Buoyancy Control for an Autonomous Underwater Vehicle

Buoyancy Control for an Autonomous Underwater Vehicle

IFAC Copyright «,'l IFAC Guidance and Control of Underwater Vehicles, Wales, UK, 2Otl3 ~ Publications www.elsevier.com/locate/ifac BUOYANCY CONTRO...

652KB Sizes 29 Downloads 156 Views


Copyright «,'l IFAC Guidance and Control of Underwater Vehicles, Wales, UK, 2Otl3


Publications www.elsevier.com/locate/ifac


Trevor Love, Daniel Toal and Colin Flanagan

University ofLimerick, Limerick, Republic ofIreland.

Abstract: This paper describes an active buoyancy control system for an autonomous underwater vehicle (AUV). The AUV 'Tethra" is a shallow-water craft designed for a target application of underwater video surveying and as a general AUV test-bed. The craft is controlled using an enhanced version of Brook's behaviour based Subsumption Architecture. Mission scenarios that might require active buoyancy control have been identified. Active buoyancy control is realised using ballast tanks and compressed air. A secondary emergency controller and an Emergency Buoyancy System (EBS) are also presented. Buoyancy control behaviours have been developed and are incorporated in the craft's autonomous control algorithms. Copyright © 2003 IFAC Keywords: Autonomous Vehicles, Autonomous Control




The addition of an active buoyancy control feature on an underwater craft adds weight and expense to a vehicle and increases control complexity. Hence the question of whether or not to include active buoyancy must be carefully considered. In the case of Tethra, mission scenarios for the target application were scrutinised. Three scenarios emerged that would make the inclusion of an active buoyancy control feature more favourable:

In late 2000 work began at the University of Limerick on developing an Autonomous Underwater The target Vehicle (AUV) named "Tethra". application of this AUV is underwater filming, for example to aid maintenance inspections of submerged structures such as ship hulls or fish-farm cages (Toal, et aI., 200 I). During the initial development stages of the AUV it was decided that the role of the craft should be expanded to that of a sea-going mission capable test-bed. This was done to facilitate testing of developed Artificial Intelligence (AI) control algorithms through practical application, as it was deemed that simulations would not be sufficient in imitating the unstructured and highly dynamic ocean environment. To aid the craft's effectiveness in achieving the main target application and to make the craft more adaptable as an effective test-bed, an active buoyancy control system was incorporated in the design. This paper discusses this system, details its practical implementation and describes the on-going development of its autonomous control.




Neutral/positive craft buoyancy drift due to changes in pressure, temperature and water salinity. Changes in water salinity apply particularly to Tethra, as the main target application of the craft would involve missions in estuaries where salinity levels can change substantially during tides and at different depths. For operations where the use of vertical thrusters is unfeasible or interferes with mission application. During filming close to a muddy seafloor the use of vertical thrusters may cause water clouding and hence be unpractical. In this scenario active buoyancy control would be used to control depth.


An emergency/fault condition. Such a condition may require a craft buoyancy change for controlled ascent.

Tethra's role as a test-bed also makes the inclusion of active buoyancy control more viable. Other conditions where active buoyancy control might be required are: • •


Buoyancy changes resulting from sample collection or payload delivery. Assistance of vertical thrusters during large changes in depth to lessen consumption of precious battery power.

Variable Ballast Systems

Most underwater vehicles use some sort of adjustable ballast to control buoyancy. In the case of smaller craft buoyancy control is normally realised using a passive fixed ballast system that may be adjusted prior to mission to make the craft slightly positively buoyant. In these cases it is assumed there will be no significant change in the buoyancy of the craft during mission and depth control is realised by means of control surfaces and vectored propulsion. Oftentimes such craft have the capability to release some fixed ballast in case of emergency, e.g. the "Explorer" AUV by International Submarine Engineering Ltd. of Canada, (Ferguson and Pope, 2000). Should an underwater vehicle require adjusting its buoyancy during mission it will need some sort of active variable ballast system to do so. An active variable ballast system normally uses water as the ballast medium. By adjusting a ratio of water and air within tanks incorporated in the craft, the buoyancy of the vehicle can be adjusted accordingly. There are two main methods of implementing such a system: water may be removed from tanks by means of a pump or piston, or water may be forced out by means of a pressure head induced from a compressed air supply.

Other sub-sea technologies also use compressed air to control buoyancy: "NO MAD", a sub-sea data collection robot, uses a compressed air method as exclusive means of depth control (Lu, 1997). As Tethra is a low-cost shallow-water craft it was decided that a pressurised air method of implementing active variable ballast would be used to realise active buoyancy control.

3 THE "TETHRA" AUV Tethra is a multi-purpose shallow-water craft of open frame design, as shown in figure 1. It is designed around the principle of an inverted tetrahedron, with the top of the craft incorporating most of the fixed buoyancy in the form of four tubes. Two tubes are solid foam filled and provide the core buoyancy for the craft. The remaining two tubes are flooded and contain foam floats on sliders. The amount of floats and positioning of the floats can be varied and thus provide a passive method of controlling stability and buoyancy. A variable ballast system uses four tanks located centrally in the main frame. The centre of the craft contains twin steel cylindrical hulls. These hulls contain the main power supply, on board computer and the craft control circuitry. Four trolling motors, manufactured by MinnKota, act as thrusters to provide propulsion. Two are mounted vertically, two horizontally. The open frame provides plenty of space for tool racks, sensor equipment, sonar, lighting, cameras and any other specialised equipment necessary for mission realisation. The electronic control hardware is designed in the form of a distributed control hardware, in which a number of separate modules are being used to perform complex tasks. The system is described extensively in (Molnar, et aI., 2002). The control software and autonomy algorithms are based on

Pumped ballast is normally used in larger deep-water vehicles such as MBARl's\ "Tiburon" Remotely Operated Vehicle (ROV) (Kirkwood and Steele, 1994) and ARPA's2 Unmanned Undersea Vehicle (UUV) (DeBitetto, 1995). Such vehicles cannot use compressed air, as the amount of air and maximum pressure head required would be unrealistic to implement. A compressed air method is viable though for smaller shallow-water vehicles, especially if cost effectiveness and power consumption are major issues. For example, the low-cost AUV "SubjuGator" uses a compressed air supply to control a variable ballast system (Novick, et aI., 1998).



Monterey Bay Research Institute Advanced Research Projects Agency

Fig. 1: The "Tethra" AUV.


General Fault Monitoring



Relief Valve

Air Out Valve




Air In Valve


0:: 0 0:: 0::





> 0 ;:;;






0 0 «



Ballast Tanks


Water valve

Fig. 2: Variable ballast system schematic. Cl Z

z~~ ....

Brook's (1999) behaviour based Subsumption hierarchy and are detailed in (Toal, et aI., 2002).




«Z 0::0

i: 0 a.








Active buoyancy control is realised using a variable ballast system. The ballast system consists of a compressed air supply, 4 ballast tanks and a series of valves and manifolds to control and distribute air and water in the system. Figure 2 shows a schematic for the system. A SCUBA air-tank provides the pressurised air-supply at 200 bar. This is regulated to lObar by a Mares regulation valve. Control of the air-supply and exhaust is done using submergible solenoid valves provided by BiS Valves Ltd.. For safety reasons submergible relief valves, also by BiS Valves Ltd., have been added on both the air-supply line and ballast tank lines. Both relief valves are set at 15 bar cracking pressure (differential). Control of water flow to and from the ballast tanks is completed using an electrically actuated ball-valve by Valpes. This valve is contained within a glass-fibre sealed enclosure, along with a pressure transducer to determine the pressure within the tanks. The enclosure and the submergible valves are mounted within the open frame. The air-lines are sub-sea hoses produced by Euroflow. The water lines are a combination of Euroflow hoses and a George Fisher PVC piping system. The four ballast tanks are mounted at half height in the craft and are equidistant from the centre, to ensure stability of the craft. Each tank has a volume of 2 litres. This provides approximately 2kgs of buoyancy per tank or 8kgs overall. The craft weights approximately l80kgs, so variable ballast accounts for 4.4% of the weight of the craft. If necessary ballast tank volume can be increased by changing tanks to give a greater +/buoyancy range. Figure 2 shows a schematic of the variable ballast system. Provisions have also been made to install a set of inflatable air-bags on the craft. These bags will form an Emergency Buoyancy System (EBS) that can be activated to bring the craft to the surface in situations of significant failure.






Fig. 3: Active buoyancy control electronics.


Active Buoyancy Control Electronic Hardware

The electronic hardware used to implement the active buoyancy control feature is split into three stages as show in figure 3. The first stage is the main CPU, on which the main robot controller operates. The second stage is a PlC based board, providing for lowlevel control functions such as timing and sequencing of valves and a platform for emergency buoyancy control. The third stage provides power circuitry for valve control and a means of manual operation of the system by radio control. It also incorporates signal conditioning for the external and ballast tank pressure transducers.


Normal Buoyancy Control Software

In order to realise buoyancy control the main robot controller operates the variable ballast system using four controlled states. State I is the normal state where the amount of ballast in the system is fixed. State 2 requires for ballast to be added while state 3 requires that ballast be removed. State 4 requires that the pressure within the tanks be equalised with the external ambient pressure. This is necessary to ensure safe operation of the system on the surface and to ensure non-gross buoyancy changes when control valves operate at depth. The CPU initiates these states and the EBS through the PlC board and interface board. The PlC board also establishes whether air should be added or removed for state 4. The current main autonomous robot controller algorithm is shown in figure 4. When deployed with this implemented controller the craft is trimmed with floats to be partially positively buoyant. Under radio control the ballast system is then used to bring it to a

r···F·Ai·Ls·AFE··BEH·Av·iolj' R·s 1 :.


Fl'l..l."bC"·m<1Inf1'no; 'rnIf'lTJlllll:~\f






---..---_._----- - , _ ...-+





. ~


---- - - - , - - - - ,.-0-











[::::::::::::~f.;;;~y~~;~~::::::::·::·I············· ""~".., soom

-------... MlIllam,~h,1Ilk




- - - -___








,.onmll:d """"


Fig. 5: Compensation behaviours.

r------ \ -

the first behaviour. Figure 5 shows both buoyancy change compensation behaviours.


-- cnr~~~


Fig. 4: Tethra's autonomous controller algorithm. positive/near neutral buoyancy and the ballast system is locked off before switching to autonomous mode. In autonomous mode the robot controller only uses thrusters for vertical motion and buoyancy control is utilised under fault conditions by means of a failsafe behaviour. However, there are envisaged mission scenarios other than fault conditions where it would be favourable to utilise buoyancy control. To achieve this specific buoyancy control behaviours are being developed. There are two main groups of behaviours: behaviours that compensate for uncontrolled buoyancy changes and behaviours that generate controlled buoyancy changes. The behaviours are designed to be incorporated within the main controller structure. Buoyancy compensation behaviours. Two uncontrolled buoyancy change compensation behaviours have been developed. The first behaviour deals with buoyancy changes due to external factors such as pressure, temperature and salinity. To detect these buoyancy changes the vertical thrusters are disabled and depth change is analysed by the controller using output from three different transducers: a pressure depth sensor, an altimeter and a vertically aligned log. As normally the craft is kept slightly positively buoyant the controller expects to read a slow ascent. If this is not the case then the behaviour will adjust buoyancy accordingly. Either symptomatic feedback from the vertical thrusters or an indicative reading from anyone of the three transducers triggers this procedure. The second behaviour deals with buoyancy changes due to intemal factors, e.g. during sample collection. In this case the behaviour can act on feedback from actuators to approximate the necessary buoyancy change required. For example, feedback from a mechanical attachment, used for collecting sediment samples, could be used to determine how much extra weight has been added to craft. An accurate trimming of buoyancy can then be completed using

Controlled buoyancy adjustment behaviours. These behaviours are used in scenarios where the controller requires a method of depth control that does not use vertical thrusters. Currently behaviours for two envisaged mission scenarios (described in section 2) have been developed: water clouding and large depth changes. These behaviours are shown in figure 6.


Emergency Buoyancy Control Software

Normal fault monitoring and fault conditions are dealt with by failsafe behaviours within the main controller. During normal operation the main robot controller monitors for fault conditions and uses the variable ballast system and EBS as appropriate. However, as the craft has a role as a test-bed for experimental control algorithms there is an increased likelihood that the CPU might crash or that behaviour/program faults would give unexpected results. Should this happen the main robot controller would become ineffective and could lead to possible craft damage or even loss. To counteract this possibility, and thereby improve the robustness of

r..·FAILS·AFE·BEH·AVioUR·S·.. ·j. :. ;. !5....--J'r---------...._ _ ~














.... :




Fig. 6: Controlled buoyancy adjustment behaviours.


, . . . - - - - - . , START


the craft performed admirably. Max speed was measured at 1.5m/s while maximum acceleration was found to be 0.75m/s 2 . It was well balanced in the water and showed good stability while under motion. The craft also proved to be agile, being able to perform turns within its own diagonal. There was no The water ingress in the hulls or enclosures. buoyancy system was also tested while in radio control mode. The four states were implemented manually and found to be operationally effective. In autonomous mode the craft was able to perform some basic behaviours such as cruise, dive, surface and turn efficiently and effectively. Owing to the confined nature of the test pool accurate measurements of dynamic characteristics of the craft have proved difficult. A series of open water tests are planned over the summer months. These tests will include a detailed analysis of dynamic characteristics such as velocity, acceleration and stability. The buoyancy control behaviours are currently in final stages of development and are being tested in a laboratory environment, the initial results are promising. Testing of the behaviours with the main controller.will soon commence in the testing pool, followed by open water trials where the behaviours can be tested within mission applications.

Fig. 7: Emergency control algorithm. 6 craft control, a simple emergency buoyancy control algorithm has been incorporated on the PlC board module. This emergency algorithm only operates in the following situations: • •


Tethra is currently undergoing a final stage of pool trials before full open-water testing commences. The results of these pool trials have shown that the practical approach adopted by the authors has been successful in developing an effective test-bed for autonomous controllers of underwater vehicles. The next phase of craft development will commence after the results of open-water trials are considered. It is hoped that the depth rating and general characteristics of the craft can be improved so as to make it more adaptable and to increase mission capability. The vehicle has been constructed to date on a tight research budget. With positive test results and extra finance the craft's performance can be significantly improved. For example, replacing the trolling motor thrusters with proper submersible robot thrusters will allow the craft's speed and depth rating to be significantly increased.

The CPU heartbeat input fails, indicating the CPU has crashed. The main controller repeatedly attempts to drive two or more variable ballast system states simultaneously. Mission time-out.

If one of these three situations occurs then the emergency algorithm will shut down the CPU and disconnect it from the buoyancy control system hardware. This is done to ensure that no malfunctioning processes can interfere with the control of the craft. Then a controlled ascent to the surface is initiated using the variable ballast system controlled states of operation, described in section 4.2. If necessary the algorithm will also utilise the EBS. A flow chart presenting the emergency algorithm operation is shown in figure 7.


Tethra is now a much needed and valuable research tool. By adopting a practical approach and providing a real world environment for application, it is hoped that research into autonomy and control at the University of Limerick can be more effective in providing real world solutions.

Tethra has been subjected to a series of tests and trials in a diving pool. The characteristics of the craft were first tested under direct radio control. Overall

Future research at UL in the area of buoyancy control will benefit greatly from the availability of a multi purpose test-bed. It is hoped that by utilising the


active buoyancy control feature on Tethra solutions to specific technology needs can be achieved. For example, Griffiths (1999) proposes a need for buoyancy control technology should AUVs develop roles as cargo carriers. In this case a post payload delivery buoyancy compensator, based on the behaviour structures detailed in section 4.2, could be developed to aid mission capability. Another possible technological goal is the development of soft grounding capabilities for underwater vehicles, similar to that presented in (Riedel, et aI., 1999) for a Naval Postgraduate School AUV.

REFERENCES Brooks, R. A. (1999). Cambrian Intelligence - The Early History ora New Al. Pages: 3-27. MlT Press, Cambridge, London, England. DeBitetto, P.A. (1995). Fuzzy logic for depth control of Unmanned Undersea Vehicles. IEEE Journal o(Oceanic Engineering, Volume: 20, pages: 242-248. Ferguson, J. and A. Pope. (2000). Explorer-a modular AUV for commercial site survey. Proceedings of2000 International Symposium on Underwater Technology. Pages 129-132. Griffiths, G. (1999) Technology needs for autonomous underwater vehicles. EUROMAR Workshop: Technologiesfor Ocean and Coastal Survey. Brussels, Belgium. Kirkwood, WJ. and D.E. Steele. (1994). Active variable buoyancy control system for MBARI's ROV. Proceedings ofOCEANS '94. 'Oceans Engineering/or Today's Technology and Tomorrow's Preservation.'. Volume: 2, pages: 471-476. Lu, Z., M. Hinchey and D. Friis (1997) Development of a small pneumatic subsea robot. IEEE Canadian Conference on Electrical and Computer Engineering. Volume: 2, pages: 442-445. Molnar L., D. Toal and C. Flanagan (2002). Evolving the control architecture for an autonomous underwater vehicle. The 33rd International Symposium on Robotics. Stockholm, Sweden. Pages 7- I I. Novick, D. K., R. Pitzer, B. Wilkers, C. D. Crane, E. de la 19lesia and K. L. Doty (1998). The Development of a Highly Manoeuvrable Underwater Vehicle. Robotics98: The 3rd international conference and exposition/ demonstration on robotics for challenging environments. Albuquerque, USA. Pages 168173. Riedel J. S., A. J. Healey, D. B. Marco and B. Beyazay (1999). Design and development of a low cost variable buoyancy system for the soft grounding of autonomous underwater vehicles. 9th UUST Symposium. University of New Hampshire, Durham, NH, USA.


Toal D., C. Flanagan, L. Molnar, S. Nolan and T. Love. (2002). Reactive control in the design of an autonomous underwater vehicle. Intelligent Engineering Systems through Artificial Neural Networks. Volume 12. Pages 527-535. ASME press. USA. Toal D., C. Flanagan, L. Molnar and G. Hanrahan (200 I). Autonomous Submersible Development for Underwater Filming Employing Adaptive Artificial Intelligence. Sensors and Their Applications Xl. London, UK. Pages 163- I68.