Vous êtes sur la page 1sur 10

Mechatronics 22 (2012) 167176

Contents lists available at SciVerse ScienceDirect

Mechatronics
journal homepage: www.elsevier.com/locate/mechatronics

Jumping like an insect: Design and dynamic optimization of a jumping mini


robot based on bio-mimetic inspiration
Fei Li a, Weiting Liu a,, Xin Fu a, Gabriella Bonsignori b, Umberto Scarfogliero b, Cesare Stefanini b,
Paolo Dario b
a
b

The State Key Laboratory of Fluid Power Transmission and Control, Zhejiang University, Hangzhou 310027, China
CRIM Lab, Polo SantAnna Valdera, Pontedera, Pisa 56025, Italy

a r t i c l e

i n f o

Article history:
Received 9 November 2010
Accepted 6 January 2012
Available online 30 January 2012
Keywords:
Bionics
Jumping robot
Articial saltatorial leg
Jumping dynamics

a b s t r a c t
This paper presents a bio-inspired design of a jumping mini robot including the theoretical analysis on
jumping dynamics based on a simplied biological model, the dynamically optimized saltatorial leg
design, the overall design of the jumping robot prototype and, as a part of the bio-mimetic research,
and the measuring and comparing of the jumping characteristics between the robot and animal. The articial saltatorial leg is designed to imitate the characteristics of a real jumping insect, kinematically and
dynamically, and proposed to reduce the contact force at tarsusground interface during jumping acceleration thus optimizes the jumping motion by minimizing the risk of both leg ruptures and tarsus slippage. Then by means of high speed camera experiment, the jumping characteristics of the theoretical
jumping model, the jumping insect leafhopper and the robot are compared so as to show the dynamic
similarity and optimization results among them. The nal energy integrated jumping robot prototype
is able to accomplish a movement of continuous jumping, of which a single jumping reaches 100 mm
high and 200 mm long, about twice and four times of its body length respectively.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
As a key issue in robotics, locomotion mode is usually the rst
decision researchers made while designing a new robot. As we
know, a proper locomotion mode may result in the improvement
on both gait efciency and travelling stability in exploration, not
only for robot design, but also in the natural world [1]. Along with
the decreasing of their body scale, jumping is much more preferred
by those small insects as their primary locomotion mode, especially when moving in long distance on a tough terrain, rather than
only being an efcient way to escape from emergency.
For those millimeter-to-centimeter-sized insects, small body
size offers them with born advantages of jumping such as less
ground contact and higher energy efciency. Such a scale effect
can be demonstrated and explained by the following equation from
previous biology research [2]:

animal muscles, the peak generated force should be proportional to


the transverse section of the muscles represented by l2. On the other
hand, considering the geometrical similarity, the inertia force
depending on the body weight is usually proportional to the body
volume represented by l3. The ratio between peak generated force
and the inertia force increases along with the scaling down of insect
size, which means that small animals show a born advantage in
jumping respect to those animals in large size [1,3]. Moreover,
jumping helps small animals launch themselves into ight, reduce
their ground contact during moving, and achieve an energetic
choice of the most efcient gait. Such a change in gait was also observed for animals with different size who tend to switch from walk
to run (or trot for quadruped) at a same Froude Number (Fr  0.4)
[1,4], which is dened as following and can be considered as the ratio between centrifugal and gravitational force acting on a stance
leg:

F max l
/ 3
mg
l

where Fmax is the maximum exerted force by animal muscles, m is


the mass of the animal, g is the gravity acceleration, and l is characteristic length of the animal. As to a given maximum yield stress of
Corresponding author.
E-mail address: liuwt@zju.edu.cn (W. Liu).
0957-4158/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechatronics.2012.01.001

Fr

v2
gl

where v represents the moving velocity of the animal. Theoretically,


ground contact is lost when Fr > 1 at which the animal is launched
into a ight phase. Apparently, compared with their large counterparts, the Froude Number is easier to approach 1 for those animals
with smaller characteristic length, which also explains why jumping locomotion is more common in the insect world (Fig. 1).

168

F. Li et al. / Mechatronics 22 (2012) 167176

Fig. 1. Jumping animals ranging with body size: (a) kangaroo; (b) rabbit; (c) frog; (d) cricket; (e) leafhopper; (f) ea.

Given inspiration from above biological observation results,


jumping is then regarded as not only a supplementary of traditional wheeled machines for roboticists, but also an evolution of
legged robots. Like what happened in the natural world, jumping
is more than a method that helps robot overcome obstacles higher
than themselves; considering the increasing of terrain roughness
resulting from their small body size, it also offers a protable locomotion solution for robots under tens of millimeters because of
improvement of their moving ability and environmental suitability
in exploring [5]. During the past decades, as a promising solution in
locomotion, possible designs based on jumping movement are applied by a lot of roboticists despite a controlling complicity. For
example, one leg hopping robots based on multi-joint design are
developed in order to better investigate the complex controlling
strategy of jumping movement [6,7]. Conversely, despite a nonoptimal jumping dynamics, integration of a simple catapult mechanism is applied by both the traditional wheeled robots [810] and
novel leg-wheeled ones [11,12], helps researchers avoid the
mechanical and controlling complication and allows rescue or
scout robots over obstacles several times their own height, improves their practicability and passing ability in complex working
conditions. On the other hand, biological investigation helps engineers better understand the intrinsic principle of insect jumping
such as their skeleton structures [13,14], jumping dynamics and
kinematics [1517]. Therefore, compared with simple catapult
mechanism, jumping robots with bio-mimetic saltatorial leg design usually show a better dynamic characteristic during acceleration besides a good jumping performance [2,1820]. Furthermore,
to benet from the natural scale effects, the concept of producing
micro jumping robot based on MEMS technology is also developed
and shows a promising prospect [21].
Based on above lessons, the GRILLO III robot, a millimeter-sized
jumping robot prototype with onboard power supply integrated
and designed as a mobile roboticized platform for environmental
monitoring, rescue or exploration in unstructured environments,
is presented in this paper. By theoretically analyzing and imitating
the catapult jumping adopted by insects (e.g. leafhopper Cicadella
Viridis), this jumping robot offers a lot of advantages such as its better jumping dynamics, improved moving ability and practicability.
2. Jumping dynamic analysis
Countermovement and catapult are two major jumping styles
adopted by vertebrates and insects respectively. The former one

is accomplished by directly contracting the muscles [22], while


the latter by storing energy in advance and then suddenly releasing
it [16]. Compared with direct muscle contraction, catapult mechanism is able to generate ground force multiples of body mass and
enable small insects to overcome obstacles even hundreds of its
characteristic length [23]. As an example, impressive jumping ability is exhibited with leafhopper insects. By extending the tarsus
along a straight line during leg elongation, the extensor of a
19 mg weight and 3.5 mm long Aphrodes is able to release 77 lJ
energy in 2.75 ms which leads to a take-off velocity of about
2.9 m/s [16]. Considering the mechanical simplicity and the high
efciency, such a catapult design is more applicable for the jumping robot prototype around tens of millimeters in size.
Indeed, due to the strategy simplication, extension of the two
saltatorial legs of a leafhopper is usually spontaneous and the
jumping angle is adjusted by the front limbs [16]. To better understand the intrinsic principle of its jumping dynamics, a simplied
theoretical model of a catapult jumping mechanism derived from
the anatomy research of leafhopper, who obtains impressive jumping ability by contracting metathoracic muscles inserted between
body and trochanter in the saltatorial leg [24,25], is rstly built
and shown in Fig. 2. As a mechanical model, the multi-sectional saltatorial leg of the insect (usually including body, coax, trochanter,
femur and tibia) is simplied to a two-bar structure composed of
only body (as a particle O with mass m xed at the end of femur
bar), femur and tibia (as massless ideal rods OA and AB with length
lf and lt respectively); natural tendon is simplied into a spring with
stiffness K and length Ls connecting between body and femur as an
energy storing element [24]. Particularly, the arms of the spring on
both femur and body are set to be ls for computation simplication;
jumping motion is restrict along y-direction in order to ignore the
inuence of jumping angle thus only vertical ground force Fy is considered; angle a and b represent the squat of femur and tibia respectively while the derivation of them represents the rotating velocity
of each rod; O is the position of the body particle and the derivation
of it represents the jumping velocity; at last, due to its symmetrical
property, calculation is only carried on the half part.
Analysis based on such a mechanical model is aimed to nd a
guiding principle for saltatorial leg design, especially a proper tibia/femur ratio at which the dynamic characteristic can be improved. Therefore, the inuence of the tibia/femur ratio Rtf on
jumping dynamics is investigated by assuming a constant overall
length of femur and tibia. Such a hypothesis can be described as
following:

169

F. Li et al. / Mechatronics 22 (2012) 167176

lf lt CONSTANT
Rtf

lt
2 1; 1
lf

Then, the kinematic restriction described by Eq. (4) ensures a


vertical jumping motion in order to ignore the inuence of jumping
angle. In a meanwhile, the relationship between a and b can be
gained as Eqs. (5) and (6) by differentiating Eq. (4) with respect
to time.

lf cos a  lt cos b 0

lf sin a
b_
 a_
lt sin b

5
2

2
l cos a 2 lf sin a  cos b 2
lf sin a  a
f
 a_
 a_  2
b
3
lt sin b
lt sin b
sin b
lt

Fig. 2. Simplied theoretical model of a catapult jumping mechanism derived from


the anatomy research of leafhopper [24].

Furthermore, the position, velocity and acceleration of body


particle O can also be deduced from the geometry relationship as
Eqs. (7) and (9):

Fig. 3. Calculation results of the theoretical jumping model.

170

F. Li et al. / Mechatronics 22 (2012) 167176

o lt sin b  lf sin a

o_ lt b_ cos b  lf a_ cos a

cos b  lt b_ 2 sin b  lf a
cos a lf a_ 2 sin a
lt b
o

Considering the energy conservation of the whole system during tendon (spring) contraction, the energy increasing of the body
particle including the kinetic energy Ek and the potential energy Ep
should always be equal to the energy releasing Es from the tendon
at each moment, which can be described as:

E_ s E_ k E_ p 0

10

As assumed the same arm distance of the spring (ls), original


length of the spring turns to be 0. Therefore, the equation variables
in Eq. (10) can be deduced as



d 1 2
2
KLs Kls a_ cos a
E_ s
dt 2

11



d 1

mo_ 2 mo_ o
E_ k
dt 2

12

d
E_ p mgo mg o_
dt

13

so that Eq. (10) becomes


2

mg o_ 0
Kls a_ cos a mo_ o

14

By substituting Eqs. (4)(9) in Eq. (14), the differential equations for calculating jumping dynamics can be obtained as:
2

Kls cos a sin b


a
 mg  Ua_ 2
C lf sina  b
1

!
15

where intermediate variables U and U satises

C lf
U lf



sin a cos b
 cos a
sin b

!
2
2
cos a cos b 1 sin a cos2 b 1 sin a

sin a

3
sin b
Rtf
Rtf sin b
sin b

16

Considering the femur rotating acceleration as a function of its


kinematic information, Eqs. (15) and (16) are used to analyze the
jumping motion of the mechanical model by means of numerical
integration. The dynamic characteristic of tibia and body particle
can also be calculated through Eqs. (6) and (9). Since the femur
OA and tibia AB are considered as massless ideal rods, the contact
force Fy at tarsusground interface during jumping acceleration
can be calculated by

mg
F y mo

17

According to above equations, the contact force Fy at tarsus


ground interface can be nally derived as a function of its jumping
kinematics including limb angle and velocity. In boundary conditions, the initial jumping velocity o_ 0 and the initial limb angle
velocity a_ 0 and b_ 0 are set to be 0; for a given tibia/femur ratio,
the simulation starts with a certain angle a0 and b0 as well as a certain body particle position o0 which should geometrically satisfy
Eqs. (3), (4), and (7); the calculation stops when the contact force
less than body weight.
Fig. 3 shows the analysis results of above model by applying different saltatorial leg structure. The jumping kinematic characteristics shown in Fig. 3a is very close to the biological observation
results of leafhopper jumping measured by Burrows in 2006 and
2007 [16,25]. Certain take-off velocity can be nally achieved with
different saltatorial leg structure because of the same energy storage (Fig. 3b). However, along with the increasing of tibia/femur ratio, body velocity during jumping acceleration tends to be linear
and the acceleration period reduced due to the different elongation
strategy resulting from varying tibia/femur ratio. Shorter acceleration period leads to an increasing of the impulse during jumping
acceleration and may result in a larger contact force which could
cause leg ruptures and tarsus slippage more easily.
The dynamic characteristic at tarsusground interface is analyzed and shown in Fig. 3c. The simulation result reveals that peak
contact force usually appears at the beginning of the jumping motion and reduces to zero with different strategies at the instant of
ground contact loosing. Both a small tibia/femur ratio and a large
one lead to a large peak force (Rtf = 1.1 or 4.0) despite a same

Fig. 4. Design of 1-DOF articial saltatorial legs: (a) schematic diagram of the spatial four-bar linkage mechanism; (b) mechanically practicable saltatorial leg design;
(c) saltatorial leg elongation and the tarsus motion orbit.

171

F. Li et al. / Mechatronics 22 (2012) 167176

take-off velocity as a jumping result. On the other hand, it shows a


better dynamic characteristic when tibia/femur ratio is around 1.7.
Fig. 3d shows the relationship between leg structure and the peak
contact force. In our opinion, at such a proper tibia/femur ratio, the
non-linear muscle stress is mapped into a nearly constant contact
force output, thus helps the insect minimize both the structure
load and the risk of slippage. For example, this optimized tibia/femur ratio is also adopted by leafhoppers who have an average ratio
of 1.72 0.15 according to the observation and measurement in
[2]. It unveils the mechanisms beneath the optimization in jumping insects, and from a biological point of view, such a coincidence
also indicates an optimization result of nature evolution.

Table 1
Parameters of segment gears.
Segment gears
(Fig. 5)

Actuating gear

Actuated gear

Material
Gear type
Module number
Teeth number

Stainless steel
Involute gear
0.3
12  2 (36 for a whole
gear)
11.40 mm
10.80 mm
9.99 mm
2.0 mm
0.80 g

23 (60 for a whole


gear)
18.60 mm
18 mm
17.19 mm
1.0 mm
1.14 g

Addendum diameter
Reference diameter
Root diameter
Teeth thick
Weight

3. Design and fabrication of jumping robot prototype


3.1. Saltatorial legs design
Therefore, the articial saltatorial leg with a tibia/femur ratio of
1.7 is then designed based on a spatial four-bar linkage mechanism shown in Fig. 4a. Femur and auxiliary bar can rotate with respect to axis 1 and 2 on the body. There are ve restrains at point O
and C, four restrains at point A and three restrains at point D, so
the DOF of the whole mechanism is set to 1 in order to simplify
the controlling. By adjusting length of the auxiliary bar, a linear
elongation of the tarsus (point B) can be achieved in order to imitate the desired leg kinematics of insect jumping. Before jumping,
the articial saltatorial legs are positioned directly beneath the
body and pressed against each other just like the levation phase
observed in leafhopper jumping [16]. Such a design is not only
for imitation of a real insect, but also reduces the chance of asymmetric thrust forces between two legs, as well as the destabilizing
rotation during the ight phase. Fig. 4b shows the mechanically
practicable saltatorial legs including a 24 mm long femur, a
42.8 mm long tibia and a 31.5 mm long auxiliary bar which result
in a 43.2 mm tarsus elongation. In addition, considering the
assembling feasibility, the restrains at point A and point D are designed as separated hinge joints (joints 2 and 3, joints 46) thus

brings a curved but tolerated changing on the tarsus motion orbit,


as shown in Fig. 3c.
3.2. Actuating system
According to biological reports, before each jumping, it takes
about 200 ms for the insect restoring enough jumping energy by
slowly extending its tendon with muscles, the jumping motion is
then carried on within tens of micro seconds by fast releasing
the restored energy. For a mechanical design, such a natural
muscletendon system is designed as a springsegmental gear
system to transmit continuous rotation of a DC motor to a
reciprocal load-release motion so as to generate continuous
jumps.
Fig. 5 shows the working strategy of the springsegmental gear
system as well as the detail structure of adopted segmental gears.
Two springs with same stiffness of 1 N/mm are used in order to restore 56.3 mJ jumping energy considering the 7.5 mm spring elongation due to the designed structure. Springs are loaded during
gear meshing period and released after meshing. The fast releasing
of the springs extends saltatorial legs thus launches the robot into
ight. In order to avoid the deformation and abrasion during the

Fig. 5. Working strategy (ad) and detail structure (e) of the springsegmental gear system, teeth of the actuating gear are left symmetrically respect to a circle in order to
generate two jumps in one rotation and unnecessary parts of the driven gear is cut for mass reducing.

172

F. Li et al. / Mechatronics 22 (2012) 167176

Fig. 6. Jumping robot prototype design and simulation model: (a) levation phase (23 s); (b) holding phase (10 s); (c) launching phase (20 ms); (d) a complete ADAMS model
of the whole jumping robot structure.

Fig. 7. Simulation results of jumping model.

F. Li et al. / Mechatronics 22 (2012) 167176

meshing period, segment gears are made of stainless steel and fabricated by high-precision wire electroerosion machining. Teeth of
the actuating gear are left symmetrically respect to a circle in order
to improve the moving efciency by generating two jumps in one
rotation. Unnecessary parts of the driven gear are also cut for mass
reducing. Other details about these two segment gears are reported
in Table 1.
3.3. Prototype design and jumping simulation
Biological observation also shows that there are three distinct
phases during insect jumping from preparing to nal launching
[16]. Usually, leafhopper jumping starts with a 1530 ms levation
phase in which the insect positions its hind legs directly beneath
the body with tarsus pressed against each other. Then it turns to
a 10200 ms holding phase in which the hind legs remain the
stance and the take-off angle is adjusted by front and middle legs.
At last the jump is generated by simultaneous extension of the two
long hind legs within 56 ms.
Fig. 6 shows the design of the jumping robot prototype. By imitating the insect jumping strategy, jumping of the robot prototype
starts with a 23 s levation phase waiting for gear meshing
(Fig. 6a). After that, the holding phase begins and usually takes
about 10 s (Fig. 6b). Finally, gear meshing ends and it turns to
the jumping phase including acceleration, ight and landing which
usually costs several seconds totally. A cephalic motor and caudal
battery are positioned to balance the mass distribution. The wing
skeleton is deigned as a support of exible wings which are linked
to the saltatorial legs and spreads after jumping acceleration in order to improve the ight stability.
In order to test the feasibility of such a design and investigate
the optimized dynamic characteristics of the jumping prototype,

173

a complete ADAMS model of the whole structure is then built


and shown in Fig. 6d. Simulation results shown in Fig. 7ac indicate the similarity of both jumping kinematic and dynamic characteristics to the optimized one derived from theoretical analysis in
Fig. 3. It reveals that, with the given leg design, contact force at tarsusground interface resulting from leg elongation tends to remain
as a nearly constant output during the launching period, which can
be considered as an optimized actuating strategy regarding a given
take-off velocity because of the minimization of structure rupture
and slippage risk avoidance. From a bionic viewpoint, such optimized jumping characteristics are mainly proted from the successful imitation of insect mechanics and achieved by an
increasing effective mechanical advantage (EMA) of the leg structure, which is also reported and explained in some previous biological research [17,26]. Fig. 7d shows the increasing EMA of the
jumping robot model. To better illustrate the dynamic similarity
between the model and animal jumping, Fig. 7d also shows the
comparison between the muscle contraction property of jumping
robot model and the one reported by Thomas in 2003 based on
the analysis on bullfrog jumping [26].
3.4. Prototype fabrication and jumping stability improvement
The nally assembled prototype is a 22 g weight jumping robot
of 50 mm  25 mm  20 mm in dimension with onboard energy
integrated (Fig. 8c). The free status and loaded status of such a robot prototype are shown in Fig. 8a and b. Most assembling parts of
the prototype is fabricated by Rapid Prototyping Manufacturing
(RPM) with plastic material, and considering the large stress
caused by fast extending of the saltatorial legs, some stress concentrated leg parts are made of aluminum (e.g. in Fig. 8a). A 30 mA h
lithium battery is positioned on the rear as an onboard energy

Fig. 8. Finally assembled jumping robot prototype GRILLO III: (a and b) jumping robot prototype before assembling the exible wings in releasing and loading status
respectively; (c) jumping robot prototype with exible wings.

174

F. Li et al. / Mechatronics 22 (2012) 167176

source which ensures 30-min continuous jumping of the robot.


Considering the loading requirements in practical applications, a
platform is assembled upon the head capacitates the robot carrying
ability of small monitoring devices or sensors. Moreover, the moving stability, including the stability during jumping acceleration,
ight and landing phase, is improved respectively by a better mass
distribution, by adding wings and a tail on the body, and by assembling of the passive forelegs and balancing midlegs. That means
even though an exact 100% simultaneous landing with both legs
cannot be ensured due to its fabrication and assembling errors,
the implementation of such preventing accessories helps correct
body attitude and much reduce the overturning risks during the
landing moment.
4. Jumping experiments
The jumping activities are then measured by high speed camera experiment. The sequential images of a single jump are rst
captured at frame rate of 1000 fps with a HotShot 512 SC high
speed camera (NAC Image Technology Inc.). Jumping kinematics,
including xy projected angles between body and femur, femur
and tibia, jumping velocity during the 40-ms acceleration, are obtained by analysis on the jumping frames from the start of jumping acceleration till the instant of ground contact loosing.

Representative frames are shown in Fig. 9. The jumping starts at


0 ms with the extending of the saltatorial legs (Fig. 9a); between
0 ms and 40 ms, the saltatorial legs elongate rapidly to force the
robot jump as a result of spring shortening (Fig. 9b); the acceleration period ends at 40 ms when the robot is launched into a
ight (Fig. 9c and d) with a measured jumping velocity of about
1.7 m/s. Thus 31.8 mJ kinetic energy is nally transformed from
the restored energy in the stretched springs. Based on the calculation mentioned in Section 3, restored jumping energy resulting
from spring contraction is about 56.3 mJ. Then considering 8.8 mJ
potential changing of the robot body during the jumping acceleration as well as the energy distribution of wings spread, the total
mechanical efciency of the saltatorial legs should be around 70%
and the energy lost mainly results from the friction at each hinge
joint because barely slippage is observed at tarsusground
interface.
The limb kinematics of a leafhopper during its jumping measured in our previous biological experiments is shown in Fig. 10
[2,17] for comparing, while the measuring results of the prototype are shown in Fig. 11. Due to the successful imitation of insect mechanics, the jumping characteristics of both the limb
kinematics (bodyfemur angle and femurtibia angle vs. time,
Fig. 11a) and the body velocity curve (Fig. 11b) show a similarity to the one derived from theoretical analysis and dynamic

Fig. 9. Representative frames of the robot jumping captured by high speed camera at 1000 fps: (a) the jumping starts with the extending of the saltatorial legs; (b) the
saltatorial legs elongate rapidly as a result of spring shortening; (c) ground contact loosing at 40 ms; (d) the robot is launched into a ight with a measured jumping velocity
of 1.7 m/s.

Fig. 10. Measuring results of the insect jumping: (a) representative frames of leafhopper jumping captured by high speed camera; (b) limb kinematics of the jumping insect.

175

F. Li et al. / Mechatronics 22 (2012) 167176

Fig. 11. Measuring results of the robot jumping.

to nd a balance between jumping performance and locomotion


stability.

Table 2
Details of the prototype design and jumping performance.
Length
Width
Height
Total weight
Motor power
Battery duration
Take-off velocity
Accelerating period
Jumping length
Jumping height
Mechanical efciency of the saltatorial legs

50 mm
20 mm
25 mm
22 g
0.3 W
0.5 h
1.7 m/s
40 ms
About 200 mm
About 100 mm
70%

simulation. The little difference between insect and robot kinematical curve may result from the DOF discrepancy at both
bodycoxa and coxatrochanter joints. Actually, due to the
conservative design required by measuring convenience and robot durability, the jumping performance can be further developed by replacing applied springs with stiffer ones. Details of
prototype design and jumping performance are also reported
in Table 2.
To better illustrate the characteristics of such jumping robot, a
comparison between this prototype and typical jumping robots
in the literature is presented in Table 3. The integration of jumping
mechanism on traditional wheeled robots helps improve their
passing ability in complex working conditions but also results in
an increase on the body size and weight, which inevitably reduces
the jumping performance. On the other hand, by simplifying the
whole actuating method, remarkable jumping performance can
be obtained despite a sacrice of its controllability and locomotion
stability. Thus we developed the jumping robot GRILLO III, aimed

5. Conclusion and future work


Taking inspiration from the natural world, an optimal locomotion solution for mini/micro sized bio-robots is derived from learning how insects choose their locomotion mode and how they
evolve to improve themselves. Based on a simplied planar model,
theoretical analysis reveals that a proper leg design may lead to an
optimization of the dynamic characteristics during jumping acceleration. By mapping the non-linear muscle contraction into a
nearly constant contact force output at tarsusground interface,
the maximum jumping force can be reduced, which helps the insect minimize both the structure load and the risk of slippage. It
shows that, for both leafhopper and jumping robot, the optimal dynamic characteristics are gained at a tibia/femur ratio of around
1.7, which also, on some level, explains the motive power of insect
saltatorial leg evolution.
By reproducing the insect optimization mechanism in the prototype, the jumping robot GRILLO III is then designed based on
the optimal articial saltatorial leg structure which is proposed
to reduce the contact force at tarsusground interface during
jumping acceleration thus minimize the risk of both leg ruptures
and tarsus slippage. By means of high speed camera experiments,
the jumping characteristics of the theoretical jumping model, the
simulation model, the insect jumping and the robot prototype
are compared in order to show the dynamical similarity and optimization results. The 22 g weight energy integrated jumping robot
GRILLO III is able to accomplish continuous operation by jumping
100 mm high and 200 mm long in each single jump.

Table 3
Comparison between GRILLO III and typical jumping robots.
Robot prototype

Miniature jumping robot [19]

GRILLO III

Mini-Whegs 9 J [11]

Characteristic body length (mm)


Weight (g)
Jumping length (mm)
Jumping performance
(jumping length/body length)
Jumping stability

50
7
1380
27.6

50
22
200
4.0

104
191
180
1.7

Uncontrolled
jumping

Stable jumping
locomotion

Stable jumping as obstacle overcoming


method, wheeled locomotion

176

F. Li et al. / Mechatronics 22 (2012) 167176

Fig. 12. Preliminary result of the electronic controlling system and bio-mimetic sensors: (a) cilia shaped ow sensors on spider limb [27]; (b) bio-mimetic articial sensory
hair made of piezoelectric polymer; (c) jumping robot prototype equipped with digital accelerometer to correct the launching angle.

As the extension of mechanical and bionic research on the


jumping robot, future work will be mainly focused on the integration of electronic controlling system and bio-mimetic sensors
which enable the robot to move autonomously in complicated
environment or endue it sensing ability in environmental monitoring. Indeed, in order to reduce the effect of wind drag, jumping angle is adjusted by front and middle legs during the holding phase
before takeoff based on the detection of cilia shaped mechanosensory system as shown in Fig. 12a [27,28]. Therefore, given inspiration from such biological observations, a pilot fabrication of
biomimetic sensing hairs with piezoelectric property based on
the improvement of thermo-direct drawing technique [29] is carried on and considered as a promising solution for improving the
intelligence and automation of mini/micro jumping robots in the
future. Fig. 12b shows the preliminary fabrication result of such
bio-mimetic sensory hair compared with its biological counterpart
(Fig. 12a). On the other hand, the digital accelerometer on the robot (Fig. 12c) back feeds the controlling system with attitude information and helps the robot approach to the right angle.
Acknowledgements
The activity presented in this paper is partially supported by
National Basic Research Program of China (973 Program,
2011CB013303) and the LAMPETRA Project (EU contract No.
216100), respectively. Authors also would like to thank Dr. Dajing
Chen of Zhejiang University for his outstanding work on the investigation of insect jumping and robot testing.
References
[1] Alexander RM. Principles of animal locomotion. Princeton: Princeton
University Press; 2003.
[2] Li F, Bonsignori G, Scarfogliero U, Chen D, et al. Jumping mini-robot with bioinspired legs. In: Proceedings of IEEE international conference on robotics and
biomimetics, Bangkok, Thailand; 2009. p. 9338.
[3] Scholz MN, Bobbert MF, Knoek van Soest AJ. Scaling and jumping: gravity loses
grip on small jumpers. J Theor Biol 2006;240:55461.
[4] Farley CT, Taylor CR. A mechanical trigger for the trotgallop transition in
horses. Science 1991;253:3068.
[5] Song SM, Waldron KJ. Machines that walk: the adaptive suspension
vehicle. Massachusetts: MIT Press; 1989.
[6] Arikawa K, Mita T. Design of multi-DOF jumping robot. In: Proceedings of IEEE
international conference on robotics and automation, Washington, DC, USA;
2002. p. 39927.

[7] Hyon SH, Mita T. Development of a biologically inspired hopping robot


Kenken. In: Proceedings of IEEE international conference on robotics and
automation, Washington, DC, USA; 2002. p. 398491.
[8] Tsukagoshi H, Sasaki M, Kitagawa A, Tanaka T. Jumping robot for rescue
operation with excellent traverse ability. In: Proceedings 12th international
conference on of advanced robotics, Seattle, USA; 2005. p. 8418.
[9] Papanikolopoulos N. Kinematic motion model for jumping scout robots. IEEE
Trans Robot 2006;22:397402.
[10] Kim DH, Lee JH, Kim I, Noh SH, Oho SK. Mechanism, control, and visual
management of a jumping robot. Mechatron 2008;18:591600.
[11] Lambrecht BGA, Horchler AD, Quinn RD. A small, insect-inspired robot that
runs and jumps. In: Proceedings of IEEE international conference on robotics
and automation, Barcelona, Spain; 2005. p. 12405.
[12] Saranli U. RHex: a simple and highly mobile hexapod robot. Int J Robot Res
2001;20:61631.
[13] Bennet-Clark HC, Lucey ECA. The jump of the ea a study of the energetics
and a model of the mechanism. J Exp Biol 1967;47:5976.
[14] Bennet-Clark HC. The energetics of the jump of the locust Schistocerca gregaria.
J Exp Biol 1975;63:5383.
[15] Alexander R. Leg design and jumping technique for humans, other vertebrates
and insects. Philos Trans Roy Soc Lond B 1995;347:23548.
[16] Burrows M. Kinematics of jumping in leafhopper insects (Hemiptera,
Auchenorrhyncha, Cicadellidae). J Exp Biol 2007;210:357989.
[17] Scarfogliero U, Bonsignori G, Stefanini C, Sinibaldi E, et al. Bioinspired jumping
locomotion in small robots: natural observation, design, experiments. Springer
Tracts Adv Robot 2009;54:32938.
[18] Scarfogliero U, Stefanini C, Dario P. A bioinspired concept for high efciency
locomotion in micro robots: the jumping robot Grillo. In: Proceedings of IEEE
international conference on robotics and biomimetics, Kunming, China; 2006.
p. 403742.
[19] Kovac M, Fuchs M, Guignard A, Zufferey JC, et al. A miniature 7G jumping
robot. In: Proceedings of IEEE international conference on robotics and
automation, Pasadena, USA; 2008. p. 3738.
[20] Kovac M, Schlegel M, Zufferey JC, Floreano D. Steerable miniature jumping
robot. Auton Robot 2010;28:295306.
[21] Bergbreiter S, Pister KSJ. Design of an autonomous jumping microrobot. In:
Porceedings of IEEE international conference on robotics and automation,
Rome, Italy; 2007. p. 44753.
[22] Burrows M, Morris O. Jumping and kicking in bush crickets. J Exp Biol
2003;206:103549.
[23] Burrows M. Froghopper insects leap to new heights. Nature 2003;424:509.
[24] Gorb SN. The jumping mechanism of cicada Cercopis vulnerata
(Auchenorrhyncha, Cercopidae): skeletonmuscle organisation, frictional
surfaces, and inverse-kinematic model of leg movements. Arthropod Struct
Dev 2004;33:20120.
[25] Burrows M. Jumping performance of froghopper insects. J Exp Biol
2006;209:460721.
[26] Roberts TJ, Marsh RL. Probing the limits to muscle-powered accelerations:
lessons from jumping bullfrogs. J Exp Biol 2003;206:256780.
[27] Barth FG. Spider mechanoreceptors. Curr Opin Neurobiol 2004;14:41522.
[28] Dangles O, Casas J, Coolen I. Textbook cricket goes to the eld: the ecological
scene of the neuroethological play. J Exp Biol 2006;209:3938.
[29] Li F, Liu W, Stefanini C, Fu X, Dario P. A novel bioinspired PVDF micro/nano hair
receptor for a robot sensing system. Sensors 2010;10:9941011.

Vous aimerez peut-être aussi