High-PowerPropulsion StrategiesforAquaticTake-offinRobotics · 2 Robert Siddall, Grant Kennedy and Mirko Kovac Fig. 1: Plunge diving AquaMAVs map a marine accident, retaking flight

High-Power PropulsionStrategies for Aquatic Take-off in Robotics

Robert Siddall, Grant Kennedy ∗ and Mirko Kovac

Abstract The ability to move between air and water with miniature robots would allow dis-tributed water sampling and monitoring of a variety of unstructured marine environments,such as coral reefs and coastal areas. To enable such applications, we are developing anew class of aerial-aquatic robots, called Aquatic Micro Aerial Vehicles (AquaMAVs),capable of diving into the water and returning to flight. One of the main challenges inthe development of an AquaMAV is the provision of sufficient power density for take-offfrom the water. In this paper, we present a novel system for powerful, repeatable aquaticescape using acetylene explosions in a 34 gram water jet thruster, which expels watercollected from its environment as propellant. We overcome the miniaturisation problemsof combustible fuel control and storage by generating acetylene gas from solid calciumcarbide, which is reacted with enviromental water. The produced gas is then combusted inair in a valveless combustion chamber to produce over 20N of thrust, sufficient to propelsmall robots into the air from water. The system for producing combustible gases fromsolid fuels is a very compact means of gas storage, and can be applied to other forms ofpneumatic actuation and inflatable structure deployment.

Robert SiddallImperial College London, UK e-mail: [email protected]

Grant KennedyImperial College London, UK e-mail: [email protected]

Mirko KovacImperial College London, UK e-mail: [email protected]

∗ R. Siddall and G. Kennedy contributed equally to this work.

1

2 Robert Siddall, Grant Kennedy and Mirko Kovac

Fig. 1: Plunge diving AquaMAVs map a marine accident, retaking flight in a cluttered aquaticenvironment using powerful bursts of water jet thrust (adapted from Siddall and Kova�c [2014])

1 Introduction

1.1 Aquatic Micro Air Vehicles

Distributed water sampling by small unmanned aerial vehicles has applications in disasterrelief, oceanography, and agriculture, in which the monitoring of water contaminants isboth time and resource intensive. In remote or dangerous marine environments, aerialvehicles can provide samples to a base station over a broad area, and can respond morerapidly than aquatic or terrestrial systems, increasing the spatial range of data that canbe collected. We have proposed that implementing aquatic locomotion capability onto afixed wing flying vehicle offers a low cost, versatile solution (Siddall and Kova�c [2014]).

We are developing a small fixed wing aquatic Micro Air Vehicle (AquaMAV) ableto dive directly into the water from the air and subsequently retaking flight. A fixed wingvehicle provides greater range and speed than hovering vehicles, particularly at the smallscale, and the plunge diving approach reduces the need for accurate control, which allowsfor platforms to be produced at lower cost and operated in larger numbers. One of themost significant challenges in the implementation of a plunge diving miniature aerialrobot is the provision of sufficient power for transition to flight from water. The provisionof highly buoyant pontoons for a floatplane take off greatly inhibits aquatic locomotion,and makes diving from flight extremely difficult. A minimally buoyant vehicle mustinstead be able to produce significant thrust in order to propel itself out of the water. Wepropose that the best way to realise an AquaMAV is with the use of a powerful, impulsivethruster that can be easily integrated onto a flying platform.

Several projects have already demonstrated the efficacy of water sampling with UAVsusing larger rotary wing vehicles (Ore et al. [2013]; Schwarzbach et al. [2014]), by takingsamples whilst hovering near the water surface. This approach relies on accurate sensingand control to maintain position while a sample probe is lowered, and has a limited range.Several large unmanned seaplanes are currently in operation (Meadows et al. [2009]; Sid-dall and Kova�c [2014]), experimental studies have shown the potential of an aerial-aquatic

High-Power Propulsion Strategies for Aquatic Take-off in Robotics 3

A B

Fig. 2: A: A jet propelled jumpgliding robot launching out of water using a CO2 powered water jetthruster. B: A timelapse image of the robot’s trajectory, with wings deployed in the final still (Siddalland Kova�c [2015a]).

robot that is propelled by adaptable flapping wings (Lock et al. [2014b]; Izraelevitz andTriantafyllou [2014]), or able to plunge dive into water (Liang et al. [2013]). Other workhas demonstrated the efficacy of jumpgliding locomotion in terrestrial robots (Desbiens etal. [2014]; Vidyasagar et al. [2015]; Woodward and Sitti [2014]), and fixed wing MicroAir Vehicles (MAVs) have been implemented with terrestrial mobility (Jones et al. [2006]).Amphibious robots have been implemented in many forms (Lock et al. [2014a]; Ijspeertet al. [2007]), but these robots are not able to cross large, sheer obstacles, and generallycan only exit the water on gentle inclines. Such robots would have limited use in situationswith large amounts of debris separating bodies of water (e.g. a flooded town). Despitethe recent research effort in this emerging topic of investigation, to the best of the authors’knowledge no Aquatic MAV (AquaMAV) has been realised to date.

1.2 High Power Density Actuation

The challenge of equipping robots with large power reserves for rapid, impulsive move-ments is a recurring problem in robotics. This is often addressed with the use of elasticenergy storage, compressed gas, or combustible fuels, all of which can release largeamounts of energy in a short space of time. A recent example is the `Sandflea’ robotproduced by Boston Dynamics, a 5kg wheeled vehicle that uses a CO2 powered pistonto jump 9m into the air. Siddall and Kova�c [2015b] created a 100 gram jumpgliding robotwith deployable wings, that was able to leap out of the water into flight using a miniatureCO2 tank to propel water through a nozzle (figure 2). However, this prototype must berecharged between actuations.

Combustion offers an alternative means of creating high pressures, without the needfor a pre-pressurised tank and release mechanism. Churaman et al. [2012] created robotsweighing only 314 mg which were able to jump over 8cm vertically using explosiveactuators. A larger robot was built by researchers at Sandia Laboratories, which usedcombustion to power piston driven jumping, achieving 4,000 2m jumps from 20 gramsof fuel (Weiss [2001]). More recent work has focussed on the use of combustion to


Fig. 3: Calcium Carbide: The 0.5 grams of fuel shown here is sufficient for 10 launches of the 34 gramthruster presented in this paper.

power soft terrestrial jumping robots; Tolley et al. [2014] and Loepfe et al. [2015] havedemonstrated the use of explosive hydrocarbon gas for locomotion in small soft robots.These systems use pressurised liquid reservoirs of butane gas as fuel, metered into acombustion chamber by electronic valves, and ignited by a spark.

However, the provision of multiple liquid fuel tanks and flow control apparatus is asignificant mass and complexity penalty for a miniature flyable thruster. In this paper,we present a method for producing water jet thrust explosively, in a system that weighsonly 34 grams, using a solid fuel reserve to produce combustible acetylene gas, whichis ignited in a valveless combustion chamber.

2 Solid Reactants as a Combustion Gas Source

Large volumes of fuel gas can be stored in a small space as a liquid under pressure.However, this means that the storage and regulation systems must be able to sustainthe large pressures required, which makes components considerably heavier, and theprovision of a pressurised container of combustible gas is a hazard in and of itself. If thecombustible fuel is instead produced by the reaction of two separately stable components,high pressures can be avoided and the fuel storage and dispensing systems can be greatlysimplified. In the particular case of an aquatic robot, water can be used as a reactant,exploiting the robot’s environment to reduce system mass.

The gas production efficiencies for several water-reacting solid fuels are summarisedin table 1, with fuel required calculated based on a stoichiometric pre-combustion mixwith air at standard temperature and pressure (STP). The energy content of the gasesgiven off (again for combustion in air at STP) are given in table 2, with the butane usedby Loepfe et al. [2015] and Tolley et al. [2014] included for reference. Energy density isbased on the fuel’s high heating value (HHV, values taken from McAllister et al. [2011]),which includes energy recovery from the condensation of steam given off by the reaction,to consider work extraction from the products down to ambient temperature.

The use of solid compounds for gas storage is common in many applications. Oneexample is the use of sodium azide (NaN3) decomposition to release nitrogen (N2) for carairbag deployment, and more recently solid alkali metal hydrides have been explored bythe fuel cell industry as a compact means of hydrogen storage. More relevant to aquaticpropulsion are studies undertaken by the US Navy (Newhouse and Payne [1981]) exam-


Table 1: Water-reactive solid fuel production efficiency, the combustible gas product is highlighted in bold.

Fuel Reaction mg fuel per cm3 µl fuel per cm3

combustor volume combustor volume

Lithium (Li) 2Li+2H2O−−→ 2LiOH + H2 0.18 0.34

Lithium Hydride (LiH) LiH+H2O−−→ LiOH + H2 0.10 0.13

Lithium AluminiumHydride (LiAH4)

LiAH4+4H2O−−→LiOH+Al(OH)3 + 4H2

0.12 0.14

Sodium (Na) 2Na+2H2O−−→ 2NaOH + H2 0.60 0.62

Sodium Hydride (NaH) NaH+H2O−−→ NaOH + H2 0.31 0.22

Potassium (K) 2K+2H2O−−→ 2KOH + H2 1.02 1.18

Calcium (Ca) 2Ca+2H2O−−→ 2CaOH + H2 1.06 0.69

CalciumCarbide (CaC2)

CaC2+2H2O−−→ Ca(OH)2 +C2H2

0.22 0.10

Table 2: Gas combustion energy content

Gas Reaction (in air 20.9% O2 by vol.) Combustion energy density(J/cm3, based on HHV)

Hydrogen (H2) 2H2+O2−−→2H2O 3.6 (100% H2)Acetylene (C2H2) 2C2H2+5O2−−→4CO2+2H20 4.2 (119% H2)Butane (C4H10) 2C4H10+13O2−−→8CO2+10H2O 3.8 (106% H2)

ining Lithium Hydrides for use in a torpedo propulsion system, or work by Borchseniusand Pinder [2010] examining solid reactants for buoyancy control. Alkali metal hydrideshave particularly high volumetric gas production per gram of reactant, and among thealkali metals and their hydrides, lithium hydride produces the largest volume of hydrogenper gram reacted with water (table 1). However, the compound itself is acutely toxic, andan inhalation hazard, because it is generally supplied as a powder.

The alkali metals of group 1 are very malleable, and so the hazards of powder canbe avoided. However, these elements have low melting points, and the reactions betweengroup 1 elements and water is sufficiently violent to melt the metal itself and ignite thehydrogen produced, even at low quantities. This would make a fuel supply vulnerable to un-planned ignitions, and requires that the fuel be stored under oil. This would add significantcomplexity and weight to fuel storage and release systems. We considered lithium, sodiumand potassium, the least reactive of the series. The higher weight elements have lower melt-ing points, react too violently and some (caesium) are radioactive, and so were excluded.

Of the fuels in table 1, calcium carbide (CaC2, figure 3) and lithium aluminium hydride(LiAH4) are the safest to store and easiest to handle. Of the two, LiAH4 has the greatestweight efficiency, but CaC2 has the greatest volumetric efficiency (reducing the fueltank size and weight). In addition, the acetylene gas produced by CaC2 has a greater


combustion energy than hydrogen (table 2). Therefore, CaC2 was selected as a fuel sourcefor the prototype presented in this paper.

3 Theory

In order to size the engine and predict thrust, we created a simple model of the combustionand water expulsion. In this section we use the subscripts 1, 2 and 3 to denote variablesrelating to the the expanding combustion gases, the air-water interface, and the nozzleoutlet respectively (figure 4A). The thrust force F produced by an expelled jet of massflow �m3 and velocity u3 is given by equation 1.

F= �m3u3 (1)

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). The water flow withthe tank is treated as quasi-1D by assuming uniform axial flow across the jet section.

u3= �V2/A3 (2)A2u2=A3u3 (3)

Where u is the water velocity, An is the jet cross sectional area and V2 is the volumeflow of water out of the tank. The unsteady form of Bernoulli’s equation (equation 4)can be recovered from Euler’s equation by integrating from the air-water interface to thenozzle exit (figure 4A). Total pressure along a streamline running from 2 to 3 is equalto the instantaneous gas pressure in the chamber:∫ 3

2

∂u∂t

ds+p1−pa

ρw+

12(u2

3−u22)=0 (4)

Where p1 is the combustion gas pressure, pa is atmospheric pressure and ρw is thedensity of water. To model the pressure of the combustion products driving jetting, theheat addition from combustion is assumed to take place rapidly, such that the enthalpy ofthe combustion gases increases by the HHV (49.92MJ/kg fuel, McAllister et al. [2011])instantaneously at the start of combustion. The precombustion atmosphere is mostly air(93% by volume), and so the combustion products are treated as air, obeying the ideal gasequation of state (equation 5). The combustion chamber has a small vent needle (inside di-ameter 0.6mm) fitted to allow it to fill passively with water (section 4.2), that vents a smallamount of the combustion products. Flow through this needle will be choked, with massflow given by equation 6. A first law energy balance is then used to account for both thework done against water pressure and the energy lost through the vent needle (equation 7).


p1(t),

patm

u3(t)x

y

1

2

3

A

0 0.2 0.4 0.6 0.8 10

2

4

6

8

10

12

Performance SpectrumFinal Engine Fraction 0.43

Bm 1

h1(t)

u2(t)

Volume Fraction of Water in Combustion Chamber

Pred

icte

d Sy

stem

Spe

ci�c

Impu

lse (m

/s)

Fig. 4: A: Jet propulsion principle: Expanding combustion products drive a jet of water through a nozzle,propelling the vehicle, with a small amount of gas lost through a choked needle. B: The total thrustermomentum change extracted from the combustion products can be maximised by finding the optimumair-water ratio in the combustion chamber using a numerical model.

p1=m1RT1/V1 (5)

�m1=γ√

γ−1

(γ+1

2

)− 12

(γ+1γ−1

)p1Aneedle/

√cpT1 (6)

�H1=−p1 �V1− �m1h1 (7)

Where γ, cp and R are the adiabatic index, heat capacity and gas constant of air, andH1 and T1 are the gas enthalpy and temperature. The inclusion of a simple conductive heattransfer model of energy loss through the chamber walls into equation 7 had negligibleeffect on predictions (1% decrease in total impulse), so the process was treated as adia-batic. The model overpredicts the total thrust production (section 5), which is interpretedas being largely due to incomplete combustion and quenching of the hot combustionproducts by the water in the chamber.

3.1 Optimum Gas-Water Ratio

For a given combustion energy density, the volume of air before launch determines thetotal energy that can be recovered as thrust. Once the stored water has been expelled,further gas expansion produces negligible thrust, but if too little gas is stored, there isinsufficient energy to expel all the water at significant speed. There then exists an optimumratio of air to water. To obtain this optimum, the system specific impulse, Isp, is used asan objective. This is numerically computed by integrating the predicted thrust profile withrespect to time, and dividing it by the total mass of the thruster before launch, including themass of water in the combustion chamber (equation 8). This maximises the momentumimparted to the thruster, and consequently the launch height that can be achieved.


(A) (C)(B) (D) (F)

Fig. 5: Illustration of the jet firing sequence: (A) Water enters combustion chamber, with air exitingthrough the central needle. (B) Water reaches central needle and further water is prevented from entering.Fuel tank plunger is actuated, drawing a drop of water onto the fuel reserve. (C) Acetylene gas isproduced, enters the combustion chamber and mixes with air. (D) A spark is created in the chamber,igniting the Acetylene. (E) After combustion, the high pressure products expel the water, producingthrust. Air flow through the narrow needle is choked and limited. After combustion, the sequence canbe repeated, and furhter fuel reserves reacted.

Isp=∫

Fdt/mtotal (8)

Where mtotal is the total mass of the vessel at the point of launch, including water.Based on the geometry of the final vessel and the HHV of acetylene, a water volumefraction of 0.45 was predicted to give the maximal performance (figure 4B).

4 Engine Design

In this section, we introduce the operation of the fabricated explosive water jet thruster(figure 5) and several key design features, including a valveless combustion chamberwith passive water fill control, and an on board ignition system. The final engine designrequires 4.5ml of acetylene for stoichiometric combustion, produced by reacting 13mgof CaC2 with 8µl of water. The CaC2 used is supplied in 1mm granules of 75% purity(figure 3), and the 1ml fuel tank can hold sufficient fuel for 10 launches. Twice as muchfuel as is necessary is stored in the system and further room is left in the fuel tank, becausethe Ca(OH)2 reaction byproduct has a tendency to passivate the CaC2 reaction afterseveral fuellings, by preventing water from reaching unreacted CaC2.

4.1 Fuelling System

The fuel system uses a simple syringe primer mechanism to produce Acetylene in smallquantities. This system consists of a narrow (�0.5mm) Teflon pipe connected to thecombustion chamber below the water, running to the fuel tank. The fuel tank is otherwise


Engine Part Mass [g] % of Total

Combustion chamber 12.1 35.5

CaC2 Fuel (10 shots) 0.3 0.9

Fuel Tank / Plunger 2.5 7.3

Fuel Servo 4.4 12.9

Piezo igniter 2.6 7.6

Igniter actuator / frame 3.5 10.3

Supercapacitor 4.2 12.3

Microcontroller 1.5 4.4

100mAh LiPo Battery 3.2 9.4

Total mass 34.1 100

Table 3: Thruster weight breakdown.

Combustion chamber

Piezoigniter

Fuel tank

Fuelling servo

Flexinol wire

Supercapacitor

Fuel line

Spark gap

Fig. 6: Fabricated prototype, with removable fueltank, fuel priming pump and shape memory alloypiezoelectric ignition system.

sealed to the outside atmosphere, and surface tension is sufficient to prevent water fromentering the fuel tank through the teflon pipe.

To force water into the fuel tank, a small plunger in the top of the tank is actuated. Asthe plunger is retracted, water is drawn from the combustion chamber through the pipe,onto the CaC2 in the fuel tank. The water then reacts with the fuel, and Acetylene gas isproduced. The resulting pressure pushes acetylene back through the pipe and into the com-bustion chamber, where it can be ignited. The 8µl of water required is near to the smallestdroplet that can be formed on the end of the teflon tube. Because of this, priming and sparkdelivery occur in a timed sequence, with 1s between the introduction of a water drop intothe fuel tank (the start of acetylene production) and combustion, This stops overproductionof acetylene from preventing ignition, in the event a larger droplet is aspirated.

To actuate the fuel system, the plunger is driven by a small 4.4g servo motor, whichis controlled by an Arduino microcontroller. By modifying the stroke of the plunger andthe dwell time at the top of the stroke, this process can be accurately controlled so thatthe correct amount of water is aspirated into the fuel tank, controlling the amount of fuelreacted and consequently the volume of acetylene produced.

4.2 Combustion Chamber

The acetylene produced is then ignited in a combustion chamber partially filled with water.This chamber has been designed not only to contain the acetylene explosion, but also tocontrol the fill level of water, ensuring that the correct gas-water ratio is maintained priorto firing (section 3). This is achieved passively without the use of valves, by exploitingthe choking of high pressure gas flow through a narrow needle. With the vehicle floating


nose up on the water surface, air can pass through the needle, allowing water to enter thecombustion chamber through the exit nozzle. The end of the needle is positioned so thatit is blocked when the water reaches the desired fill point, and the chamber fills no further,leaving sufficient air volume for combustion.

The air remaining inside the chamber can then be mixed with acetylene and ignited.During engine firing, pressure inside the combustion chamber reaches up to 10 bar within5 ms. Consequently, the outflow through the narrow needle used to allow water to enterwill choke, limiting the mass flow rate through the needle. Because of this, high pressure ismaintained inside the combustion chamber for a short time, and the water inside the cham-ber is forced through the main nozzle, generating thrust. A small amount of water is alsoejected through the needle during firing, but this will only further restrict needle outflow.

The fabricated combustion chamber is constructed in two parts from high impactpolystyrene. The lower section is a cylinder terminated by a 20o cone, which forms thewater outflow nozzle. The upper section is a hemispherical dome which contains theignition spark gap and the regulator needle. The regulator needle has an inside diameter of0.6mm. Performance would be improved if this were smaller, but narrower needles had atendency to block with soot and prevent refilling of the chamber with water after ignitions.

4.3 Ignition System

Experiments with hotwire and spark systems showed that a spark produced ignition morereliably, and a hot wire was prone also prone to breaking during firings. The fabricatedprototype uses a piezoelectric igniter, which can produce electric arcs of up to 1cm inlength. The igniter is connected to two spark gaps in series, placed at different points inthe chamber, to produce two smaller sparks and ensure ignition where the air and fuelhave mixed poorly. The central needle vent forms the final contact.

To produce a spark, the piezoelectric igniter requires 30N of force over a 5mm stroke,which is produced by two Flexinol NiTi shape memory alloy (SMA) wires in series. Thewires are sheathed in teflon and the tubes sealed with grease to allow the wires to actuateunderwater. Wire actuation is driven by an aerogel supercapacitor, triggered by the samemicrocontroller used to drive the servo. The wires are mounted through carbon fibre tubes,which support the compression forces on the igniter, and are also used to mount the otherengine components (figure 6).

5 Combustion Tests

The operation of the fabricated final prototype was tested statically, with the combustionchamber and control components mounted to a perspex sheet, which was mounted to afixed force sensor. During tests, the jet was mounted vertically, and immersed in a 5 litreglass tank. The fuelling and ignition systems were then actuated, force data was collectedat 2500Hz (figure 8B), and the engine was filmed at 480fps. The recorded data shows a


0 ms +14ms+12ms+10ms+8ms+6ms+4ms+2ms +16ms +18ms

Fig. 7: Video sequence of the jet combustion, filmed at 480fps: The acetylene explosion rapidly increasesthe internal pressure to up to 10 bar, expelling water and producing over 20N of thrust.

0 10 20 30 40 50 60 70 800

5

10

15

20

25 Combustion Test Thrust Pro�les

Time (ms)

Thru

st (N

)

Prediction, Total Impulse = 0.43 Ns1st Ignition, Total Impulse = 0.25 Ns2nd Ignition, Total Impulse = 0.30 Ns3rd Ignition, Total Impulse = 0.27 Ns

Fig. 8: Experimental Data: Static thrust measurements of prototype engine compared to theoreticalthrust prediction. The prediction assumes combustion takes place instantaneously, while in practice peakpressure occurs around 5-10ms after ignition, and the predicted thrust curve has been shifted left to offera clearer visual comparison.

good level of consistency between successive actuations, although maximal pressure wasnot achieved repetitively. Video data showing gas bubbles leaving the nozzle exit indicatedthat all water was expelled during each explosion. The recorded thrust time histories wereintegrated to give the total impulse, which were 0.25Ns, 0.30Ns and 0.27Ns for the first,second and third ignitions respectively.

6 Discussion

6.1 Combustion Tests

The shape and duration of the measured thrust profiles compare well to theory, but totalimpulse is reduced by 28%-40% from prediction. The fall off in thrust after the initial peakis much steeper in the measured data. This means that pressure is dropping in the chamberfaster than predicted by the change in gas volume and needle outflow. This is interpretedas the both result of incomplete combustion and quenching of the hot combustion productsby the water they are expelling, neither of which are accounted for by the model.


While the tests show a level of consistency sufficient for design, there is also a vari-ation in total impulse of approximately 10% between successive tests. Examination ofthe data suggests that this is due to incomplete combustion, due to inhomogeneities in thepre-combustion atmosphere: In the two tests which showed less than the maximum peakpressure, oscillations in thrust at approximately 500Hz can be observed. The oscillationsare characteristic of `knock’ in internal combustion engines, caused by autoignition ofunburned gases compressed by the combustion products. In internal combustion engines,the tendency for a fuel to autoignite is expressed in terms of octane number, with a lowernumber indicating a lower level of compression for autoignition, and a stronger tendencyto knock. Acetylene has a Research Octane Number (RON) of 50 (McAllister et al.[2011]) indicating a high susceptibility to autoignition (this can be compared to commonautomobile petrol with an RON of 100). The presence of secondary ignitions towards theend of the gas expansion indicates that the initial combustion was not complete, resultingin reduced pressure. This is borne out by the lack of secondary ignitions in the higherpeak thrust profile.

The incomplete combustion does not have a significant effect on the overall perfor-mance, but the actuation consistency could nonetheless be improved by modificationof the gas dispersal system. Discharging the acetylene through a smaller tube will buildhigher pressure in the fuel tank and result in smaller, more scattered gas bubbles, orthe splitting of the fuel line to enter the engine at two points could increase the mixhomogeneity before ignition.

6.2 Engine design

The prototype has demonstrated the functionality of the key systems of the engine. Thefuelling system is able to meter the gas accurately enough for powerful combustion, andthe combustion chamber’s valveless design allows repeated refuelling and ignition. Thepresence of water, and the short duration of the combustion means that the polystyrenecombustion chamber showed no signs of melting or burning after over 20 actuations.

The most significant loss in performance comes from the choked needle, which ventsenergy from the combustion chamber. Making the needle narrower, to reduce this losswould improve performance, but prevents the chamber from refilling with water reliably.The losses could instead be greatly reduced by the addition of a sprung check valve with aminimum closing pressure, but this would add mass and complexity, and the functionalityof a sprung valve is liable to degrade with repeated exposure to explosions, unlike theimplemented choking system, which has no moving parts.

Importantly, the fuelling and ignition systems (50% of total engine mass) representthe device’s `mass capital’, and only the fuel tank will increase in size if the combustionchamber is scaled up. The system could therefore be readily resized for a larger payloador higher jumps, as a mission demands and achieve increased performance to weight. Thetotal cost of the engine is $35, including the arduino and lipo battery, enabling the deviceto be treated as disposable.


The fuel supply of the device is also a fraction of the total vehicle mass (0.9%), and anincrease in fuel capacity would allow the engine to function as an aquatic jump-glider, withfuel for a similar number of jumps to many terrestrial jumping microrobots. This wouldenable aquatic locomotion over obstacles in flooded or littoral areas in a similar capacityto terrestrial jumpers. Both for gliding locomotion and propelled flight, the engine designlends itself well to integration onto an airframe. The ignition and fuel systems are both onlylinked to the chamber by flexible wires and tubes, and so may be freely positioned withina vehicle, facilitating both stable mass distribution and compact design. This is also impor-tant because the filling and ignition sequence described in section 4.2 relies on the vehiclebeing upright, and the ability to position mass to achieve this passively will be important.

6.3 Comparison to Another Water Jet Thruster

Siddall and Kova�c [2015a] previously demonstrated aquatic escape with a similar system,using a tank of compressed CO2 instead of an acetylene explosion to drive the expulsionof water. This system weighed 40 grams, and was able to propel a 100 gram flying robotto a speed of 12m/s from beneath the water. Comparing the thrust profiles from Siddalland Kova�c [2015a] with the measured thrust from the acetylene engine (figure 9) it canbe seen that while the peak thrust is much higher, the acetylene engine has a smaller totalimpulse. However, the CO2 thruster can only produce a single shot, and when the 10 totalshots of the combustion prototype are taken into account, the mass efficiency is muchgreater (table 4). With further design improvements, and an increase in engine size, thecombustion prototype is expected to be a far better system for aquatic escape.0 50 100 150 200 250 300 350 400

0

5

10

15

20

Time (ms)

Thru

st (N

)

Comparison of explosive CaC2 thruster with compressed CO2 powered thruster

Valved 57 bar CO2 driven jet (Siddall and Kovac, 2015)Valveless acetylene explosion driven jet

0 20 40 60 80 100 120 140 160 1800

5

10

15

20

Time (ms)

Thru

st (N

)

Comparison of explosive CaC2 thruster with compressed CO2 powered thruster

Valved 57 bar CO2 driven jet (Siddall and Kovac, 2015), IT = 0.80NsValveless acetylene explosion driven jet IT = 0.30Ns

Fig. 9: Comparison of the explosive thruster presented here with the single shot CO2 aquatic escapethruster presented by Siddall and Kova�c [2015b].


Table 4: Explosive CaC2 water jet system compared to the compressed CO2 thruster presented by Siddalland Kova�c [2015b].

Thruster Type Mass Peak Thrust Total Impulse(per shot)

System Specific Im-pulse (inc. total shots)

Compressed CO2 Thruster 40.1 g 5.1 N 0.80 Ns 19 ms-1

Explosive CaC2 Thruster 34.1 g 21.5 N 0.30 Ns 88 ms-1

6.4 Other Potential Combustible Fuel Sources

Fuels were chosen principally based on their source reactant, rather than the performanceof their product gas. Other combustible gases such as methane and hydrogen have a muchbetter resistance to autoignition and may ultimately be better fuel choices for the designif consistency is a strong design consideration. In addition, hydrogen and methane canbe sourced directly from the environment, rather than a fuel reservoir, methane throughthe use of methanogenic bacteria, and hydrogen by direct electrolysis. The latter processalso generates pure O2 and so could better use the available combustion volume. Eitherof these approaches would potentially created a fully renewable energy source for aminiature robot, while the engine presented here has a limited fuel capacity, although itis comparatively very compact.

While hydrogen and methane as fuels offer the potential for fully renewability, acety-lene has an different advantage in that it is also one of the most detonateable fuels. If anarrow tube of fuel and air is ignited at one end, the flame front will accelerate down thetube, and if the front is able to travel a sufficient distance, it will transition to a travellingnormal shockwave (deflagration to detonation transition). This shockwave precompressesthe reactants and results in much higher combustion pressures, on the order of 20bar, whichhas the potential to enhance the maximum power output of the device. By modifying thechamber geometry, the calculations of Silvestrini et al. [2008] applied to combustion on asimilar scale to the prototype presented here suggest that a deflagration to detonation tran-sition of the flame front would be achievable in a high aspect ratio combustion chamber,if the fuel could be appropriately premixed. This will be explored further in the future.

7 Conclusions

In this paper we have demonstrated the use of CaC2 as a fuel source to repeatably generateexplosions in a miniature robot. These explosions can be harnessed to produce very highpower to weight ratios, which are otherwise unachievable with conventional miniaturesystems. The fabricated thruster is entirely self contained and weighs only 34 grams, withsufficient fuel for 10 controlled explosions. The device scales well, and an the power orfuel capacity of the thruster could be increased without significantly increasing the totalsystem mass.


The fuelling and ignition processes used for the thruster could also be used to increasethe performance of other systems, such as a pneumatic piston or soft robot. The productionof gas from a solid reactant offers a highly compact means of gas storage, even more sowhen environmental water is used as a reactant. The solid liquid reaction used is also anaccurate means of metering gas production, as it is easier to accurately dispense liquidsthan gases. If the produced is not ignited, a similar solid reactant system could also beused in inflatable structure deployment.

References

J Borchsenius and S Pinder. Underwater glider propulsion using chemical hydrides. InOCEANS 2010 IEEE-Sydney, pages 1–8. IEEE, 2010.

Wayne Churaman, Luke J Currano, Christopher J Morris, Jessica E Rajkowski, SarahBergbreiter, et al. The first launch of an autonomous thrust-driven microrobotusing nanoporous energetic silicon. Microelectromechanical Systems, Journal of,21(1):198–205, 2012.

A L Desbiens, M T Pope, D L Christensen, E W Hawkes, and M R Cutkosky.Design principles for efficient, repeated jumpgliding. Bioinspiration & biomimetics,9(2):025009, 2014.

A J Ijspeert, A Crespi, D Ryczko, and J-M Cabelguen. From swimming to walking with asalamander robot driven by a spinal cord model. science, 315(5817):1416–1420, 2007.

J Izraelevitz and M Triantafyllou. A novel degree of freedom in flapping wings showspromise for a dual aerial/aquatic vehicle propulsor. arXiv preprint arXiv:1412.3843,2014.

K Jones, F Boria, R Bachmann, R Vaidyanathan, P Ifju, and R Quinn. Mmalv - themorphing micro air-land vehicle. In IROS 2006, 2006.

J Liang, X Yang, T Wang, G Yao, and W Zhao. Design and experiment of a bionicgannet for plunge-diving. Journal of Bionic Engineering, 10(3):282–291, 2013.

R J Lock, S C Burgess, and R Vaidyanathan. Multi-modal locomotion: from animal toapplication. Bioinspiration & Biomimetics, 9(1):011001, 2014.

R J Lock, R Vaidyanathan, and S C Burgess. Impact of marine locomotion constraintson a bio-inspired aerial-aquatic wing: experimental performance verification. Journalof Mechanisms and Robotics, 6(1):011001, 2014.

M Loepfe, C M Schumacher, U B Lustenberger, and W J Stark. An untethered, jumpingroly-poly soft robot driven by combustion. Soft Robotics, 2(1):33–41, 2015.

S McAllister, J Chen, and A Fernandez-Pello. Fundamentals of Combustion Processes.Mechanical Engineering Series. Springer, 2011.

G Meadows, E Atkins, P Washabaugh, L Meadows, L Bernal, B Gilchrist, D Smith,H Van Sumeren, D Macy, R Eubank, et al. The flying fish persistent ocean surveillanceplatform. In AIAA Unmanned Unlimited Conference, 2009.

H Newhouse and P Payne. Underwater power source study. Technical report, DTICDocument, 1981.


John-Paul Ore, Sebastian Elbaum, Amy Burgin, Baoliang Zhao, and Carrick Detweiler.Autonomous aerial water sampling. In The 9th Intl. Conf. on Field and Service Robots(FSR), 2013.

M Schwarzbach, M Laiacker, M Mulero-Pazmany, and K Kondak. Remote watersampling using flying robots. In Unmanned Aircraft Systems (ICUAS), 2014International Conference on, pages 72–76. IEEE, 2014.

R Siddall and M Kova�c. Launching the aquamav: bioinspired design for aerial–aquaticrobotic platforms. Bioinspiration & biomimetics, 9(3):031001, 2014.

R Siddall and M Kova�c. Fast aquatic escape with a jet thruster. Mechatronics,IEEE/ASME Transactions on (Submitted), 2015.

R Siddall and M Kova�c. A water jet thruster for an aquatic micro air vehicle. In Roboticsand Automation (ICRA), 2015 IEEE International Conference on. IEEE, 2015.

M Silvestrini, B Genova, G Parisi, and FJ Leon Trujillo. Flame acceleration and ddtrun-up distance for smooth and obstacles filled tubes. Journal of Loss Prevention inthe Process Industries, 21(5):555–562, 2008.

M Tolley, R F Shepherd, M Karpelson, N W Bartlett, K C Galloway, M Wehner, R Nunes,G M Whitesides, and R J Wood. An untethered jumping soft robot. In IntelligentRobots and Systems (IROS 2014), 2014 IEEE/RSJ International Conference on, pages561–566. IEEE, 2014.

A Vidyasagar, J-C Zufferey, D Floreano, and M Kova�c. Performance analysis ofjump-gliding locomotion for miniature robotics. Bioinspiration & Biomimetics, inpress, 2015.

P Weiss. Hop hop hopbots!: designers of small, mobile robots take cues fromgrasshoppers and frogs. Science News, 159(6):88–91, 2001.

M A. Woodward and M Sitti. Multimo-bat: A biologically inspired integrated jumpinggliding robot. The International Journal of Robotics Research, 2014.

Documents

High-PowerPropulsion StrategiesforAquaticTake-offinRobotics · 2 Robert Siddall, Grant Kennedy and Mirko Kovac Fig. 1: Plunge diving AquaMAVs map a marine accident, retaking flight