1748-3190_10_1_016005

This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 194.27.228.12

This content was downloaded on 14/05/2015 at 20:42

Please note that terms and conditions apply.

A bioinspired multi-modal flying and walking robot

View the table of contents for this issue, or go to the journal homepage for more

2015 Bioinspir. Biomim. 10 016005

(http://iopscience.iop.org/1748-3190/10/1/016005)

Home Search Collections Journals About Contact us My IOPscience

Bioinspir. Biomim. 10 (2015) 016005 doi:10.1088/1748-3190/10/1/016005

PAPER

A bioinspired multi-modal flying and walking robot

LudovicDaler1,3, StefanoMintchev1,3, Cesare Stefanini2 andDario Floreano1

1 Laboratory of Intelligent Systems (http://lis.epfl.ch) at Ecole Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne,Switzerland

2 BioRobotics Institute, Scuola Superiore SantõAnna, Polo SantõAnnaValdera, Viale Rinaldo Piaggio 34, I-56025 Pontedera (Pisa), Italy3 LDaler and SMintchev contributed equally to this work.

E-mail: [email protected]

Keywords: adaptivemorphology,multi-modal locomotion, flying robot

Supplementarymaterial for this article is available online

AbstractWith the aim to extend the versatility and adaptability of robots in complex environments, a novelmulti-modalflying andwalking robot is presented. The robot consists of a flyingwingwith adaptivemorphology that can performboth long distance flight andwalking in cluttered environments forlocal exploration. The robotʼs design is inspired by the common vampire batDesmodus rotundus,which can perform aerial and terrestrial locomotionwith limited trade-offs.Wings’ adaptivemor-phology allows the robot tomodify the shape of its body in order to increase its efficiency duringterrestrial locomotion. Furthermore, aerial and terrestrial capabilities are powered by a single loco-motor apparatus, therefore it reduces the total complexity andweight of thismulti-modal robot.

1. Introduction

Robots capable of hybrid air and ground locomotioncould be used for different search-and-rescuemissions[1, 2], exploration of hostile environments andenvironmental monitoring [3]. In particular, thecombination of forward flight and ground locomotionbrings dual advantages of travelling quickly over longdistances and thoroughly exploring a specific region ofinterest on the ground in order to, for example, searchfor victims trapped in partially collapsed buildings [4].

Multi-modal locomotion is a feature that increasesthe environmental adaptability, the locomotion versa-tility, and the operational flexibility of robots [5].Although multi-modal locomotion has a potentiallyhigh impact in robotics and has recently attractedmuch attention [3, 6], robots that successfully demon-strate competences in diverse environments are still atan early stage. Current prototypes show that theimplementation of any additional locomotion modecan potentially lead to performance losses (i.e. man-oeuvrability, speed, energetic efficiency) [7–10].Indeed, similarly to animals [6, 11–13], multi-modalrobots are subject to various trade-offs due to conflict-ing requirements imposed by locomotion on differentsubstrates. Therefore, the main challenge is to identify

design strategies that maximize performances over abroad range of substrates.

Current implementations of flying and walkingrobots are mainly based on an additive strategy (seefigure 1(a)), where secondary locomotion modes areobtained by using additional actuators and mechan-isms [7, 14, 15]ordedicated appendices, such aswheels[16], cylindrical cage [17], spherical cage [18] or legs[9] that are not directly used during flight. These addi-tions have an impact on weight and drag during flight,decreasing aerial efficiency and manoeuvrability. Fur-thermore, most of these robots have a morphologyhighly optimized for a primary locomotionmode, andare therefore less effective in the secondaryone.

Many animals can perform multi-modal locomo-tion by using an integrated strategy: this means that asingle locomotor apparatus, composed of actuators(i.e. set of muscles for animals or a motor for robots)and appendices (i.e. limbs for animals or mechanicalstructures for robots), is used for multiple modes oflocomotion. For example, seabirds of the Alcidaefamily exploit wing propulsion for both flight andswimming [19] and some bat species use their wingsboth for flying and walking [12]. In robotics, an inte-grated strategy is potentially advantageous because itlimits the number of mechanisms, actuators and sen-sors, hence the overall weight and complexity.

RECEIVED

24 June 2014

ACCEPTED FOR PUBLICATION

12November 2014

PUBLISHED

19 January 2015

© 2015 IOPPublishing Ltd

However, an integrated strategy is more challengingthan an additive strategy because a single locomotorsystem must accommodate potentially conflictingdynamics. The integrated approach has been adoptedin a preliminary version of the flying and walkingrobot [4] described here and in a jumping and glidingrobot [20].

Furthermore, animals such as flying snakes, sea-birds and some salamanders adapt their morphologyto the desired locomotion mode in order to enhanceperformance. Flying snakes, for example, flatten theirbody in order to increase lift and to travel longer dis-tances during gliding [21, 22]. Before diving into thesea, alcids partially fold their wings into a shape bettersuited to swimming. Johansson and Aldrin [23] sug-gested that the partly folded wings of the alcidsmay actas efficient aft-swept wingtips, reducing the induceddrag and increasing the lift-to-drag ratio.

The salamander is another example of animal thatuses adaptive morphology. Salamanders retract theirlegs along their body when they transition from walk-ing to swimming [24]. Some other species of sala-manders are also capable of rolling. Their body,originally shaped for walking, can take the shape of alargewheel to rapidly roll downhill [25].

Among all the animals with mixed aerial and ter-restrial locomotion capabilities, the common vampirebatDesmodus rotundus is a relevant case study. In gen-eral, batʼsmuscles andmorphology are highly adaptedfor flapping flight [26] and, compared to other mam-mals, bats move awkwardly on the ground. D. rotun-dus is a notable exception because, most probably dueto its blood-based diet [12], it evolved remarkable ter-restrial capabilities such as running [27] and jumping[28] with its wings. According to the literature, theseterrestrial competences do not appear to negatively

affect its flight ability, although further biological datawould be necessary to validate this hypothesis [29].D.rotundus has evolved an integrated strategy to multi-modal locomotion: a single locomotor apparatus, thepectoral muscles and the wings, are used to locomotein the air and on the ground. In addition, this animaladapts the morphology of the wings during the transi-tion from flight to terrestrial locomotion. Thisremarkable combination of an integrated strategy withan adaptive morphology explains the small trade-offsin the multi modal capabilities of this animal and isour source of inspiration for the design principles ofrobots capable offlying andwalking.

In a preliminary exploration of flying and walkingrobots, we proposed a robot [4] that can use the tips ofits wings as whegs [30] (see figure 1(b)). The advan-tage of this approach lies in the weight mitigationachieved by using the same structure for two locomo-tion modes. However, this explorative prototype usesadditional actuators dedicated only to walking and hasafixedmorphology.

The present paper describes the implementationof a novel flying wing with additional walking cap-abilities, which fully exploits the aforementioned bio-logical design principles. According to an integratedstrategy, the robot adopts the same actuators andappendices, called wingerons, for both flight controland walking on the ground. In addition, this robot hasfoldable wings, which provide morphological adapta-tion for switching from a wing shaped for flight to amore compact morphology adapted to ground loco-motion: deployed wings maximize lift during flight,while folded wings enhance the efficiency of the roboton the ground by increasing the grip of the wingerons.With the proposed design, terrestrial competences aresuccessfully endowed on a flying wing while losses of

Figure 1.Examples ofmulti-modal land and air robots.

2

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

aerial performance are minimized in terms of flightmanoeuvrability and cost of transport (COT) duringflight.

2. Platformdesign

We aim at a robot that could be used in search-and-rescue missions where it will have to cover longdistances in order to reach remote areas, therefore itsprimary mode of locomotion will be forward flight.Then, the robot should be capable to land and explorethe environment in order to locate victims, thereforeits secondary mode of locomotion will be terrestriallocomotion. Among all the possible configurations offlying vehicles, we have chosen a flying wing becausethe absence of fuselage and tail makes it simple toconstruct and robust in landings. The next step is toinvestigate how the basic components of a flying wingcan be usedwhenmoving on the ground. According tothe integrated design strategy, the same set of actuatorsshould be used for both flight control and groundlocomotion. In the proposed design (see figure 2), thetwo extremities of the wings, called wingerons, areused to control the pitch and the roll axes of the robotduring flight and are also used as whegs to power theground locomotion in unstructured environments.These wingerons are designed in such a way that thetwo modes of locomotion have compatible dynamics.Moreover, the robot has foldable wings in order toimprove its performance on the ground.

2.1.Dual usewingeronsThe effective use of the wingerons for multiplelocomotion modes in a flying wing (see figure 3(A))has been achieved using a two-step design process:

• The first step was to identify the best shape of thewingerons in order to accommodate the differentconstraints imposed by flight control and groundlocomotion.

• The second step involved the selection of suitableactuators and the sizing of wingerons for their

double use with the smallest possible flight effi-ciency loss.

Regarding the first step, three important aspectsconstrain the shape of thewingerons (seefigure 3(B)):

(1)The axis of rotation of thewingeron should be closeto its aerodynamic centre. The aerodynamic centreof any airfoil is the point where the pitchingmoment coefficient does not change with the angleof attack; this is also the point where the lift isapplied [31]. Thus, if the axis of rotation of thewingeron goes through the aerodynamic centre,the torque that must be applied by the motor toturn the wingeron is minimized and it also makesthe calculations easier since this torque will notchange with the angle of attack. The aerodynamiccentre is located at approximately 25% of theMeanAerodynamic Chord (MAC) [31].

Figure 2.Novelmulti-modal flying andwalking robot. The robot is equippedwithwingerons that seamlessly integrate flight controland terrestrial walking. Foldable wings adapt themorphology of the robot either for aerial or ground locomotion.

Figure 3. (A) Thewingeron is the portion of thewing used tocontrol theflight. (B) Zoomonone of thewingerons, whichshows the different parameters which have to be dimensionedand the constraints; (1) the centre of aerodynamic pressure(black dot) should be on the rotation axis, (2) the axis ofrotation should be in the centre of thewingeron ( =w w1 2)and (3) the trailing edge should be horizontal. (C) Thesolution that fulfils all these requirements.

3

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

(2)For the ground locomotion the axis of rotation ofthe wingeron has to be in its centre in order tominimize the peak torque required by themotor.

(3)The trailing edge of the wingeron should be parallelwith its axis of rotation, in order to maximize thegrip on the ground.

In order to satisfy these three requirements, we aregoing to demonstrate below that the wingerons shouldhave a triangular shape, as in figure 3(C). The MAClength is given by the following equation:

= −− +

+

⎡

⎣⎢⎢

⎤

⎦⎥⎥

( )MAC A

A B B

A B

2( )

3( ), (1)

A

2

where A is the root chord of the wingeron and B is itstip chord (see figure 3(B)). The swept distance atMAC, c, is given by:

= ++

c SA B

A B

2

3( ), (2)

where S is the swept distance at the tip of the wingeron(see figure 3(B)). The following equations must besatisfied in order to, respectively, minimize the motortorque, have an horizontal trailing edge, and have thecentre of rotation at 25%of theMAC distance:

=w w , (3)1 2

= +A S B, (4)

= +w cMAC

4. (5)1

Thus, by replacing c andMAC by their expressionandwith + = +w w S B1 2 wefind:

− + =S AS A2 0. (6)2 2

This equation has only one solution: S = A andtherefore B = 0. It follows that the wingerons shouldhave a triangular shape as shown in figure 3(C). Thistriangular shape has a large swept angle which creates avortex at the tip of the wingeron. This vortex preventsstalling of the wingerons at large angles of attack andthus increases the performance during aerobaticsmanoeuvres that require high deflection of thewingerons.

Regarding the second step, namely the selection ofactuators and the scaling of the wingerons, it is impor-tant to consider that flight manoeuvrability requiresrapid wingeron movements and low torque, whileground locomotion demands high torque and lowerrotational speed. These differences in torque and rota-tional speed requirements make the selection of a sin-gle actuator difficult.

To appreciate this issue, letʼs consider the operat-ing range of a conventional dc motor (see figure 4). Itis constrained by twomain factors: the torque that canbe continuously delivered by the motor is limited byheat dissipation, and the maximum speed is primarilylimited by the wear effect in the commutations

systems and in the bearings. Moreover, the maximumcontinuous torque decreases with the speed, thus thereis a trade-off between torque and speed that can bedelivered by the motor. Usually, if two locomotionmodes have very different dynamics (i.e. one needsvery high speed and the other very high torque), anactuator dimensioned for one locomotion mode willnot be suited for the other and vice-versa. For exam-ple, an actuator dimensioned for terrestrial locomo-tion (slow with high torque, point B), which operatingrange is shown in figure 4, is not able to continuouslyprovide the high speed required by flight controlbecause its working point A is outside of the con-tinuous workspace of this actuator. Alternatively, anactuator dimensioned for flight control, which wouldhave a different operating range compared to the oneshown in figure 4 (i.e. higher nmax and lower con-tinuous torque), would overheat during ground loco-motion because of the too high torque.

A similar problem is encountered in the skeletalmuscles required for animal locomotion. Skeletalmuscles generate theirmaximumpower and efficiencyin a small range of fibre strain, contraction speeds andload [32]. When birds need less power to fly due tofavourable environmental conditions, instead of redu-cing the contraction speed of the muscles, they alter-nate between gliding and flapping phases in order tomaintain an optimal contraction speed of the muscleduring the flapping phase [11, 33]. Using this strategy,avian muscles work in their optimal range, maximiz-ing efficiency and output power. Similarly, when tran-sitioning between substrates with different physicalproperties [34], animals, like robots have to addressthe limitations of their biological actuators. In

Figure 4.Operating range of a dcmotor suited for terrestriallocomotion (rotational speed n versus torqueM). Thedynamics offlight control and terrestrial locomotion areshown at pointA andB, respectively. The diagonal lineillustrates the dynamics of a dcmotor operating at

=V Vnominal, whereV is the voltage applied to themotor andVnominal is themaximumvoltage recommended by themanufacturer. The continuous torque is themaximumtorque that can be applied continuously by themotorwithoutoverheating and nmax is themaximum speed of themotor.The optimization of the wingerons’ sizemoves the dynamicsof flight control within the continuousworkspace of theactuator dimensioned for terrestrial locomotion.

4

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

robotics, three solutions exist to address the limitedoperating range of dcmotors:

(1)The use of a single oversized dc motor that issimultaneously compatible with both locomotionmodes. A single dc motor that matches bothdynamics has to be simultaneously fast and strongresulting in a heavier solution than the use of twodifferent motors each independently optimized fora single working point. Hence, an integratedapproach with a single locomotor apparatus can beheavier than using two locomotion systems assuggested by the additive strategy. A numericalexample below supports this conclusion.

(2)The use of a transmission with controllable gearratio which would allow a single actuator to matchthe dynamics imposed by flight control andterrestrial locomotion by changing this ratio. Thisapproach involves additional components (i.e.clutch, gear shift actuator and multiple gear trains)that increase both the complexity and the weight ofthe robot [14]. Therefore, the advantage withrespect to an additive strategy is questionable.

(3)The rotational speed of the wingerons during flightcontrol could be reduced until it is compatible withthe operating range of the actuator suited forground locomotion (as shown by the dotted arrowin figure 4). This can be achieved without loss offlightmanoeuvrability if the size of thewingerons isincreased proportionally to the reduction of speed.

In the third solution, the two locomotion modesbecome dynamically compatible within the limitedoperating range of a single actuator; hence they can beseamlessly integrated in a single locomotor systemoptimized for one of the mode of locomotion. In thiscondition, the additional terrestrial competences havea minimal impact on the aerial performances of therobot. Furthermore, because the wingerons arealready part of the flying wing, additional weight anddrag are minimized, hence reducing the impact ofwalking on the aerial COT.

2.1.1. Analysis of solution 1When two locomotion modes have differentdynamics, the use of a single apparatus is notconvenient. To this aim, the dynamic data reported intable 1(a) are considered, and two actuation strategieshave been compared:

• A single actuator dimensioned to match thedynamics of bothworking points (integrated designstrategy).

• Two different actuators for flight control and forgroundwalking (additive design strategy).

Weight is a good metrics to compare the differ-ent strategies since it is a key parameter in thedesign of flying robots. Weight comparison is basedon the fact that biological and artificial effectors(the system that converts an input energy into anoutput mechanical work) have constant power den-sities [35, 36]: their weight increases with ratedpower capabilities. Considering a complete actuator(i.e. electromagnetic effector + reduction stage), it isstill possible to assume that its weight increases withits rated power. According to this assumption, theweights of two strategies can be compared by evalu-ating the maximum power associated with theactuators that are involved.

Dc motors have been selected for actuating thewingerons since they can be easily implemented andbecause their control techniques are well established.Nevertheless, the overall methodology presented herecan be generalized to other actuators (e.g. SMAs,EAPs). The dynamic behaviour of a dcmotor is descri-bed by the following equation, which expresses therotational speed, n (in radians per second), as a func-tion of themotor output torque,M:

= −n M n kM( ) , (7)0

where n0 is the no-load speed and k is the speed/torque gradient. The output power of the motor, P, isgiven by:

= = −P M n M M n kM M( ) ( ) ( ) . (8)0

The maximum power, Pmax, that can be deliveredby the actuator is then evaluated as follows:

= −P M

Mn kM

d ( )

d2 , (9)0

= = =⎜ ⎟⎛⎝

⎞⎠P P M

n

k

n

k2

1

4. (10)max

0 02

The two strategies of actuation are illustrated inthe diagram in figure 5, which shows rotational speed,n, versus torque,M. The points A and B represent theworking points of the wingerons during flight controland walking, respectively. According to equation (7),each dc motor is associated with a line connecting theno-load speed to the stall torque. An actuator is suitedfor a specific working point if it lies on the actuatorline. The additive strategy requires a single actuator foreach locomotion mode: actuator a for flight controland actuator b for terrestrial operations. In the inte-grated design approach a single actuator c can be usedfor both.

Considering at first the strategy of multiple actua-tors, the motor parameters (n0 and k) can be opti-mized in order to minimize the peak power of eachactuator, and therefore their weight. Indeed, consider-ing a generic working point (nx, Mx), the actuatordynamics are described by:

= −n n k M (11)x x x x0,

5

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

and, according to equation (10):

=+( )

Pn k M

k

1

4. (12)x

x x x

xmax,

2

P xmax, has a minimum value for =kxn

Mx

xcorrespond-

ing to:

=+

=( )

Pn n

n Mn M

1

4. (13)x

x x

x xx xmax,

2

In conclusion, power expressions associated withoptimized actuators that work in the dynamic condi-tionsA andB,P amax, ,P bmax, and +P a bmax, , are:

=P n M , (14)a a amax,

=P n M , (15)b b bmax,

= ++P n M n M , (16)a b a a b bmax,

where ni and Mi are respectively the speed and thetorque outputs of a motor i. This example shows that,within the additive design approach, in order tominimize weight, an optimal actuator delivers itsmaximumpower in theworking point of interest.

By adopting an integrated design approach, a sin-gle actuator c is used.With this strategy, the twomotorparameters (n c0, and kc) are constrained by the follow-ing equations:

= −= −

⎧⎨⎩n n k M

n n k M

,

,(17)

a c c a

b c c b

0,

0,

which leads to:

=−−

kn n

M M, (18)c

a b

b a

=−−

nn M n M

M M. (19)c

a b b a

b a0,

Finally, the maximum power associated with theactuator c, Pc

max, can be evaluated according toequation (10) as:

= =−

− −

( )( )( )

Pn

k

n M n M

M M n n

1

4

1

4. (20)c

c

c

a b b a

b a a b

max,0,2

2

Considering the values of torque and speed asso-ciated with the two working points A and B [4], andaccording to equations (14)–(16) and equation (20),the overall power consumption of the two actuationstrategies are summarized in table 1(a).

It can be observed that the power required by thesingle actuator strategy is 37% higher than the overallpower required by multiple actuators. Since power isassumed as an index of weight, the conclusion is thatan integrated design approach with a single actuator isnot optimal in terms of weight. This is due to the factthat the two operational points have very differentdynamics, and therefore a single dc motor that mat-ches both requirements has to be both fast and strong,thus resulting in a heavier solution than two singlemotors each optimized for a single working point. Forexample, considering motors from Maxon (Sachseln,Switzerland), good candidates are the following: RE8with a 64:1 reduction, weighs 7.8 g for flight control;DCX 10 L with a 400:1 reduction, weighs 18.7 g (con-sidering the weight of the available 1024:1 reductionstage) for ground locomotion; RE-max 13 with a 100:1reduction, weighs 31 g with plastic gears and 41 g withmetal gears (considering the available 67:1 reductionstage) for both locomotion modes. In summary, theweight associated with a single actuator strategy is19–57% heavier than the solution with two differentmotors, which is in good agreement with the calcula-tion presented above.

2.1.2. Analysis of solution 3In the authors’ opinions, the best solution for asuccessful integrated design is illustrated in figure 4. Itconsists of translating the working point A to point ′Ain order to match the properties of actuator b. In fact,the working points ′A and B are dynamically wellmatched to actuator b, which can be effectively selectedto power both locomotion modes. By doing this thevelocity of the wingerons decreases, thus the man-oeuvrability of the flying wing may be compromised.The question is then if it is possible to optimize the sizeof the wingerons in order to make the two operationalpoints dynamically compatible, with minimal impacton flight manoeuvrability. To answer this question, letus consider the lift generated by awingeron, L:

ρ=L v A C1

2, (21)L

2wn

where ρ represents the air density, v is the air speed,Awn

is the area of the wingeron and CL is the lift coefficient.The lift coefficient of an airfoil in a steady airflow can beexpressed as a function of its angle of attack,α, as:

α= +C K C , (22)L c L0

where Kc is a parameter that allows to evaluate liftvariation depending on the angle of attack and CL0 is

Figure 5.Working pointA of wingerons during flight,working pointB of wingerons during ground locomotion anddynamic capabilities ofmotors a, b and c. Data experimentallymeasured from thefirst DALERprototype.

6

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

the lift coefficient for α = 0. Since the aim is topreserve flight manoeuvrability, the wingerons mustgenerate the same lift variation (that is ultimatelyresponsible for pitch and roll control) in the same timeinterval, and therefore:

ρ α

ρ

=

= =

L

tv A K

v A K n

d

d

1

2˙

1

2const, (23)

c

c

2wn wn

2wn wn

where nwn (α̇wn) is the rotational speed of thewingeron. This equation demonstrates that the flightmanoeuvrability is not compromised when the win-geron is slowed down if its area is increased accordingto the following relationship:

=A n const. (24)wn wn

Concerning torque requirements, in first approx-imation, wingerons producing the same manoeuvr-ability need the same actuation torque. For example,when the area of a wingeron is doubled its lift doublesas well. Nevertheless, the stroke (angle α) is reducedand consequently the velocity is divided by two. Thesetwo effects compensate each other, resulting in a con-stant torque requirement for each wingeron thatensures the same level of manoeuvrability. Thus, asshown by the dashed arrow in figure 4, the workingpoint A can be translated vertically until it crosses in ′Aby simply increasing its width L (see figure 3). Torqueand velocity of the working points ′A and B, ′Ma , ′na ,Mb and nb are reported in table 1(b) as well as the asso-ciated power requirements.

2.2. FoldablewingsIn the previous section 2.1, a single locomotorapparatus suited for both flight control and walkinghas been described. However, the challenge of adapt-ing a morphology mainly optimized for flight to onemore suited for terrestrial locomotion is discussedhere. For example, flight and walking require differentpositions of the centre of mass (CM) of the robot. Forground locomotion the CM of the robot must be closeto the centre of rotation of the wingeron to avoid thewingerons slipping on the ground [4] (seefigure 7(B)). In flight the CMmust be instead far fromthe centre of rotation of the wingeron in order tocreate torques required to control the flight (see

figure 7(A)). Furthermore, for flight stability reasons,the CM must be in front of the aerodynamic pressurecentre of thewing. In this prototype, the use of foldablewings to adapt the morphology of the robot to eitherflight or ground locomotion is proposed. Withreference to figure 7, when the wings are deployed, therobot has aflight-adaptedmorphology:

• The lift is augmented due to the largewingspan.

• The distance between the axis of rotation of thewingerons and the CM is increased, enhancingflightmanoeuvrability.

When the wings are folded, the morphology of therobot becomes suitable for ground locomotion:

• A short wingspan improves the robotʼs agility incluttered terrestrial environments.

• The axis of rotation of the wingerons is closer to theCM, thus enhancing the grip between the winger-ons and the ground.

The use of these deployable wings allows adapta-tion of the morphology of the robot and therefore itsatisfies the requirements on the relative positioning ofthe CM and of the wingeron axis of rotation for thetwomodes of locomotion.

2.3.Mechanical designThis subsection presents the mechanical design of thesecond DALER prototype, which is based on thetheoretical considerations presented above. As illu-strated in figure 6 the DALER comprises five mainbody sections: a central frame housing the propeller,electronics and battery, two foldable sections and twowingerons forflight control and ground locomotion.

The robot has a wingspan of 72 cm and a weight of393 g. The frame of the DALER is designed in order tominimize weight, while providing enough stiffness forefficient flight. To this aim, the frame has a centralbody and multiple ribs connected together by carbonfiber spars. This frame is covered with Icarex, a light-weight polyester fabric, inextensible and resistanttowear.

Table 1.Dynamic requirements.

7

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

Flight control and ground locomotion rely on asingle locomotor system composed of two indepen-dent wingerons and their actuators. The dcmotors arehoused in the external ribs (rib B) and are coupledwith the wingerons using a synchronous belt. For con-trol purposes, the angular position of the wingerons ismeasured by Hall effect sensors, which detect theorientation of small magnets mounted on the winger-ons’ axis of rotation.

The two foldable sections of the robot are equip-ped with an articulated frame controlled by a single dcmotor. The foldable frame is composed of two carbonspars that are serially connected together and also toribs A and B through revolute joints. Each foldablesection is controlled by two separate cables, one toopen and the other to close the wings. The two cablesare connected to a double arm that is directly con-trolled by a dc motor. The arm is used instead of apulley because it acts as a self-locking mechanism thatprevents unwanted rotation of the motor when thewings are fully deployed or collapsed. The cableresponsible for wing folding is connected to a springand when the wing is deployed, the spring is pulledand the covering fabric is tensioned. This is of para-mount importance in order to pre-load the fabric toavoid fluttering, thus maximizing flight efficiency.Furthermore, the spring can absorb energy, limitingdamage to the wings in case of a frontal collision. If awing collapses due to a collision, the spring is stret-ched. Each spring can absorb a maximum energy of12.5 J which corresponds to a collision at 8.4 m s−1.Wingmorphing can be performed in 1–2 s. The lengthof each foldable section can be reduced from 17 to6 cm (65%); this corresponds to a 30% reduction inthe overall wingspan.

3. Results

This section presents the analysis of the performanceof the robot on the ground as well as in the air, theanalysis of the level of mass integration of the two

modes of locomotion and the analysis of the versatilityand complexity of the robot.

3.1. Ground locomotion analysisThis subsection presents the measurements of theCOT of the robot on the ground. In figure 8(a) theCOT of the robot on a wooden floor as a function ofthe speed can be seen for different openings of thewings. For each measurement two complete revolu-tions of the wingerons were performed and thefollowing parameters weremeasured:

• The time (s).

• The travelled distance (m).

• The current in the dcmotors (A).

• The voltage applied to the dcmotors (V).

A PID controller running at 100 Hz controls therotational speed of the wingerons. At each update ofthe controller (every 10 ms) the current in bothmotors is measured along with their voltage. The elec-trical power is computed at each step and then low-pass filtered. At the end of the run, the mean power iscomputed and multiplied by the time of the run inorder tofind the total energy used for the run (in joules(J)). Finally, the COT (in joules per kilogramme andper metre (J Kg−1 m−1)) is computed by dividing theenergy by the mass of the robot and by the travelleddistance (speed is computed as the ratio between thetravelled distance and the time of the run).

Concerning the experiments presented infigure 8(a), the wingerons were set to seven differentrevolution speeds and each experiment was repeatedfive times. The smaller dots represent all the measure-ments and the larger dots represent the mean COT atthe mean speed of the robot for the seven imposedrotational speeds of the wingerons. The three differentmarkers’ shapes (diamonds, triangles and squares)represent the configurations with the wings open,half-closed and fully closed respectively. With

Figure 6. 3Dmodel of theDALER. (A) The left wing is deployed, while the right one is completely folded. (B) Zoomon the centralframewhich shows themotor for wing deployment. (C) Zoomon thewingeronʼs drivemechanism.

8

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

reference to figures 6 and 7 the width of the foldablesection for the three different openings was set to 17,11 and 6 cm respectively and creating a distancebetween the CM and the axis of rotation of the winger-ons ( −d daxis CM) of 13, 11 and 9 cm.

It can be seen that the three lines show the sametrend; a high COT at low speeds, then a minimal COTat around 75% of the maximum rotational speed ofthe wingerons, and again a high COT at high rota-tional speeds of the wingerons. At high rotationalspeeds there is a sharp increase of COT due to the factthat the wingerons slip more on the floor. The COT ofthe robot when the wings are closed (i.e. folded) ismuch lower than theCOT for openwings, especially athigh speeds. Furthermore, the speed of the robot ismuch higher for the same rotational speed of the win-gerons (35% increases). These two arguments clearlydemonstrate the need and the advantage of having anadaptivemorphology (i.e. foldable wings).

Another interesting observation is that at speedslower than 0.04 m s−1 the configuration with half-closed wings has a lower COT than the one with fullyclosed wings. This indicates that the morphology ofthe robot (i.e. wings opening) should be continuouslyadapted as a function of the speed to minimize theCOT. This can be explained by the fact that, on onehand, when the wings are folded the CM of the plat-form has to rise higher at each step than when thewings are open, while on the other hand, at low speedsthere is static friction between the ground and thewin-gerons while at very high speeds the friction is mainlydynamic (i.e. lower), therefore the wings start to slipmore. Thus, by adapting the morphology of the robotthe optimal trade-off between CM lifting and winger-ons slippage thatminimizes theCOT can be found.

Figure 8(b) shows the same experiment but with asmall wheel at the tip of the robot that is used to reducethe friction between the central frame of the robot andthe ground. It can be seen that if this friction is reducedto almost zero the adaptive morphology is requiredless than previously and also that the robot travelsmuch faster for the same rotational speed of the win-gerons. In outdoor, field applications the friction ofthe central body cannot be so drastically reduced, thusthis situation will not happen, however it demon-strates the importance of maximizing the friction onthe wingerons andminimizing the friction on the cen-tral frame. From these experiments it can be seen thatCOT values between 15 and 20 J Kg−1 m−1 at speeds ofmore than 0.1 m s−1 can be reached, similar to smallrunning animals [37].

Table 2(a) summarizes the performance of theDALER prototype presented in this paper. Differentexperiments have been performed in order to evaluatethe capabilities of the robot on the ground. The max-imum gap that the robot can overcome repeatedly(100% success over five trials) is 9 cm, which corre-sponds to 0.25 body-length (BL); above this distancethe robot gets stuck or falls in the gap. The maximum

step that the robot can climb is 6 cm, which corre-sponds to 1 body-height (BH). Themaximumupwardslope, on a wooden floor, that the robot can walk on is9°. The maximum forward speed measured on a flatwooden floor is 7 cm s−1, which is 0.2 BL s−1 and themaximum rotational speed of the robot (on spot) is24° s−1 (15 s for one complete revolution). The auton-omy of the robot is very much dependent on the typeof terrain; as demonstrated above, the COT changesdrastically with the variation of the friction betweenthe robot and the ground.On a flat wooden surface themaximum autonomy has been measured at close to60 min and in rough terrains the robot can walk forabout 30 minwith a full battery (3 cells LiPo, 1 Ah).

3.2. Flight analysisThe drag force during flight is the same as on a wingcapable of only flying since no additional appendiceshave been added for ground locomotion. When thefabric that covers the wings is properly stretched by thedeployable mechanism, the lift produced by the wingsis similar to the one produced by a rigid wing. The onlydifference to a regular flying wing is the weight addedfor the deployable wings, which represents less than10.7% of the weight of the platform, 42 g over 393 g(see table 2(b)). Thus, for retaining the same wingloading (i.e. the same total weight) as on a regularwing; the weight of the battery must be reduced by theweight of the deployable wings (42 g). The DALER hasa 1 Ah (3 s) battery which weighs 90 g, which meansthat a 1.5 Ah (approx. 130 g) battery can be used in arobot without deployable wings, resulting in a differ-ence in battery capacity equal to 33%.

The minimum flight speed of the robot, beforestalling, has been measured at 6 m s−1 and thereforethe robot can easily be launched by hand. The max-imum flight speed of the robot has been measuredabove 20 m s−1. The autonomy of the robot at cruisespeed (approx. 12 m s−1) has been measured atbetween 15 and 20 min with a 1 Ah battery. The

Figure 7. (A) Robot seen from abovewith deployedwings.(B) Robot seen from abovewith foldedwings. (C) Robot seenfrom the sidewalking on a flat terrain.

9

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

maximum constant pitch and roll rates, measured byperforming an inside loop manoeuvre and a full roll

manoeuvre, are 120° s−1 and 180° s−1 respectively.The robot shows excellent flight performances thatcan be seen in the accompanying video.

3.3. Integrated design analysisThe working area of the wingerons’ motors has beenmeasured during flight. For a standard flight, the speedof the wingerons varies between 0 and 25 rpm and thetorque varies between 100 and 150 Nmm. This torqueis mainly caused by friction in the belt transmissionand does not change with the speed of the robot in theair. The speed during ground locomotion variesbetween 0 and 15 rpm (which is the speed thatminimizes the COT) and the torque varies between650 and 850 Nmm depending on wing opening, speedof the robot and friction with the ground. Figure 9shows the two designed working points, ′A and B, forflight and ground locomotion respectively and the twomeasured working areas of the motor during flightand ground locomotion are also shown. The two

measured areas are below the line of the actuator andthis demonstrates that the samemotor can be used forthe two modes of locomotion and that these modes oflocomotion are dynamically compatible.

Table 2(b) provides the weight distribution of theDALER prototype. All the components are sorted intothree categories; the components shared for bothmodes of locomotion, the components used only forground locomotion and those used only for flight.Shared components are: the frame which is composedof all the mechanical parts which are needed to build afixed wing without deployable wings (central body,ribs, carbon rods, wingerons and fabric), the winger-ons’ control which includes the actuation and thetransmission used to control the wingerons, the elec-tronics which includes the autopilot board and the dcmotor board, and finally the battery. These sharedcomponents weigh a total of 301 g. The componentswhich are added only for ground locomotion are partof the foldable wings, which includes all the joints,bearings and springs which are added to enable thefolding of the wings. They together weigh 42 g, which

Figure 8.Cost of transport versus speed of the robot on the ground.

Table 2.Performance andweight analyses.

10

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

brings the total for the ground locomotion to 343 g.The parts used only for the flight are the propulsion sys-tem (brushless motor, propeller and speed controller)and themotor holding. These components weigh 50 g,which brings the totalmass for the flightmode of loco-motion to 351 g. The total mass of the robotweigh 393 g.

According to themass integrationmetrics recentlyintroduced in [20], which is ‘a measure of the percen-tage of the total integrated robot mass required tocombine the modes without any integration betweenthem’, the mass integration of the DALER prototype is1.77. This results shows that 77%of themass is used byboth modes of locomotion; the robot is 43.5% lighterthan it would bewith an additive design approach. Theother multi-modal robot evaluated with this metric isthe MultiMo-Bat [20], which scores 1.69. The resultachieved by the DALER is even more remarkable con-sidering that it combines active aerial and terrestriallocomotion, while in theMultiMo-Bat flight is passive(gliding), and therefore it does not require actuators.In the DALER, such high level of integration can beenachieved since the same actuators are used for bothflight control andwalking on the ground.

3.4.Multi-modal locomotion analysisThe versatility and complexity metrics defined in [5]have been used to evaluate the DALER. The metricsused to evaluate the versatility of a mobile robot isdefined as ‘an extension of mobility that includesoperation in and transition among multiple domains’,and the complexitymetrics is defined as the number ofactuators multiplied by the number of degrees offreedom of the robot. The objectiveness of theversatility metrics is questionable since arbitrarygrades are given by the person doing the evaluation ofa given robot. However, to the best of the authors’knowledge, there is no objective metrics to evaluatethe versatility of amobile robot.

Based on the rules defined in [5], the versatility ofthe DALER prototype has been evaluated and com-pared to the ones of the first DALER prototype [4], ofBOLT [9] and of MMALV [15]. These robots havebeen chosen since they are the only robots with wingsthat can also move on the ground and thus they can be

compared to our prototype. Figure 10(a) shows theresults of these evaluations (refer to [5]). The robotsare graded for their capabilities in the aerial and terres-trial domains and for their capability of transitionbetween these domains; they obtain a grade between 0(cannot do it) and 2 (does it well) for each ‘mobility’.The mobilities in the terrestrial domain are furthercategorized between the manoeuvrability and thecapacity to overcome different types of obstacles.Some of these grades are based on measured valuesgiven in table 2(a) and some are evaluated by theauthors. The versatility of the DALER is 0.183 which ishigher than the other prototypes; 0.141 for the firstDALER, 0.128 for MMALV and 0.118 for BOLT. Thefirst DALER prototype obtains a lower versatilitybecause it has lower flight performances compared tothe newDALER prototype. Furthermore, according tofigure 4 of [5], which shows the versatility of manymobile robots, none of them has a versatility higherthan 0.16. This result shows that our prototype has avery high versatility compared to state of the artmobile robots.

The complexity of these robots has also been eval-uated and the results are given in figure 10(b). Thecomplexity of the DALER is 16; it has four actuators(the two dc motors for the wingerons, the brushlessmotor for the propeller and the dc motor for theadaptive morphology) and has as many degrees offreedom. The previous DALER prototype had a com-plexity of 49 (seven actuators and seven DOFs), BOLThas a complexity of only 12 (three actuators and fourDOFs; its wings and legs are power with a single actua-tor) and MMALV has a complexity of 36 (six actua-tors and six DOFs). Figure 10(c) gives the versatilityversus the complexity of this four robots. It can beseen that the new DALER prototype has a higher ver-satility and a lower complexity than the first DALERprototype and than MMALV. BOLT has a lower com-plexity but also has a lower versatility compared to theothers.

4.Generalized designmethod

This section proposes a design method extracted fromthe example presented in this paper. For the design of arobot capable of multi-modal locomotion two differ-ent strategies can be adopted: an integrated strategy oran additive strategy. Nature provides many examplesof animals using one of these two strategies [6].However, when it comes to robotics an integratedstrategy will allow a higher reduction of total weightand complexity, and will consequently lead to betterperformances in the different modes of locomotion.Themain idea behind using an integrated strategy (i.e.using a multi-purpose locomotor apparatus) is tomaximize the versatility of the robot by minimizingthe complexity of the mechanical structure and thenumber of actuators required by each additional

Figure 9.Designedworking points ′A ofwingerons duringflight andB during ground locomotion andmeasuredwork-ing areas of themotor during the twomodes of locomotion.

11

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

locomotion mode. However, effective implementa-tion of an integrated approach is challenging; while inan additive strategy each apparatus can be separatelyoptimized for a specific locomotion mode, a dual useapparatus must simultaneously fulfil the differentdynamics imposed by each locomotion mode. Thesteps to follow for the design of an effective multi-modal robot based on the proposed methodology aregiven below:

(1)Define themission requirements of the robot.

(2)Define what is the principal mode of locomotionand choose an appropriate basic design.

(3)Define how the existing structure and actuators ofthis basic design could be reused for the secondmode of locomotion. If necessary modify theplacement of the locomotor apparatus to fit theneeds of the twomodes of locomotion.

(4)Optimize the shape and dimensions of the locomo-tor apparatus; the two modes of locomotion musthave compatible dynamics.

(5)Use adaptive morphology if the two modes oflocomotion have different constraints on the place-ment of theCM.

(6)Take advantage, if possible, of the adaptive mor-phology to continuously adapt the shape of therobot in order to optimize the performance of therobot for different locomotion tasks.

The first three steps were presented in greaterdetail in [4] while the steps 4–6 are covered by thispaper. The six steps were followed for the design of theprototype presented in this paper and the authorsbelieve that they can be applied to the design of othermulti-modal locomotion robots.

5.Discussion

Nature has evolved multiple strategies to implementmulti-modal locomotion. These strategies can besuccessfully applied to the development of robots withlocomotion capabilities in multiple environmentswithminimal compromises.

A comparison can be made between animals thatexploit an additive strategy with multiple single-uselocomotor apparatus, or an integrated strategy with asingle apparatus with competences in multiple sub-strates. For robots, it is shown that the latter strategy isconvenient if the two locomotion modes imposedynamics that are compatible with the operating rangeof the actuator used in the single locomotor apparatus.In this condition, secondary locomotionmodes can beadded with minor impact on the primary locomotionmode. In addition, the overall structural mass andcomplexity of the robot are minimized as well. Forexample, in the proposed prototype, ground locomo-tion can be performed with wingerons, introducingminimal losses in flight manoeuvrability and minimalincrease in robotweight.

Figure 10.Versatility versus complexity ofmulti-modal robots computedwith themetrics defined in [5]. Only robots capable ofwinged-flight and ground locomotion have been compared in this study.

12

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

Furthermore, many animals exploit adaptivemorphologies in order to accommodate the require-ments imposed by different modes of locomotion. Wehave shown that also this strategy is a good solution tominimize trade-offs. Indeed, a robot optimized forflight can improve its terrestrial capabilities with fold-able wings.

In addition, multi-modal locomotion in roboticscould provide new perspectives for understandingmulti-modal locomotion in animals. For example, theconcept of dynamically compatible locomotionmodescould explain why the Desmodus rotundus does notapparently show compromises caused by terrestrialcompetences in a body optimized for flight. It is possi-ble to speculate that this bat evolved a unique runninggait compatible with a locomotor apparatus evolvedfor flight [29]. The concept of adaptive morphology issupported by biological observations of animals thatactively control their body to improve locomotionperformance inmultiple environments [23].

Our aim in the near future is to deepen the under-standing of the biological principle of multi-modallocomotion, also integrating ad hoc control techni-ques and multi-purpose perception capabilities intothe robot. Concerning the locomotion flexibility of theDALER, we would like to add hovering capabilities,which are needed for vertical take-off and flight incluttered environments. Furthermore, we are cur-rently developing a mass-and-power model of thisfuture version of the DALER prototype in order to beable to optimize its geometrical parameters (e.g. wing-span, chord, etc) to differentmission scenarios.

Acknowledgments

This work was supported by the Swiss NationalScience Foundation through the National Centre ofCompetence in Research Robotics.

References

[1] Tadokoro S 2009Rescue Robotics: DDTProject onRobots andSystems forUrban Search and Rescue (Berlin: Springer)

[2] Nagatani K et al 2013 Emergency response to the nuclearaccident at the FukushimaDaiichi nuclear power plants usingmobile rescue robots J. Field Robot. 30 44–63

[3] Siddall R andKovačM2014 Launching the aquamav:bioinspired design for aerial–aquatic robotic platformsBioinsp. Biomim. 9 031001

[4] Daler L, Lecoeur J, Hählen PB and FloreanoD 2013A flyingrobot with adaptivemorphology formulti-modal locomotionIEEE/RSJ Inter. Conf. Intelligent Robots and Systems (IROS)1361–6

[5] Nie C, Pacheco-CorchoX and SpenkoM2013Robots on themove: versatility and complexity inmobile robot locomotionIEEERobot. Autom.Mag. 20 72–82

[6] Lock R J, Burgess S C andVaidyanathan R2014Multi-modallocomotion: from animal to applicationBioinsp. Biomim. 9011001

[7] Kossett A, DʼSaR, Purvey J and PapanikolopoulosN 2010Design of an improved land/airminiature robot IEEE Int.Conf. onRobotics andAutomation (ICRA) (Piscataway,NJ:IEEE) pp 632–7

[8] Kovac M, Hraiz W, Fauria O, Zufferey J-C and Floreano D2011 The epfl jumpglider: a hybrid jumping and glidingrobot with rigid or folding wings IEEE Int. Conf. on Roboticsand Biomimetics (ROBIO) (Piscataway, NJ: IEEE)pp 1503–8

[9] PetersonK and Fearing R S 2011 Experimental dynamics ofwing assisted running for a bipedal ornithopter IEEE/RSJ Int.Conf. on Intelligent Robots and Systems (IROS) (Piscataway,NJ:IEEE) pp 5080–6

[10] Li F, LiuW, FuX, Bonsignori G, ScarfoglieroU,Stefanini C andDario P 2012 Jumping like an insect: designand dynamic optimization of a jumpingmini robot based onbio-mimetic inspirationMechatronics 22 167–76

[11] Lovvorn J R, CrollDA andLigginsGA1999Mechanical versusphysiological determinants of swimming speeds in divingBrunnichʼs guillemots J. Exp. Biol. 202 1741–52

[12] RiskinDK, Parsons S, SchuttWA,Carter GG andHermanson JW2006Terrestrial locomotion of theNewZealand short-tailed batMystacina tuberculata and thecommon vampire batDesmodus rotundus J. Exp. Biol. 2091725–36

[13] Elliott K H, Ricklefs R E, Gaston A J, Hatch S A,Speakman J R and Davoren G K 2013 High flight costs, butlow dive costs, in auks support the biomechanical hypoth-esis for flightlessness in penguins Proc. Natl Acad. Sci. 1109380–4

[14] Kossett A and PapanikolopoulosN 2011A robustminiaturerobot design for land/air hybrid locomotion IEEE Int. Conf. onRobotics and Automation (ICRA) (Piscataway,NJ: IEEE)pp 4595–600

[15] BachmannR J, Boria F J, VaidyanathanR, Ifju PG andQuinn RD2009A biologically inspiredmicro-vehicle capableof aerial and terrestrial locomotionMech.Mach. Theory 44513–26

[16] ItasseM,Moschetta J-M, AmehoY andCarr R 2011Equilibrium transition study for a hybridmav Int. J.Micro AirVeh. 3 229–46

[17] Kalantari A and SpenkoM2013Design and experimentalvalidation of hytaq, a hybrid terrestrial and aerial quadrotorIEEE Int. Conf. on Robotics and Automation (ICRA) (Piscat-away,NJ: IEEE) pp 4445–50

[18] Briod A,Kornatowski PM, Zufferey J-C and FloreanoD 2014A collision resilient flying robot J. Field Robot. 31 469–509

[19] GastonA J and Jones I L 1998TheAuks: Alcidae (Oxford:OxfordUniversity Press)

[20] WoodwardMAand SittiM 2014Multimo-bat: a biologicallyinspired integrated jumping-gliding robot Int. J. Robot. Res. 331511–29

[21] Socha J J 2002Kinematics: gliding flight in the paradise treesnakeNature 418 603–4

[22] Socha J J, OʼDempsey T and LaBarberaM2005A3d kinematicanalysis of gliding in aflying snake, chrysopelea paradisi J. Exp.Biol. 208 1817–33

[23] Johansson LC andAldrin B SW2002Kinematics of divingatlantic puffins (Fratercula arctica l.): evidence for an activeupstroke J. Exp. Biol. 205 371–8

[24] Ijspeert A J, Crespi A, RyczkoD andCabelguen J-M2007 Fromswimming towalkingwith a salamander robot driven by aspinal cordmodel Science 315 1416–20

[25] García-ParísM andDeban SM1995Anovel antipredatormechanism in salamanders: rolling escape inHydromantesplatycephalus J. Herpetology 29 149–51

[26] NorbergUMandRayner JM1987 Ecologicalmorphologyandflight in bats (mammalia; chiroptera): wing adaptations,flight performance, foraging strategy and echolocationPhil.Trans. R. Soc.B 335–427

[27] RiskinDK andHermanson JW2005Biomechanics:independent evolution of running in vampire batsNature 434292–292

[28] SchuttW,Altenbach J S, ChangY, CullinaneDM,Hermanson JW,Muradali F andBertram J 1997The dynamicsof flight-initiating jumps in the common vampire batDesmo-dus rotundus J. Exp. Biol. 200 3003–12

13

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al

[29] RiskinDK, Bertram J E andHermanson JW2005Testing thehindlimb-strength hypothesis: non-aerial locomotion by chir-optera is not constrained by the dimensions of the femur ortibia J. Exp. Biol. 208 1309–19

[30] Quinn RD,NelsonGM, BachmannR J, KingsleyDA,Offi J T,Allen T J andRitzmannRE 2003 Parallel complementarystrategies for implementing biological principles intomobilerobots Int. J. Robot. Res. 22 169

[31] Etkin B andReid LD 1996Dynamics of Flight: Stability andControl (NewYork:Wiley)

[32] Rome L, FunkeR, Alexander R, LutzG, AldridgeH,Scott F and FreadmanM1988Why animals have differentmuscle fibre typesNature 335 824–7

[33] Pennycuick C 1991 Adapting skeletal muscle tobe efficient Efficiency and Economy in Animal

Physiology (Cambridge: Cambridge University Press)pp 33–42

[34] Lock R J, VaidyanathanR andBurgess S C 2014 Impact ofmarine locomotion constraints on a bio-inspired aerial-aquatic wing: experimental performance verification J.Mech.Robot. 6 011001

[35] Hunter IW and Lafontaine S 1992A comparison ofmusclewith artificial actuators Solid-State Sensor andActuatorWork-shop 1992. 5th Technical Digest., IEEE (Piscataway,NJ: IEEE)pp 178–85

[36] Huber J, FleckN andAshbyM1997The selection ofmechanical actuators based on performance indices Proc. R.Soc.A 453 2185–205

[37] Tucker VA 1975The energetic cost ofmoving aboutAm. Sci.63 413–9

14

Bioinspir. Biomim. 10 (2015) 016005 LDaler et al