rolling robots review

Upload: durgesh

Post on 01-Jun-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/9/2019 Rolling Robots Review

    1/14

    Available online

    at

    w sciencedirect com

    ScienceDirect

    Journal of Bionic Engineering 3 (2006) 195-208

    Rolling in Nature and Robotics:

    A

    Review

    Rhodri H. Arm our, Julian

    F.

    V. Vincent

    Centre or Biomitnetic and Natural Technologies, Department ofMechanica1 Engineering,

    University o Bath, Bath BA2 7AY, U K

    Abstract

    This paper presents a review of recent rolling robots including Rollo from Helsinki University of Technology, Spherical

    Mobile Robot from the Politecnico of Bari, Sphericle from the University of Pisa. Spherobo t from Michigan S tate University,

    August from A zad University

    of

    Qazvin and the University of Tehran. Deformab le Robot from Ritsumeijan Un iversity, Kickbot

    from the Massachusetts Institute

    of

    Technology, Gravitational Wheeled Robot from Kinki University, Gyrover from Carnegie

    Mellon University, Roball from the Universite de Sherbrooke. and Rotundus from the Angstrom Space Technology Center.

    Seven rolling robot design principles are presented and discussed (Sprung central member. Car driven, Mobile masses,

    Hemispherical whee ls, Gyroscopic stabilisation. Ballast mass ixed axis, and Ballast mass moving axis). R obots based on

    each of the design principles are shown and the performances of the robots are tabulated. An attempt is made to grade the design

    principles based on their suitability for movement over an unknown and varied but relatively smooth terrain. The result of this

    comparison su ggests that a rolling robot based on a mobile masses principle would be best suited to this specific application.

    their own rolling motion or external forces cause their rolling.

    Keywords:

    Rolling robot, tumbleweed, wheel, ballast, Gryoscope

    Copyright

    2006, Jilin University.

    Published by Science

    Press

    and Elsevie r Limited.

    All

    nghts

    reserved.

    Som e wonderful rolling organisms are introduced and defined as active or passive depending on whether they generate

    1

    Introduction

    It is important to distinguish between a true rolling

    robot and one w ith jus t large wheels and a reaction point

    with the ground. It seems suitable to define a rolling

    robot as one that rolls on its entire outer surface rather

    than just external wheels and does not need to react any

    of the rotating torque against the ground. Thus they tend

    to be spherical or cylindrical in form and have a single

    axle (or no axle) and a com pletely active outer surface

    i.e. a surface that is completely involved in the move-

    ment.

    A spherical rolling robot also has the following

    advantages over a traditional wheeled robot o r walking

    robot:

    It

    is

    possible to enclose the entire robot system

    inside the spherical shell and thus provide mechanical

    and perhaps even environmental protection to compo-

    Correspondingauthor:

    Rhodri

    H.

    Armour

    E-mail:

    [email protected]

    nents and equipment.

    There are no body extremities that can hang-up

    on obstacles.

    The entire outer body is driven (or rotates) help-

    ing the device to cover uneven or soft surfaces.

    Spherical robots have no side to fall over upon

    from which they ca nnot recover.

    Spherical robots have a lower ground contact

    pressure d ue to a w ider footprint than a simple wheel or

    foot

    so

    they can be used for travelling on soft substrates

    such as sand, snow, brush, or vegetation and for paddling

    through water.

    Rolling spheres can potentially move in any di-

    rection and therefore can turn in place when meeting

    obstacles over which they cannot roll. This makes their

    control relatively simple.

    Spherical robots can recover from collisions

    easily.

  • 8/9/2019 Rolling Robots Review

    2/14

    196

    Journal of Bionic Engineering

    (3006)

    vo1.3 No.3

    Significant numbers of rolling robots are under

    development but i t is still in the early stage mainly for

    the proofs of concepts being undertaken rather than

    optimisation for maximum perform ance. The proof of

    concept includes the design of the mobility system but

    also takes significant consideration of the control s ystem

    and path planning issues. Few of the prototypes have a

    specific application in mind but there is obviously wide

    research interest in rolling robots. Suitable applications

    can be found once the rolling and control issues have

    been solved.

    2

    Rolling in nature

    By

    looking toward nature. perhaps some lessons

    regarding rolling can be learned and applied to rolling

    robots. However. nature has not adopted a rolling m otion

    in

    the main and exam ples are rare. This is not to say

    there are no rolling organism s, but only a few have been

    discovered.

    Rolling seems to be a secondary form of motion for

    all known rolling organisms, the Tumbleweed plant

    being

    a

    possible exception. There are also two variants

    of rolling

    in

    nature which are here referred to as pas-

    sive and activ e. Passive rolling requires extern al

    forces, such as wind or gravity. to drive the movement.

    Active rolling is where the organism expends its own

    energy in achieving the rolling and thus controls its

    rolling such that it is able to move

    in

    a specific direction.

    2.1 Passive rolling

    Passive rolling organisms include the Tumbleweed

    plant, the Web-toed Salamander, Namib Golden Wheel

    Spider, and the Woodlouse.

    Rolling is a secondary form of movement for all

    known passive rollers except the tumbleweed, since the

    animals have legs for normal movement such as

    searching for food. Rolling motion is only adopted as an

    escape during attacks from predators since it allows

    much higher speeds for these animals. The web-toed

    salamand er curls itself into a hoop sh ape, with its dorsal

    side outermost, and rolls down slopes with rocky

    surfaces tar faster than would be possible by simply

    running down the slope. The Namib golden wheel

    spider (Fig. I ) cartwheels down sand dunes when at-

    tacked by its nemesis he Poinpilidae tararitirla wasp.

    It rolls down slopes at m K

    or

    20 rotations per sec-

    ond. Som e types of woodlouse , curl into an armoured

    ball when attacked or startled, and roll passively if on a

    sloped surface.

    There is very little information regard-

    ing the speeds and mechanics

    of

    these animals rolling

    motions.

    The tumbleweed is novel i n

    its

    rolling since it

    harnesses the force of the wind to roll over a wide area of

    smoo th and flat land. Tum bleweed is a weed that lives

    for only one season. Its most recognised form is one

    that looks like a rounded skeleton of a normal shrub

    (Fig.

    2).

    However, there was a time in its life when it

    was a normal brightly green coloured rooted plant.

    During the spring and summer, the plant grows very

    much like a normal shrub. How ever.

    in

    autumn , after the

    dry sum mer, a specialised lay er of cells in the plant stem

    allows the dried plant to break away from its roots and

    begin its wind driven journe y. During rolling, the plant

    disperses up to

    250

    000 seeds over a wide area.

    Each

    seed is a small coiled embryonic plant wrapped by a

    thin skin which is very different from typical hard se eds

    Fig. 1 Namib wheeling spider rolling.

    Fig. 2 Photograph of tumbleweed.

  • 8/9/2019 Rolling Robots Review

    3/14

    Rhodri

    H. Amo ur , Julian

    EV. Vincent: Rolling in Nature

    and

    Robotics:

    A

    Review

    197

    containing large nutrient stores. When moisture levels

    increase later in the season, the seed germinates and

    begins to grow starting the annual cycle.

    It is the form of tumbleweed that ensures its ability

    to roll. Th e plant is almost spherical in its rolling state

    and its large surface area generated by the intricate

    branch structure harnesses enough wind force to over-

    come the plants weight. The centre of gravity of the

    whole plant is not at the centre of the sphere and it is

    thought that this causes the plant to bounce as it rolls

    encouraging seeds to be released, and improving its

    speed by keeping it airborne and removing the rolling

    resistance with the ground31.

    Tumbleweeding is pursued as a form of locomo-

    tion for a range of rolling robots being developed by

    NASA

    Langle~~.]nd ESA. The direction control of

    these devices is not considered in detail but it is felt that

    using the wind as a free power source should enable

    large unexplored areas to be discovered with very little

    energy expended on the movement.

    2.2 Active rolling

    If developing an autono mous rolling robot, then i t

    is probably active rolling that is most interesting since

    robots using this sort of movement

    are

    able to choose in

    which direction to travel. After an exten sive literature

    search, only two active rolling organisms have been

    discovered he caterpillar of the Mothe r-of-pearl Moth

    and the stomatopod shrimp Nannosquilla decemspinosa.

    2.2.

    I

    Mother-of-Pearl M oth caterpillar

    The caterpillar of the Mother-of-pearl Moth

    Pleum0,a rwalis)

    when attacked with sufficient ag-

    gression, fixes its tail onto a surface and pushes its

    foremost segm ents backward quickly with its front legs.

    Once the head reaches the tail, the caterpillar rolls into a

    wheel shape (Fig.

    3)

    with its back outermost and con-

    tinues to roll for up to five complete revolutions[61.

    Speed of around

    40

    cm.s- has been measured w hich is

    about

    40

    times faster than its normal walking speed.

    This means that the caterpillar of 25 mm long will have

    moved itself 125 mm in around

    0.3 s,

    enough to outrun a

    predator. This form or retreat obviously surprises a

    predator and give s the caterpillar valuable time to escap e.

    Fig3 Photographs

    of

    the M other-of-pearl caterpillar

    during rolling[61.

    Th e series of rolls are all driven by a single imp ulse so it

    cannot be considered a continuous rolling motion,

    however, an external force, such as gravity, is not ap-

    plied. Table 1 shows the details of the Mother-of-pearl

    Moth caterpillar.

    Table

    1

    Performance capability of M other-of-pearl

    moth caterpillar

    Capability Measurement

    Diameter

    (m) 0.008

    Speed

    d s ) 0.4

    2.2.2

    Stomatopod shrimp annosquilla decemspinosa

    The

    Nannosquilla decemspinosa

    is a type of shrimp

    that adopts an active rolling motion when washed up

    onto a beach. They spend m ost of their time underwater

    in their burrow s where they wait for prey. They swim

    only a sma ll distance, up to one body length, out of their

    burrows to collect food. Because of their elongated

    bodies and short laterally projecting legs, adults cannot

    walk when ou t of the water. Their long and low bodies

    rest on the substrate and the friction is too large for legs

    to drag them along an d the legs are not strong enough to

  • 8/9/2019 Rolling Robots Review

    4/14

    198

    Journal of Bionic Engineering

    (2006)

    Vo1.3 No.4

    lift their body from the surface. When a wave washes

    them from their burrows onto the beach and out

    of

    the

    water. they perform up to

    2

    m of backward somersaults

    (20 40 rotation s) returning to the The rolls

    do not occur uninterrupted since the animal ends up

    lying on its back after each cycle and then inputs another

    pulse of energy.

    A diagram of their rolling motion is

    shown in Fig. 4. They are very flat

    in

    the dorsoventral

    plane which imp roves rolling stability and, when rolling.

    tend

    to

    lean downbeach so will eventually reach the

    water,

    The stability of each roll is good since for m uch of

    the time the centre of gravity of the animal is close to the

    ground

    such

    that

    it

    doesn't topple over. There is a defi-

    nite pause after each somersault which also gives a

    useful stable resting point. Table 2 shows the details of

    the

    Nannosquilla

    decemspino

    su

    Table 2 Performance capabilityofNurinosquillu

    decenispinosa

    Capabil i ty Measurement

    Diameter ( n i l 0.007

    Length rn) 0

    073

    Speed d s ) 0-1

    Static slope climbing ( ) 10- 30

    3

    Design principles of rolling robots

    The eleven robots reviewed here provide the cur-

    rent state-of-art of rolling robots. All are based on the

    principle of moving the centre of gravity of a wheel or

    sphere outside of the contact area which causes the

    wheel or sphere to fall in that direction and thus roll

    along. This is best displayed diagram matically as shown

    in Fig.

    5 .

    Of the eleven robots, there are only seven funda-

    men tally different design princ iples. It is difficult to

    directly compare the individual performances of each

    robot since none have been optimised for use specifi-

    cally as

    an

    autonomous mobility platform o factors.

    such as speed. obstacle traversabilty, or slope climbing

    ability, are far from maximised. A discussion and

    evaluation of the seven design principles follows with

    details of the robots that have applied them.

    3.1 The first design principle prung

    central mem ber

    The first design uses a single driven wheel on a

    sprung member with a passive wheel on the other end

    (Fig.

    6).

    Each wheel makes contact with the inner sur-

    face of the sphere at opposite s ide and the sprung central

    mem ber applies the contact force. Steering of the device

    involves rotating the driven wheel around on its contact

    patch within the sphe re using the inertia of a fixed mass

    on the member to react the steering torque. The central

    member has two motors. One drives the driven wheel

    and the other steers the driven wheel.

    A s the driven

    Stationary

    Rolling

    \ ontact parch

    /

    Fig.

    5

    Rolling by moving centre

    of

    gravity.

    Fig. 6

    Sprung central member concept, solid arrow shows

    rolling source, open arrow shows steering source.

  • 8/9/2019 Rolling Robots Review

    5/14

    Rhodri H.

    Amour ,

    Julian

    F.V. Vincent:

    Rolling in

    Nature

    and Robotics: A Review

    199

    wheel rotates, the drive mechanism moves within the

    sphere unbalancing the sph ere and causing it to fall for-

    ward and roll. Due to the arrangement of the masses on

    the central member, the centre of gravity of the whole

    sphere is situated beneath the geometric centre which

    ensures stability.

    This design should enable the device to stop in one

    position and continue in an entirely new direction

    without rotating or turning. However, of the two robots

    developed,

    roll^[^^

    is now based on another principle

    and the Sph erical Mobile Robot '] has not yet move d

    from the existing test cylinder that has a supported cen-

    tral axle. It seem s that this design is not popular with

    research teams and the one that has pursued it to the

    highest level have rejected it in favour of the Ballast

    mass moving axis concept (section

    3.7).

    This is likely to

    be due to the difficulties of ensuring a constant force on

    the driven wheel whilst keeping the sprung member

    running through the geometric centre of the sphere.

    Steering may also be a problem as

    it

    relies on being ab le

    to react the torque of the turning motors attempting to

    rotate the driven wheel around on its contact patch. This

    reaction can only occur through inertia of a mass so will

    tend to increase the overall weight of the design. With

    optimization it could be m ade to work but there are more

    simple solutions.

    Fig. 7 Rollo prototype one shown without spherical

    enc~osure[~].

    3.1.1 The Ro llo robot

    Rollo robot was developed by Helsinki University

    of Technology

    &

    Rover Com pany Ltd[ ].

    It

    is a ball

    shaped exploratory robot platform. The initial autono-

    mous prototype operated on the sprung central member

    concept but this design was rejected in later prototypes

    (Fig. 7). A specific explanation of why this occurred has

    not been mentioned and no performance details of this

    first prototype were presented.

    3.1.2The Spherical Mobile Robot (SMR)

    The S MR was developed by Politecnico

    of B a ~ i ~ ' .

    The design of the SMR is based on the sprung central

    member concept (Fig.

    8).

    However, the prototype exists

    only as a cylindrical de vice of

    160

    mmdiameter with the

    drive mechanism pivoted around its central axis (Fig. 9).

    The drive mechanism uses a DC motor to power two

    Fig. 8 SMR design concept '].

    Fig. 9 SMR cylindrical prototype ''.

  • 8/9/2019 Rolling Robots Review

    6/14

    200

    Journal of Bionic Engineering

    (2006)

    Vo1.3 No.4

    drive wheels by radio contro l. There is no speed control

    since a specific voltage will move the device at

    a

    quicker

    speed

    on

    a rigid surface than on a soft surface, such as

    sand. This feature enables the robot to feel the terrain.

    The device achieved a speed of 0.6 m. s but this was to

    validate models of it performance so higher speed may

    be possible. A complete summary of its capability is

    shown in Table

    3.

    Table

    3

    Perform ance capability

    of

    SMR prototype

    Capability Measurement

    also has the added effect of smoo thing the surface for the

    enclosed car.

    The Sph ericle from University of Pisa / Siena is

    an example which is an untethered sph erical vehicle that

    can roll around on the floor (Fig. 1 1 .

    Its movement is

    based on the car driven concept with a small wheel

    driven unicycle ca r placed

    on

    the inside surface of the

    sphere. The mov ement of the car causes the whole robot

    to accelerate, steer and decelerate. There are no details

    of its performance presented in the literature.

    Diameter (m)

    0.16

    Speed

    d s ) 0.6

    Other notes Cylindrical prototype only

    3.2

    The second design principle ar driven

    The second design concept uses a small car

    placed within the sphere allowing the whole robot to

    move when the car drives around the inside his

    is

    very similar to the way in which a hamster, placed inside

    a ball, moves around a room (F ig.

    10).

    It allows a device to stop in one position and then

    move off in another direction without turning in place

    so

    long as the car inside can turn

    in

    place.

    There are

    drawbacks however.

    The car cannot be guaranteed to

    remain in contact with the inner surface of the sphere

    and if the device were to tumble down a slope or step,

    the car could become dislodged and end up resting on its

    back within the sphere without being able to turn over

    thus stranding the whole device.

    When the device is

    moving quickly over

    a

    non-smooth surface, the friction

    between the car wheels and the inner surface of the

    sphere will vary and perhaps loose contact all together.

    This makes control difficult since simp le encoders can-

    not be used on the cars wheels and there is no other

    stable platform for range mea surements

    or

    cameras. The

    simplicity of this design does enable a very lightweight

    solution since the driving mass

    of

    the sphere is as far

    from the geometric centre as possible and can therefore

    be made as light as possible. It should allow for the

    largest slopes and obstacles to be overcom e for this same

    reason although the motion of the car is limited to the

    lower half of the sphere. If considering how t he car

    alone would move across a rough surface, the sphere

    3.3

    The third design principle Mobile masses

    The third concept uses the movement of masses

    along radial axles of

    a

    sphere to alter the centre of

    gravity

    of

    the sphere such that can achieve movement

    (Fig.

    12).

    The design allows movement

    in

    any direction

    at any time since the centre of gravity of the device can

    be located at almost any position within the sphere. Two

    Fig. 10 Car driven concep t, solid arrow sho ws rolling

    source, open arrows show steering source.

    Fig. 11 Close up photograph

    of

    unicycle car within

    Sphericle.

  • 8/9/2019 Rolling Robots Review

    7/14

    Rhodri

    H.

    Armour,

    Julian

    EV.

    Vincent: Rolling in

    Nature

    and Robotics: A Review

    20 1

    Fig. 12 Moving masses rolling principle, steering is not shown but can be considered if a 4th radial axis

    protrudes

    out

    of

    the drawing.

    of the robots based on this design have the same solution

    each with four radial axes with masses moving along

    them, but the deformab le robot uses a similar principle

    of moving the centre of gravity by reducing the length of

    shape memory alloy (SM A) actuators.

    Of

    all the possi-

    ble design conce pts this one is the most difficult to create

    linear movement in a plane since the movem ent of mul-

    tiple masses need to be controlled simultaneously. In

    some ways this can be considered as complex as a

    multi-legged robot with numerous actuators per leg.

    However, once suitable algorithms have been developed,

    remote control should be easy but autonomy would be

    complex due to the absence

    of

    fixed points for sensors.

    Using moveable masses does en able the centre of grav-

    ity to be located at the geometric centre of the sphere

    when required and this could be useful if a wind driven

    tumbleweeding motion is also desired.

    3.3.

    I The Spherobot

    The Spherobot from Michigan State Univer~ity ~'

    is a spherical omni-directional robot designed using the

    mobile mass concept. In this conceptual robot four

    radial spokes are fixed within a hollow sphere and

    positioned w ith equal angles between them in 3D space

    (Fig. 13). Along each spoke is a mass which can be

    moved from one extremity (centre sphere) to the other

    (sphere shell). Moving the masses radially along each of

    their spokes in a systematic way would enable the sphere

    to accelerate, roll at a constant speed, in a straight or

    curved path or stop at a specific point.

    No

    working

    prototype has yet been produced using this principle but

    a significant amount of work has been done on the mo-

    tion planning.

    Fig. 13 Spherobo t design concept 31.

    3.3.2 The August robot

    The August robot by Azad University of Qazvin

    and University of Tehran 41 s another example based on

    the moving mass concept which has been prototyped

    (Fig.

    14).

    It is autonomous via a radio link and control

    base station observing the sphere from the outside. The

    centre of mass of the device with the moving m asses in

    equally spaced locations is exactly at the geometric

    centre of the sphere thus the robot tends not to tip over

    on flat surfaces when stopped. The four radial spokes

    are located at 109.47 to one another. Each spoke has a

    1.125 kg weight positioned along it that can be moved

    using a stepper motor with 200 steps per revolution. An

    external camera is used as feedback for the positioning

    and control system. Th e device has successfully moved

    across a surface in straight and curved lines and a com-

    plete summary of its capability is shown in Table

    4.

  • 8/9/2019 Rolling Robots Review

    8/14

    202

    Journal

    of Bionic Engineerin? (2006)Vo1.3 No.4

    Fig.

    14 August design concept .

    Table 4 Performance capability of August robot

    ~

    Capabilit) Measurement

    for the SM A spokes is via

    a

    cable tether

    so

    it is not truly

    autonom ous. The robot has successfully crawled up a

    20 incline and jumped 80 mm vertically. The maxi-

    mum speed is 260 mm over 10

    s.

    Possibly the most

    interesting aspect of this design is that i t is also able to

    jump meaning that with further development and an

    on-board power supply, it could make an ideal all-terrain

    robot. complete summ ary of its capability

    is

    shown in

    Table

    5 .

    Table 5 Perform ance capability of Deformab le

    robot

    Capability Measurement

    Mass

    (kg) 0.003

    Diameter

    (m)

    0.04

    Speed

    (

    d s 0.026

    Static slope climbing ( ) 20

    Other Notes: Can jump

    0.08

    m (twice diameter)

    Mass

    (kg ) 12.13

    Diameter m i 0.29

    Speed

    (m/s

    (but

    this is not maximum)

    3.3.3 The Deformable robot

    The Deformable robot was developed by Rit-

    sumeikan University' I This 40 mm diameter tethered

    deformable robot can roll and jum p using shape mem-

    ory alloy (SMA) spokes within

    a

    soft rubber shell

    (Fig

    .15).

    Both cylindrical and spherical versions exist

    and are both based upon the moving mass design con-

    cept. Although these prototypes do not have additional

    masses as such. when a voltage is applied to the SMA

    spokes they contract. moving the centre

    of

    mass of the

    robot toward the rubber hoop.

    By controlling which

    S M A actuators con tract, the entire robot is able to move

    along. The robot weighs only 3g but the v o h g e supply

    3.4

    The fourth design principle emispherical

    wheels

    The fourth design type uses the two hemispheres as

    large all-encompassing wheels each driven individually

    from their centres (Fig.

    16).

    This is perhaps the simplest

    form of rolling robot.

    It has a single axle and uses two

    motors driving each wheel to control acceleration, de-

    celeration and steering. Th e torque of

    the

    driving motors

    is resisted by the pendulum mass of the enclosed chassis.

    Steering is achieved by rotating each w heel at

    a

    different

    speed

    or

    by counter rotating them for turning on the spot.

    Whe n stationary, there is

    a

    known fixed vertical hanging

    orientation of the central chassis allowing for the known

    location of sensors. This design does have

    a

    number of

    drawback s however. Each side of the sphere needs to be

    able to rotate independently of the other so there cannot

    Fig. 15 Photograph s of the Deformab le robot climbing a slope .

  • 8/9/2019 Rolling Robots Review

    9/14

    Rhodri

    H.

    Armour,

    Julian F.V. Vincent: R olling in Nature and Robotics: A Review

    203

    Side Front

    Fig.

    16 Spherical wheels concept, solid arrow shows rolling

    source, open arrow s show steering source.

    be an effective environmental seal between them allowing

    moisture and debris into contact w ith the control circuits.

    There is a specific axis through the d evice which m eans

    som e areas of its surface is not active and the device

    could become stuck if resting on those axle end points.

    The ends of the axle do however provide an excellent

    location for external sensors.

    3.4.1 The Kickbot

    Kickbot by Massachusetts Institute of Technol-

    ~ g y ' ~ ~ 's an autonomous robot which rolls around its

    environmen t and invited passers-by to kick it as a play-

    thing (Fig. 17). The project had three objectives:

    Build an autonomous robot from scratch, focus-

    sing on a new control system rather than an off the shelf

    one.

    Create a robot w here falling over is not a failure

    Experiment with various emergent behaviours.

    The prototype is based on a hamster ball and a

    polystyrene cha ssis. The two hem i-spheres of the ham-

    ster ball act as large wheels and sit either side of a central

    disc containing the control circuits, motors and sensors.

    The central chassis also includes a cou nterweight which

    is critical to the movement of the device, since as the

    counterweight is raised by the motors, the centre of

    gravity of the robot moves in front of its contact patch

    and the robot falls forward. The two hem ispheres d o not

    meet when assembled

    so

    the ball does not have a con-

    tinuous rolling surface although it is able to roll in all

    directions when acted upon externally or example

    after a luck from a bystander. Each hemisphere is driven

    mode (either self-orienting or orientless).

    Fig. 17 Photographof Kickbot 61.

    by a dedicated motor allowing a form of differential

    steering so is able to turn in place. There are no details

    of performance presented but given the objectives above

    the project was con sidered a success by its inventors.

    3.4.2 The Gravitational wheeled robot

    The Gravitational w heeled robot by K inki Univer-

    sity'17] is anothe r exam ple using this design principle.

    Although this device has two wheels, it does not react

    the torque of the two driving motors against the ground,

    rather it uses a hanging pendulum mass on the central

    chassis. The prototype has successfully climbed up

    slopes and over obstacles (Fig. 18). It does not comply

    with the entire definition of a rolling robot a bove but if

    the wheels were replaced by half spheres then it would

    have an entirely active outer surface. A summary of its

    capability is shown in Table 6 and Table 7.

    Fig.

    18

    Photograph of Gravitational wheeled r obot[171.

  • 8/9/2019 Rolling Robots Review

    10/14

    204

    Journal of Bionic Engineering (2006) Vo1.3

    No.3

    Table 6 Performance capability of Kickbot

    Capability Measurement

    Diameter

    m)

    0.28

    Other N otes: 0.03 m wheelbase

    Tahle

    7

    Performance capability

    of

    Gravitational

    Wheeled Robot

    Caoabi i

    t v

    Measurement

    ~~

    D i amet e r i m) 0.28

    Mass ( k p )

    Speed

    rid\) 0.45

    2 Wheels @ 1.13kg

    Static slope climbing (') 10

    Static obstacle climbing

    m)

    0.0

    3.5

    The fifth design principle yroscopic

    stabilisation

    The

    f if th

    design concept uses a gyroscope to stabi-

    lise the rolling axis of a wheel and a separate motor for

    forward movement (Fig. 19). An internal gyro spins in

    the same plane and direction as the direction of travel but

    when its axis is twisted to be non-parallel with the axis

    of the wheel,

    i t

    causes the wheel to lean over and so

    allows radiused turns. A separate motor rotates the outer

    wheel around the chassis and the gyro using the chassis'

    mass as a pendulum to react the torque and produce

    acceleration and braking.

    A

    design based on this prin-

    ciple cannot turn

    in

    place due to the fixed rotation ax is so

    turns can only be made when moving along. The inte-

    gral function of the gyroscope means that the device is

    unlikely to be light weight and the stabilisation and

    steering gyroscope require energy thus competing with

    the energy required to achieve motion. Again the device

    has a fixed axle through it mean ing that there are som e

    Sidc

    Front

    Fig. 19 Gyroscopic stabilisation concept. solid a rrow shows

    rolling source, open arrows

    show

    steering source.

    non-active parts of its surface but the gyroscope should

    be able to rotate the device off those points

    if

    required.

    Gyrover by C arnegie M ellon University 81 is a

    single wheel robot which is steered and balanced by a

    single gyroscope (Fig. 20). It is not completely spherical

    but the designers argue that its form gives better steering

    response and highe r speed d ue to its smaller frontal area.

    It has no problem recovering after falling on its side

    since rotating the axis of the g yro self-rights the device.

    Two radio controlled prototypes have been produced.

    The first is able to traverse laterally ac ross a slope of 45

    and has moved through a pile of gravel. The second

    prototype has much improved electrical efficiency

    mainly from improvements to the design of the gyro

    e.g. it is now enclosed in a vacuum. The second proto-

    type has also successfully moved across water by in-

    flating a large tyre and paddling over the surface.

    Summaries of the two Gyrover prototypes are shown in

    Table

    8

    and Table

    9.

    Fig. 20 Photograph of Gyrover in action 81.

    Table

    8

    Performance capab ility

    of

    Gyrover one

    Capability Measurement

    Diameter

    (m) 0.29

    Mass (kg) 2

    Static

    slope

    climbing ( )

    14

    Speed d s ) 2.78

    Traversed a 45 ramp moved

    through a pile of gravel

    ther Notes:

  • 8/9/2019 Rolling Robots Review

    11/14

    Rhodri H.

    A m o u r ,

    Julian F.V. Vincent: Rolling in Nature and Robotics:

    A

    Review

    205

    Table

    9

    Performance capability

    of

    Gyrover two

    Capability Measurement

    Diameter (m) 0.34

    Mass (kg) 2

    400%

    improved battery life from redes-

    igned gyro. Pneumatic outer tyre. In-

    flatable tyre for movement on water.

    Other Notes:

    3.6 The sixth design principle allast mass

    fixed axis

    The sixth design principle rotates an off-axis mass

    around within a sphere for linear rolling. Moving the

    same mass toward one end of the rotation axis, either

    linearly or by swinging, causes the device to lean and

    therefore steer around a comer (Fig.

    21).

    It is not pos-

    sible to turn in place w ith such a design sin ce the rotation

    axis is fixed within the sphere. There fore, if the device

    stops at a point, it must continue in that initial heading

    whilst making a radiused turn. This solution has a dis-

    tinct axis through the device and so some non-active

    surface points.

    3.6.1 The Roball

    The Roball by UniversitC de Sherbrooke is an

    autonomo us indoor mobile robotic toy (Fig. 22). It has

    been specifically designed to engage children in their

    play and has a sophisticated control system that deter-

    mines its behaviou r. It is the mobility platform that is

    interesting from a rolling point of view since it adopts a

    rolling motion when moving across flat surfaces. It

    moves by rotating its spherical shell around its internal

    components on an axis that travels through the device.

    Swinging a ballast mass from side to side enables the

    robot to steer a path and for it to fall onto one side with

    its face upperm ost for interaction with the childre n.

    Fig.

    22

    Photograph of R~ball ~'.

    The almo st spherical shell has a small flat band area

    around it which ensures straight line movement. It has

    been tested only on hard surfaces but is able to

    roll

    up a

    slope of

    8

    (Fig. 23) and climb over a

    0.79

    mm obstacle.

    Its speed has not been recorded in the literature but from

    videos it has been estimated as

    0.5

    m 4 ' .

    A

    complete

    summ ary of its capability is show n in Table 10.

    Fig.

    23

    Photograph

    of

    Roball climbing a ~lop e ~' .

    Table 10 Performance capability

    of

    Roball

    Caoabilitv Measurement

    Diameter (m) 0.15

    Speed d s ) 0.5

    Static slope climbing

    ( )

    8

    Mass

    (kg)

    1.81

    Static obstacle climbing

    (m) 0.00079

    Side

    Front

    Fig. 21 Ballast mass fixed axis concept, solid arrow shows rolling source, open arrows show steering source.

  • 8/9/2019 Rolling Robots Review

    12/14

    206

    Journal

    of

    Bionic Engineering

    (2006)

    Vo1.3

    No.4

    3.6.2 The Kotundux

    Rotundus (Fig.

    24)

    was developed at the Angstrom

    Space Technology Center but is now produced by a

    Swedish company and is marketed as a autonomous

    rolling spherical robot for security applications specifi-

    cally in hazird ous environm ents. The robot is based on

    the ballast mass fixed axis concept but details

    of

    its in-

    ternal construction. power supply and control system are

    axis concept (Fig. 26). The central axis is supported on a

    track around the equator of the sphere allowing the bal-

    last mass to rotate

    in

    any vertical plane. This prototype

    has been tested on grass and sandy surfaces and works

    well. It is able to climb slopes and roll over small ob-

    stacles as predicted by design calculations although the

    exact size of these has

    not

    been mentioned. A summary

    of its capability is shown in Table 12.

    held confidential.

    It

    has been tested on hard surfac es and

    on snow and is completely sealed from the environm ent.

    The device can travel at up to 15mph but no detailed

    information is presented

    in

    the public literature. sum -

    mary

    of

    its published capability

    is

    shown

    in

    Table

    11.

    Side Frotit

    Fig. 25 Ballast mass moving axis concept,

    solid

    arrow shows

    rolling source, open arrows show steering source.

    Fig.

    24

    Photograph

    of

    Rotundu s moving across a

    snow covered surface'20'.

    Table

    11

    Performance capability

    of

    Rotundus

    Capability Measurement

    Diameter (m )

    Speed

    ( i d s ) 6.7

    Other Notes:

    0.5 (estimated from photographs)

    Completely sealed

    from

    the environment.

    A

    product intended

    for

    security applica-

    tions

    3.7

    The seventh design principle allast

    mass

    moving

    axis

    The seventh solution is sim ilar to the sixth solution,

    except that the axis

    is

    capable of being moved w ithin the

    sphere and this achieves a change

    in

    direction (Fig. 25).

    This enables the device to move off

    in

    any direction after

    stopping but it is not possible to make radiused turns at

    all since the direction of travel is always perpe ndicular to

    the rotation axle.

    The most recent prototype of the Rollo spherical

    robot by Helsinki University

    of

    Technology

    &

    Rover

    Com pany Ltd ). is based on the ballast inass moving

    Fig.

    26 Photographs

    of

    Rollo

    prototype

  • 8/9/2019 Rolling Robots Review

    13/14

    Rhodri H. Armou r, Julian F.V. Vincent: Rolling in Na ture and Robotics:

    A

    Review

    207

    Table 12 Performance capability of Rollo prototype three

    Capability Measurement

    Diameter (m) 0.24

    Mass (kg) 3

    Speed

    d s )

    0.5

    Other Notes: Peak current consumDtion

    0.8

    A

    4 Conclusion

    From these different designs, it can be seen that

    there are numerous solutions to the problem of rolling

    and the most suitable design solution depends on the

    specific application and requirements chosen by the

    developers. If autonom y is impo rtant, then the most

    robust solution is desirable and that is likely to be the

    third or seventh solutions, Mobile masses or Ballast

    mass on a moving axis. If movement in a confined space

    is required then the first (Sprung central member), third

    (Mobile masses) or seventh (ballast mass moving axis)

    designs are suitable since they can turn in place rather

    than having to make a curved turn. If wind driven

    tumbleweeding is required as another mobility solu-

    tion, then only the third design (Moving masses) is

    suitable since the centre of gravity can lie at the centre

    of the devic e. If move ment in wide open and flat

    areas is required, then a more simple solution can be

    used such as the second (C ar driven), fifth (Gyroscopic

    stabilisatio n) and sixth (Ballast mass fixed axis) types. If

    mobility across water is important then only enclosed

    solutions can be used which rejects the fourth type

    (Hemispherical wheels).

    If each of the designs is scored on a scale of 1 to 3

    (l=poor, 2=average, 3=good) against a series of pa-

    rameters defining the suitability for use in an unknow n

    and varied but relatively smooth terrain such as the

    surface of Mars (Fig. 27) it is possible to determine the

    best overall design for that specific application. The list

    of parameters is not comprehensive and the scoring

    system is somewhat subjective

    so

    re-evaluation may be

    required once a working environment is selected.

    Al-

    though adding the scores may not give the perfect design

    for a specific application, it does give a good overview

    of the potentially best all-rounders. The scores are

    shown in Talbe 13.

    Fig.

    27

    Surface

    of

    Mars.

    Table 13 Design principle scoring

    Parameter

    1

    Design

    2ld

    Design

    3rd

    Design

    4Ih

    Design 5 I h Design 61hDesign 7IhDesign

    Mobile . Spherical Gyroscopic Ballast mass Ballast mass

    wheels stabilisation fixed axis moving axis

    prung Car driven

    central axis

    Obstacle Avoidance

    Simplicity

    of

    design

    Easy of autonomy

    Weight

    Solid or flexible roll-

    ing surface

    Tumble-weeding

    Slope Climbing

    Obstacle Climbing

    Movement in enclosed

    areas

    Solid

    No

    2

    2

    3

    1

    2

    3

    Flexible

    possible

    No

    3

    3

    2

    Flexible

    Yes

    3

    3

    3

    Solid

    No

    1

    1

    2

    Flexible

    N o

    2

    2

    1

    Flexible

    Possible

    2

    2

    I

    Solid

    Possible

    2

    -

    3

    Total 14 15 17 13 14 I 16

  • 8/9/2019 Rolling Robots Review

    14/14

    208

    Journal of Bionic Engineering (200 6) Vo1.3 No.4

    From this summary.

    it

    can be determined that the

    mobile masses solution perhaps offers the most flexi-

    bility for its use. Given that the deform able robot pre-

    sented above (section 3 . 3 . 3 is also able to jump, then

    using this type of solution offers an interesting all-terrain

    mobility system

    so

    long

    as

    the energy supply situation

    can be resolved.

    This work has been conducted as part of the de-

    velopment of a rough terrain mobility system based on

    rolling and jumping movements at the University of

    Bath.

    Acknowledgement

    Thanks to Professor Julian Vincent for his guidance

    throughout the progression

    of

    this work.

    References

    [ I ]

    Garcia-Paris M. Deban S M. A novel antipredator

    mechanism in salamanders: Rolling escape i n hydro-

    mantes platycephalus.

    JoLrniril .fHerpetology. 1995,

    29.

    149-1

    5

    1

    [2 ] Henschel J. Feature creature he golden wheel spider.

    GobnBeb Times. 2005. 1

    3.

    [ 3 ]

    Antol J . Calhoun P. Flick J, Hajos Ci Kolacinski R.

    Minton D. Owens R, Parker J. Low Cost Mars Surface

    Exploration: The Mars Tumbleweed. NASA Technical

    Report, NASA Langley Research Center, NASAlTM-

    2003-2 I24

    1

    [4]

    Antol J, Chattin R

    L.

    Copeland B M. Krizan

    S

    A. The

    NASA

    langley mars tumbleweed rover prototype.

    Pro-

    cerclirig of

    the 44th

    A I M

    At~rospiic.e

    Sciences

    Meeting

    mid

    E.ukibit.

    Reno. Nevada. USA. 2006. AIAA-

    2006-0064.

    1.51 Jakubik P. Suomela J. Vainio

    M ,

    Yilkorpi T, Halme A.

    ARIADNA A04532-03/6710.

    B io log i cu l l~ Inspired

    So1rrrion s iw Robotic.

    Sirrf2ifirc e

    Mobility.

    Helsinki Univer-

    sity

    ot

    Technolo:) utoma tion techno logy laboratory.

    Helsinki. 2004.

    161

    Brackenbury

    J

    Caterpillar kinematics.

    Nature,

    1997.390.

    453.

    [ 7 ]

    Full R. Earls

    K.

    Wong M, Caldwell R. Locomotion like a

    u:heel'l NLl tuw. 1993.365.

    405.

    [8] C;ildwell

    R I-. A

    unique form of locomotion in a

    stoniatopod ackward somersaulting.

    Ncitiire.

    1979.

    282.71-73.

    [9] Halme A. Schonberg T, Wang

    Y.

    Motion control of a

    spherical niobile robot. 4th IEEE Intem~itioric i l orkshop

    on

    Advn i i ccd Morion Control. Mie. Japan. 1996, 1

    259-264.

    [ 101 Reina

    3,

    Foglia M. Milella A, Gentile A . Rough-terrain

    traversability for a cylindrical shaped tnobile robot.

    Pro-

    ceedirigs of rhe IEEE

    I i i t e i -n~ i t i onul Conference

    on

    Mechntronics.

    IEEE. New York, USA, 148-153.

    [ 1 I 1 Rover Company Limited. Ball-shaped Robot. St Peters-

    burg Russia. 1996.

    [ 121

    Bicchi A. Balluchi A , Prattichizzo D . Gorelli A. Intro-

    ducing the '5PHERICLE': An experimental testbed for

    research and teaching in nonholonomy.

    IEEE

    Infer-ria-

    t i o r i onference

    on Robotics cord ALitornutiori,

    Albu-

    querque, NM, USA, 1997.3,2620-2615.

    [ 131

    Mukherjee R. Minor M A. Pukrushpan J T. Simple mo-

    tion planniny strategies for sphero bot: A spherical mobile

    robot.

    Procretl i t igs

    qf the 38th Cor@xwce on Decision

    mid Confro/.

    Phoenix. AZ.

    USA.

    1999.

    3,

    21 32-2 137.

    [

    141

    Javadi A H A, Mojabi P. Introducing

    august:

    A novel

    strategy for an omnidirectional spherical rolling robot.

    IEEE Interrintiotrnl Conference

    on

    Robotics and Airto

    itintion,Washington. DC. USA. 2002, 3527-3533.

    151

    Sugiyama Y, Hirai S . Crawling and jumping by a cle-

    fomiable robot.

    IEEE/RSJ I n / e r t i t i t i n r i ~ i l

    Confererice

    o n

    Iiirelligerit Robots ~ i n d

    ysterns. Sendal. Japan.

    2004. 4.

    3276-3?8 1

    161 Batten

    C ,

    Wentzlaff D. Kickbot: A Spherical Autonomous

    Robot. Massachusetts Institute of Technology,

    [2OCm3-X1,

    http://www.iiiit .edu/-cbatten/wor~kickbot-enibint01

    pdf

    [ 171 Kiyoshi Ioi H 1 Ayanobu Murakam i. Design

    of

    a gravi-

    tational wheeled robot.

    Advuriced Robotics, 2002. 16

    785-793.

    [I81

    Brown H B Jr ,

    Xu

    Y. A single-wheel. gyroscopically

    stabilized robot. IEEE Ir~ternrrrional

    Conference

    on Ro-

    botics

    u i i d

    Airrornnrion,

    Minneapolis, MN, USA, 1996, 4

    3658-3663.

    [

    191

    Michaud

    E

    Laplante J-E Larouche H, Duqiiette A, Caron

    S . Letourneau D. Masaon P. Autonomous spherical mo

    bile robot to child development studies. IEEE Trcrnsac-

    tions

    Oil

    s\..stPIrl.s.

    Mnn, f i n

    C~Dernefic .S .

    005.

    35.

    71480 .

    [20]

    Rotundus Rotundus. Durable mobile robot for outdoor

    serveillance. Upp sala University. Angsrriini Space Tech-

    nology Center.

    2002.