracy and error estimation. The use of this type of systems requires the deployment of a field of acoustic transponders (LBL) or the use of an acoustic positioning system from a mother ship (USBL, HiPaP). (3) The altitude range for the DVL can be extended by using a model with lower frequency however the maximum range is still limited to 100 m and it also causes a loss of accuracy of

128

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

Fig. 20. Deep water Yo–Yo mission.

the velocity measurements. A new type of velocity sensor called correlation velocity log (CVL) is going to be used in underwater systems like the AUV. The CVL sensor should be capable of providing velocity measurements at an altitude up 500 m with a good accuracy even at low speed. The profile of a typical REA mission is shown in Fig. 20. The OEX-C AUV was tasked to collect environmental data while approaching the beach (shallow water area) from a distant position in a deep water area (  150 m). The vertical profile has an ‘Yo–Yo’ shape to optimize the acquisition of salinity and temperature data. The vehicle was running with the deadreckoning system based on a magnetic compass. A prototype of the inertial navigation system was also installed and running in background logging all the navigation data. The REA mission described by Fig. 20 is critical from the navigation point of view, because it starts in an area where the velocity of the vehicle respect to the sea bottom cannot be measured by standard DVL sensors. This results in a less accurate initial alignment and the INS runs in pure inertial mode during the diving phase. This problem was addressed in post-processing by using the water-reference velocity from the DVL during the periods when ground-referenced velocity is unavailable. The error model for the DVL was changed according to the type of measurements provided by it. Despite of the larger error affecting the water-referenced velocity

(currents, noise, etc.), the system achieved a good accuracy, especially when compared to the dead-reckoning solution. The availability of further aiding sensors: CVL and transponders or aiding systems: terrain navigation (Hagen and Hagen, 2001), concurrent mapping and localization (Ruiz, de Raucourt, Petillot, & Lane, 2004) based on advanced techniques extends the capabilities of the AUVs to perform a wider range of missions, however the choice of the sensor set should always be a trade-off between the desired accuracy, the system complexity and the mission requirements (covertness, environmental factors, and mission duration). 6. Conclusions Autonomous underwater vehicles (AUV) have reached sufficient maturity to be considered for military applications. After the successful completion of the GOATS Joint Research Programme (Bovio et al., 2001), the Centre has initiated a longterm programme to explore and demonstrate the operational benefits and limitations of AUVs for battlespace preparation. This activity is based on multi-national, multi-disciplinary sea trials to evaluate the performance of commercially available AUVs in comparison with current military assets for MCM and REA applications. In addition, similar to the GOATS series of experiments, the Centre studies with research partners the key technologies required for successful AUV deployment. The

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

experiments carried out during the first test at sea, during May– June 2002 in the Tyrrhenian and Ligurian seas, have been highly successful and demonstrated the clear advantage of using autonomous vehicles for a variety of REA and MCM missions (Bovio et al., 2003). In particular COTS AUV technology has been evaluated in MCM operations against terrorist mining of ports. Four experiments demonstrated that current AUV technology is sufficiently mature to complement existing MCM assets (mine hunters and EOD divers) and improve their performance. Of particular interest is the capability to ship overnight small AUVs anywhere a crisis might occur and to place the appropriate sensors (sonar, optical, magnetic) in close proximity of mines without risking human lives. The work will continue in the following years in cooperation with the research partners and NATO navies to reach the final goal of assessing the value of AUV networks for operational use. References Bovio, E., Coelho, E., & Tyce, R. (Eds.). (2003). Maritime recognized environmental picture (MREP) conference proceedings, no. CP-47. Bovio, E., Tyce, R., & Schmidt, H. (Eds.). (2001). Autonomous underwater vehicle and ocean modelling networks: GOATS 2000 conference proceedings, no. CP-46. Craven, P., Sutton, R., & Kwiesielewicz, M. (1998). Neurofuzzy control of a nonlinear multivariable system. UKACC international conference on control, Vol. 1. Edwards, J., Schmidt, H., & LePage, K. (2001). Bistatic synthetic aperture target detection and imaging with an AUV. IEEE Journal of Oceanic Engineering, 26, 690–699. Feng, Z., & Allen, R. (2002). H 1 autopilot design for an autonomous underwater vehicle. In Proceedings of the 2002 international conference on control applications, Vol. 1. Fossen, T. (2002). Marine control systems. Trondheim, Norway: Marine Cybernetics. Goheen, K., & Jeffreys, E. (1990). Multivariable self-tuning autopilots for autonomous and remotely operated underwater vehicles. IEEE Journal of Oceanic Engineering, 15(3), 144–151. Hagen, O. K., & Hagen, P. E. (2001). Terrain referenced integrated navigation systems for underwater vehicles. In E. Bovio, R. Tyce, & H. Schmidt (Eds.), Autonomous underwater vehicle and ocean modelling networks: GOATS 2000 conference proceedings, no. CP-46. Healey, A., & Lienard, D. (1993). Multivariable sliding mode control for autonomous diving and steering unmanned underwater vehicles. IEEE Journal of Oceanic Engineering, 18(3), 327–339. Jalving, B. (1994). The NDRE–AUV flight control system. IEEE Journal of Oceanic Engineering, 19(4), 497–501. Jalving, B., Gade, K., & Bovio, E. (2003). Integrated inertial navigation systems for AUVs for REA applications. In E. Bovio, E. Coelho, & R. Tyce (Eds.), Maritime recognized environmental picture (MREP) conference proceedings, no. CP-47. Juul, D., McDermott, M., Nelson, E., Barnett, D., & Williams, G. (1994). Submersible control using the linear quadratic gaussian with loop transfer recovery method. In Proceedings of the 1994 symposium on autonomous underwater vehicle technology. LePage, K., & Schmidt, H. (2000a). Laterally monostatic backscattering from 3D distributions of sediment inhomogeneities. In M. Zakharia, P. Chevret, & P. Dubail (Eds.), In Proceedings of the 5th European conference on underwater acoustics. LePage, K. & Schmidt, H. (2000b). Spectral integral representations of multistatic scattering from sediment volume inhomogeneities. In Proceedings of the 140th ASA meeting/NOISE-CON 2000. Abstract published in Journal of the Acoustical Society of America, 108, 2564.

129

LePage, K., & Schmidt, H. (2000c). Spectral integral representations of volume scattering in sediments in layered waveguides. Journal of the Acoustical Society of America, 108, 1557–1567. LePage, K., & Schmidt, H. (2002a). Bistatic synthetic aperture imaging of proud and buried targets using an AUV. IEEE Journal of Oceanic Engineering, 27, 471–483. LePage, K., & Schmidt, H. (2002b). Spectral integral representations of monostatic backscattering from three-dimensional distributions of sediment volume inhomogeneities. Journal of the Acoustical Society of America, 113, 789–799. Maguer, A., Bovio, E., Fox, W., & Schmidt, H. (2000). In situ estimation of sediment sound speed and critical angle. Journal of the Acoustical Society of America, 108, 987–996. Maguer, A., Fox, W., Schmidt, H., Pouliquen, E., & Bovio, E. (2000). Mechanisms for subcritical penetration into a sandy bottom: Experimental and modeling results. Journal of the Acoustical Society of America, 107, 1215–1225. Moran, B. (1999). GOATS 98 AUV network sonar concepts for shallow water mine countermeasures. In Proceedings of the 11th international symposium on unmanned untethered submersible technology. Ruiz, I. T., de Raucourt, S., Petillot, Y., & Lane, D. M. (2004). Concurrent mapping and localization using side-scan sonar. IEEE Journal of Oceanic Engineering, 29(2), 442–456. Schmidt, H., & Bovio, E. (2000). Underwater vehicle networks for acoustic and oceanographic measurements in the littoral ocean. MCMC2000: 5th IFAC conference on maneuvering and control of marine crafts. Schmidt, H., Edwards, J., & LePage, K. (2000). Bistatic synthetic aperture sonar concept for MCM AUV networks. International workshop on sensors and sensing technology for autonomous ocean systems. Schmidt, H., Maguer, A., & Bovio, E. (1998). Generic oceanographic array technologies (GOATS) 98—bistatic acoustic scattering measurements using an autonomous underwater vehicle. Technical report SR-302, NATO Undersea Research Centre, La Spezia, Italy. Song, F., & Smith, S. (2000). Design of sliding mode fuzzy controllers for an autonomous underwater vehicle without system model. OCEANS 2000 MTS/IEEE conference and exhibition, Vol. 2. Spina, F., Bovio, E., & Canepa, G. (2001). Seafloor classification for MCM with AUV mounted sensors. In E. Bovio, R. Tyce, & H. Schmidt (Eds.), Autonomous underwater vehicle and ocean modelling networks: GOATS 2000 conference proceedings. Spina, F., & Grasso, R. (2003). Unsupervised sea bottom classification from side-scan sonar images using multi-resolution transform features. Technical report SR-372, NATO Undersea Research Centre, La Spezia, Italy. Veljkovic, I., & Schmidt, H. (2000). Experimental validation of numerical models of 3D target scattering and reverberation in very shallow water. In Proceedings of the 140th ASA meeting/NOISE-CON 2000. Abstract published in Journal of the Acoustical Society of America, 108, 2485. Zerr, B., Bovio, E., & Spina, F. (2003). Bathymetric sidescan sonar for covert and accurate MCM REA. In Proceedings of the UDT 2003 conference. Edoardo Bovio, after graduation in electrical engineering at the University of Genoa in 1976, joined the SHAPE Technical Centre in The Hague where he worked in communications and radar. He then worked for Hewlett Packard specializing in engineering applications of signal processing techniques. Since 1980 he is with the NATO Undersea Research Centre (NURC) where he performed several scientific and managerial tasks related to the development of active sonar systems. Since 1997 he is responsible for the Centre’s AUV program. This work evolved from basic research on vehicle technology in collaboration with many research institution and NATO Navies, to focused studies that quantify the benefits of AUVs for military applications, with special emphasis on harbor protection. Bovio has published more than 20 papers on the applications of autonomous underwater vehicles. Daniele Cecchi received the Master of Science in Electronic Engineering and the Ph.D. degree from the University of Pisa in 2000 and 2004, respectively. Since 2004 he was with the Department of Electrical Systems and Automation of the University of Pisa as contractor and post-doctoral fellow. His research

130

E. Bovio et al. / Annual Reviews in Control 30 (2006) 117–130

activity is about guidance, control and simulation of autonomous underwater vehicles. Francsco Baralli received the Master of Science in Computer Engineering and the Ph.D. degree from the University of Pisa in 1999 and 2004, respectively.

Since 2003 he was with the NATO Undersea Research Centre (NURC) in La Spezia as scientific consultant until 2005, then as staff member of the Expeditionary Mine Countermeasure and Port Protection Department. His research activity is about inertial navigation and guidance of autonomous underwater vehicles.