2015 IEEE International Conference on Robotics and Automation (ICRA) Washington State Convention Center Seattle, Washington, May 26-30, 2015

A Water Jet Thruster for an Aquatic Micro Air Vehicle Robert Siddall and Mirko Kovac Aerial Robotics Laboratory Imperial College London, United Kingdom [email protected], [email protected]

Abstract- Water sampling with autonomous aerial vehicles has major applications in water monitoring and chemical accident response. Currently, no robot exists that is capable of both underwater locomotion and flight. This is principally because of the major design tradeoffs for operation in both water and air. A major challenge for such an aerial-aquatic mission is the transition to flight from the water. The use of high power density jet propulsion would allow short, impulsive take-offs by Micro Air

Vehicles (MAVs). In this paper, we

present a high power water jet propulsion system capable of launching a 70 gram vehicle to speeds of llmls in 0.3s, designed to allow waterborne take ofT for an Aquatic Micro Air Vehicle (AquaMAV ). Jumps propelled by the jet are predicted to have a range of over 20m without gliding. Propulsion is driven by a miniaturised 57 bar gas release system, with many other applications in pneumatically actuated robots. We will show the development of a theoretical model to allow designs to be tailored to specific missions, and free flying operation of the jet.

I. INTRODUCTION

Locomotion in unstructured terrain is one of the most significant challenges to robots operating in an outdoor environment. Whilst many amphibious robots exist [1], these robots are not able to cross large, sheer obstacles, and can only exit the water on a gentle incline. Floatplane-style Micro Aerial Vehicles (MAV s) [1] are also equally inhibited by the presence of obstacles or waves on the water, requiring a large, clear area to take off, preventing them from entering confined spaces. We are aiming towards the development of an Aquatic Micro Aerial Vehicle (AquaMAV ), a fixed wing vehicle designed to fly to a target, dive into the water and subse­ quently execute an impulsive leap from the water surface, transitioning back to flight (figure 1). This robot will find use in disaster relief [2] and oceanography [3], [4], particularly in areas such as flooded collapsed buildings, or rocky, littoral ecosystems, where obstacles impede the free movement of conventional aquatic vehicles and prevent close observation by purely aerial robots. Several unmanned seaplanes are currently in operation [5], and studies have shown the poten­ tial of an aerial-aquatic robot that is propelled by adaptable flapping wings [6], or able to plunge dive into water [7]. Other work has demonstrated the efficacy of jumpgliding locomotion in terrestrial robots [8], [9], [10]. However, to the best of the author's knowledge, no AquaMAV has been realised to date. Use of high power density propulsion would allow a vehicle to take off from a confined space of water (figure 1). This would mean that an AquaMAV could dive into an

978-1-4799-6923-4/15/$31.00 ©2015 IEEE

Fig. l. A plunge diving AquaMAV can enter confined spaces of water for data collection, before retaking flight with a high power impulsive take-off

Fig. 2. The presented C02 powered water jet, with a shape memory wire valve and supercapacitor trigger circuit.

Fig. 3. Jet propulsion is used as a means of take-off by Flying Squid, to initiate extended jumpgliding leaps (background). The fabricated jet will be extended with soft folding wings, to allow similar leaps by small plunge diving AquaMAVs. A future prototype is illustrated here in the field. Background: Bob and Deb Hulse

3979

isolated area of water, where it could collect water samples and return underwater video footage. The vehicle could then perform a short take-off, and return to its launch site to submit collected samples and data for analysis. This would enable a fast, targeted response to emergency scenarios such as a stricken ship or a tsunami event, that could not be matched by conventional aquatic robots. In nature, a wide variety of animals are capable of crossing aquatic obstacles with ease by leaping from the water surface [11]. This can be achieved by swimming at high speed through the surface (flying fish, salmon), foot propelled surface jumping (mole crickets, basilisk lizards) or by flying appendages (seabirds). Uniquely, several species of squid are able to initiate gliding leaps by expelling a pressurised jet of water [12] (figure 3). This jet propelled launch is uniquely applicable to short take-offs by AquaMAV s. Jets of mass have a very rapid thrust response, unlike swimming leaps, and unlike foot propelled jumps, a jet continues to produce thrust in both air and water because it does not rely on external reaction forces. This allows a vehicle to lift free of the water and continue accelerate when airborne, where drag is dramatically reduced compared to locomotion in water [11]. While this thrust behaviour could be readily achieved with the use of com­ bustible rockets, typical small scale rocket propellants are hazardous both to those in the robot's operating environment and the user. Water jet propulsion offers a clean and safe alternative. In this paper we present a 40 gram jet thruster powered by compressed gas (figure 2), which can be rapidly recharged from a small carbon dioxide (C02) cannister, allowing a flying robot to perform several water sample return missions in succession. The thruster uses a shape memory alloy actuated valve to control the release of stored CO2, which is used to expel propellant water mass through a nozzle. In the following sections, we introduce the physical prin­ ciples behind water jet propulsion, and detail the key design features of a jet-propelled jumping robot. Consistent static thrust from the fabricated device and free flying launch is then demonstrated. II.

DE SIGN

Here we detail the key objectives in the vehicle design process, and the underlying theory behind water jet propul­ sion. We then present the implemented design solutions and the prototype that resulted. A.

Objectives

The robot is to form a critical component of a plunge diving AquaMAV, producing the thrust required to leap from the water surface after diving. Because it is intended for use in sustained flight, the device must be as lightweight as possible. A maximum mass budget of 40 grams has been allocated for the water launch system of the planned AquaMAV After a dive, the device will be actuated on the water surface and must propel the vehicle to a velocity sufficient for sustained flight. This means that the actuation

e

Fig. 4. Jet propulsion principle: Compressed gas released from a high pressure tank expels water through a nozzle, propelling the vehicle. Circled numbers correspond to the locations indicated by equation subscripts

of the device must be entirely self contained, and thrust production must not be sensitive to the variations in attitude due to the motion of the water surface. Water jet propulsion is often used in science education in the form of simple plastic water rockets [13]. However, the water rockets common in schools and amongst hobbyists are generally launched vertically, and use an off-board launchpad to both contain the pressure and actuate the rocket. These devices also contain air and water free to mix within a single chamber, relying on gravity and the rocket's acceleration to separate the two. If these rockets are at too shallow an angle the compressed gas can be expelled whilst water remains inside, resulting in little thrust and unexpelled mass. In order to be able discharge the device effectively at any angle or during 'sloshing motion', air and water must be kept separately. Implementing this as a concept for the AquaMAV propulsion system also minimises device weight, as it allows propellant water to be collected in situ at launch, rather than carried. We decided to design a vessel that can be charged using commonly available liquid CO2 cannisters, to create a powerful, but practical and portable device. CO2 is used because it is more compressible than air, and so stores more energy for a given pressure and volume. Liquid C02 capsules also offer the advantage of producing gas at a constant pressure (the liquid vapour pressure) until the tank is exhausted, so that a single capsule can recharge the device repeatedly without performance loss. The 16g capsules used to charge the device contain sufficient liquid for over 20 charges. B.

Theory

The thrust produced by an expelled jet of mass flow and velocity U4 is given by:

3980

m

(1)

A

B

Predicted Thrust 6·

5

5·

ro4

Z 4.

Cl..

� 3

....... III

23·

.!:

I-

Tank Pressures

::J III

�

2.

2

Cl..

lOa

Fig. 5.

200 Time (ms)

300

400

200 Time (ms)

lOa

a function of the pressure ratio between the tanks (equation 6). In both cases, the mass flow is dependent on the valve flow coefficient, Cv' To measure Cv, a valve was fixed in the open position with epoxy and the discharge rate of a column of water was measured. (5)

P2 PI

(2)

(3) 4

1 au P2 2 -ds+ - + -(u4 2 at Pw

(4)

Where u is the water velocity, P2 is the pressure of gas in the water tank, V2 is the gas volume, An is the jet cross sectional area and Pw is the density of water. The pressure acting on the water must be built up by the gas released from the CO2 tank. With a 57 bar reservoir, the tank gas outflow will initially be choked, and will be a function of only the reservoir conditions (equation 5). As the pressures at 1 and 2 equalise, and their ratio becomes non-critical, it will also be

)

(6)

To determine the variation of gas conditions in the gas and water tanks, a first law energy balance is used. The gas exchange is treated as being in quasi-equilibrium and as adiabatic, because the jetting takes place over too short a timescale for significant heat transfer to take place. This leads to an equation in which the enthalpy flux from the gas tank is equivalent to the enthalpy flux into the water tank, minus the final kinetic energy of the air (equal to the water velocity at the interface) and the pdV work done against the water pressure (equation 7). Gas conditions in the tank obey the ideal gas equation of state.

� dt

3

400

Simulated jetting process for water rocket with separated chambers: (A) Thrust. (B) Gas pressures in water and gas tanks.

If a gas is used as the propellant mass, its low density means that thrust production is negligible without very high exit velocities, and for efficient propulsion from a limited reservoir, a heavier propellant must be used. For an Aqua­ MAV, water is a propellant that can easily be collected before launch. Because of this, the presented robot has been designed to use water as its propellant mass, driven by the expansion of compressed gas. In this section we use the subscripts 1, 2, 3 and 4 to denote variables relating to the main gas tank, the gas within the water tank, the air-water interface, and the nozzle outlet respectively (figure 4). The incompressibility of water means the expelled jet will be at atmospheric pressure, and the gas expansion rate will equal the water outflow (equation 2). Upstream of the jet, the water flow can be treated as quasi-ID by assuming uniform flow across the jet section, with purely axial fluid velocity. By mass continuity, the local velocity is then a function of cross­ sectional area (equation 3). The unsteady form of Bernoulli's equation (equation 4) can be recovered from Euler's equation by integrating from the air-water interface to the nozzle exit (figure 4) [14]. Total pressure along a streamline between these two points is equal to the instantaneous gas pressure in the water tank. The air-water interface is assumed to remain planar throughout jetting, and viscous effects are neglected.

1

300

(

m2u� 2

)

(7)

Combining equations 2-7 leads to a system of 5 non-linear first and second order differential equations in V2(t), hI (t), h2(t) and mI (t). The solutions are obtained by numerical integration, which allows thrusts and launch trajectories to be computed. To calculate thrust, an additional force term, Fint(t) is included to account for the force experienced by the vessel due to the internal acceleration of fluid mass (equation 8) [14]. Fint

=

J �� Pw

dVwater

(8)

After all water is expelled, the release of the remaining gas produces a small amount of thrust. This can be computed from the stagnation quantities of the remaining gas, using equation 1 with an additional term, A4P4, added to the right

3981

A

C

D

+

Insulated steel bolt Brass contact

B

GRP Rods 1.6mm

1cm

SMA Wire Spring steel wires

First movement

Opening

Fully open

Schrader valve core

Fully closed

Fig. 6. (A) CAD rendering of the fabricated jet, with transparency added to show the valve actuation system inside the gas tanJe (B) Valve actuation photographs. (C) Illustration of the valve opening mechanism, with the vessel body forming a positive earth. (D) Fabricated valve actuator.

TABLE I

hand side to account for the fact that the nozzle outflow is not at atmospheric pressure. Nozzle conditions are calculated using standard Mach number relations (equations 9 and 10, where I is the gas adiabatic index, cp the gas heat capacity, and M the nozzle Mach number). Equation 7 here no longer holds and the mass flow out of the nozzle must also be included in the gas thermodynamic calculations.

WEIGHT BUDGET

Part Water Tank Gas Tank

P02

(10)

The simulated results for the prototyped geometry and valve are shown in figure 5. The gas discharge from the tank can be seen to be initially choked, unchoking at 70ms, and a small amount of additional gas thrust can be seen after the water is fully expelled at 160ms. After the gas built up in the water tank is expelled, further gas release through the valve produces little thrust, due to the restricted flow rate. C.

Prototype

The fabricated prototype consists of an air and water tank, both with sealed screw connections to a connecting centre­ piece, which contains a poppet valve core. This comprises most of the prototype's mass (table II-C). It was decided that in order for the device to be practical, the pressure release system must also be reusable, which precluded the use of 'single shot' systems such as a bursting diaphragm. To contain and release the gas, a Schrader poppet valve has been used, actuated by NiTi shape memory alloy (SMA) wires. The valve actuation is contained within the pressurised CO2 gas tank, so that it does not impede attachment to a canister for charging (figure 6A). The gas pressure vessel is constructed from 7075-T6 aluminium according to the European standard for high pressure vessels (EN 13445) [16], with an additional safety factor of 2, due to laboratory health and safety concerns. The water tank is not pressurised to the full 60bar during jetting

Carbon Fibre

+

weight [g]

Aluminium

Aluminium

11.05 11.53

Aluminium

+

Steel

5.84

Valve Actuator

NiTi, GFRP

+

Acetal

2.19

Connection seals

Steel

Centrepiece and Valve

Supercapacitor

P4

material

+

Rubber

Aerogel

3.93 4.20

MicrocontrolJer

1.4

Total mass prototype

40.14

(figure 5), and is not required to sustain pressure for extended periods. As a result, it is instead constructed from a woven CRFP tube with a 0.5mm wall thickness. This is bonded with epoxy to a aluminium screw connection and a vacuum formed polystyrene conical nozzle at either end (figure 6A). The system has a deliberate modular construction, with the centrepiece and its valve actuation system entirely self contained, so that gas and water tank sizes can be varied according to final mission design, and readily interchanged. The valve is opened by moving the valve stem a vertical distance of 1.6mm. When under 60 bar of internal pressure, the force required to open the valve is 24N, (19N of pressure force, 5N from the valve internal spring). These high force, short stroke actuation requirements are well suited to the use of Nitinol SMA wire, and a high pressure valve actuation system was designed, employing two 0.51mm Flexinol wires. These wires can produce repeatable contraction forces of up to 35.6N upon heating to 90°C [17], [18]. To provide sufficient stroke and force, a single length of 104mm flexinol wire is threaded through the valve stem, effectively creating two 501mn wires connected in series (figure 6C). The wires pull against a frame formed by glass fibre rods mounted into the centrepiece, guyed to the centrepiece by kevlar twine to hold the structure in compression (figure 6D).

3982

Static Thrust Data vs. Theory 6

..... -Theory 1st Charge - 2nd Charge . - 3rd Charge

��:?!!5S������ �

5

.

Z4 ..., VI

.

.

.

.

.

-

23 1- 2

..s::::.

50

100

150

200 Time (ms)

250

300

350

400

Fig. 7. Sequence of 3 thrust tests. The thrust tests were performed in succession, charged from the same capsule, and the cooling of the capsule during charging reduced the liquid C02 vapour pressure, causing a reduction in the gas tank initial pressure. The results otherwise show a high level of consistency.

After contraction, the wires must be stretched as they cool to their original length to reseal the tank. This re­ quires a minimum stress of 69MPa, which corresponds to a minimum total force of 23.2N in addition to the internal spring. To achieve this with minimal mass and volume, a pair of buckling spring steel wires with pinned ends are used. The wires are treated as elastic beams, and the large deformation elastica problem is solved analytically using elliptical integrals [19]. To achieve the required average force over the valve stroke, with the wire length constrained to be 200mAh), While this is not a weight efficient solution for actuation of the jet alone, it would be a mass saving when the jet is integrated into a flying vehicle requiring a larger

3984

power source. The spraying of water at the end of jetting, while not greatly detrimental to the overall thrust production, produces eccentric thrust, which has a strong tendency to spin the vehicle in flight. This may prove to be a problem for vehicles during takeoff. However, the effects could be mitigated by a reduction in the nozzle exit diameter, which would reduce the velocity of the water within the chamber, and delay the onset of spraying. This would also increase the jetting duration, and allow the vehicle more time to stabilise before it is perturbed by the final stage of the thrust. V.

CONCLUSION S AND FUT URE WORK

In this paper we have presented the design and fabrication of a novel water jet propelled robot. The modular design of this device means that it can be readily modified, and a theoretical model has been developed which will allow water jet thrusters to be optimised depending on the desired overall size and performance of the planned vehicle. The presented prototype has demonstrated powerful jumps, and has sufficient power to propel a small fixed wing vehicle with conventional propeller propulsion to flight speed. This will allow the implementation of a plunge diving water sampling and reconnaissance vehicle or a 'self .. recovering' long term monitoring buoy, either of which would drastically reduce the resource and manpower costs of marine data collection. Future work will focus on the implementation of folding wings into the vehicle to allow diving into water and to minimise drag during lift off. Design of the airframe for the vehicle will be integrated with a final optimisation of the jet geometry based on the presented theory. By varying the tank and nozzle diameters as well as the valve flow coefficient, it will be possible to maximise the work extracted from the gas, minimise system mass and tune the thrust profile to the aerodynamic requirements of a short take off. The completed vehicle will then allow the mechanics of short take .. offs from water to be characterised for miniature robots, and provide a pathway for the development of a new spectrum of Aquatic Micro Aerial Vehicles. V I.

Fig. 9.

[4] F. Shkurti, A. Xu, M. Meghjani, 1. C. Gamboa Higuera,

Y. Girdhar, P. Giguere, B. B. Dey, 1. Li, A. Kalmbach, C. Prahacs, et aI.,"Multi­ domain monitoring of marine environments using a heterogeneous robot team," in Intelligent Robots and Systems (IROS), 2012 IEEEIRSJ International Conference on. IEEE, 2012,pp. 1747-1753. [5] G. Meadows, E. Atkins, P. Washabaugh, L. Meadows, L. Bernal, B. Gilchrist, D. Smith, H. VanSumeren, D. Macy, R. Eubank, et aI., "The flying fish persistent ocean surveillance platform," in A1AA Unmanned Unlimited Conference,2009. [6] R. 1. Lock, R. Vaidyanathan, S. C. Burgess, and 1. Loveless, "Devel­

opment of a biologically inspired multi-modal wing model for aerial­ aquatic robotic vehicles through empirical and numerical modelling of the common guillemot, uria aalge," Bioinspiration & Biomimetics, vol. 5,no. 4,p. 046001,2010. [7] 1. Liang, X. Yang, T. Wang, G. Yao, and W. Zhao, "Design and experiment of a bionic gannet for plunge-diving," Journal of Bionic Engineering, vol. 10,no. 3,pp. 282-291,2013. [8] M. Kovac, 1.-C. Zufferey, and D. Floreano, "Towards a self-deploying and gliding robot," in Flying insects and robots. Springer, 2010,pp. 271-284. [9] A. L. Desbiens, M. T. Pope, D. L. Christensen, E.

W. Hawkes, and M. R. Cutkosky, "Design principles for efficient, repeated jumpglid­ ing," Bioinspiration & biomimetics,vol. 9,no. 2,p. 025009,2014. [l0] M. A. Woodward and M. Sitti, "Multimo-bat: A biologically in­ spired integrated jumpinggliding robot," T he International Journal of Robotics Research,2014. [11] 1. Rayner, " Pleuston: animals which move in water and air," Endeav­ our,vol. 10,no. 2,pp. 58-64,1986. [l2] R. ODor, 1. Stewart,

ACKNOWLEDGEMENT S

The authors would like to acknowledge the technical staff in the Imperial College Aeronautics department, in particular Roland Hutchins, Andrew Wallace and Mark Grant, whose help was indispensable during the project. This project is being funded by the UK Engineering and Physical Sciences Research Council.

[13] [14]

[l5] [16] [17]

REFERENCES

[18]

[l] R. Siddall and M. Kovac, "Launching the AquaMAV: Bioinspired

design for aerial-aquatic robotic platforms," Bioinspiration and

W. Gilly, 1. Payne, T. Cerveira Borges, and T. Thys, " Squid rocket science: how squid launch into air," Deep Sea Research Part II: Topical Studies in Oceanography,2012. 1. M. Prusa, "Hydrodynamics of a water rocket," SIAM review,vol. 42, no. 4,pp. 719-726,2000. C. 1. Gommes, "A more thorough analysis of water rockets: Moist adiabats, transient flows, and inertial forces in a soda bottle," American Journal of P hysics,vol. 78,p. 236,2010. P. Thomas, Simulation of Industrial P rocesses for Control Engineers. Elsevier Science, 1999. E. DIN, "13445-3," Un red P ressure Vessels-Part,vol. 3,2005. "Technical characteristics of flexinol actuator wires," Dynal!oy, Inc. L. M. Schetky, " Shape-memory alloys," Kirk-Othmer Encyclopedia of Chemical Technology,1982.

[l9] C. Wang, Applied elasticity.

Biomimetics,2014. [2]

Timelapse image of the jet being launched from an outdoor lake.

S. Kawatsuma, M. Fukushima, and T. Okada, "Emergency response by robots to fukushima-daiichi accident: summary and lessons learned," Industrial Robot: An International Journal, vol. 39, no. 5, pp. 428435,2012.

[3] P. Stevenson, S. McPhail, M. Tsimplis, and E. Higgins, "Air launched

platforms-a new approach for underwater vehicles," in Oceans 2009, 2009,pp. 1-8.

3985

McGraw-Hill, 1953.