Japan Advanced Institute of Science and Technology

Japan Advanced Institute of Science and Technology

JAIST Repository

https://dspace.jaist.ac.jp/

Title

に関する研究

Author(s)

Citation

Issue Date

Type

Text version

URL

Rights

Description

Takehito Sugawara

Scho ol of Information Science

Japan Advanced Institute of Science and Technology

February 14, 1997

Keywords: mobile robot, rob ot manipulator, visual servo, singularity avoidance.

In this thesis, we consider a visual servo control problem of mobile manipulator(we

callvisual servocontrolmobilemanipulatoras rob otwithintegratedlocomotion andma-

nipulatorinfuture). Weprop oseacontrolmethodforveryuseful themobilemanipulator

of end-eector mounted the camera toposition arobot arm.

The robot manipulator and the mobile rob ot has b een develop ed from request of

industrial so cietyas automation production. Construct of line of afactory becomes only

a few of a limit from moving the robot manipulator and easy alteration of line. But,

workas movedis hard. Because workpointmotionmake awrongforecast of unexpected

motion than work such as no mobile manipulator. Visual feedback can help overcome

it problem, because a vision sensor is a non-contact and can provide information on a

much larger area of the work-cell than a force sensor provides. Hence it is necessary

to the mobile robot mounted the robot manipulator mounted the robot manipulator of

end-eector mounted camerato position arobot arm.

The previous researches isdescribedin the pages that follow.

There is researches which incorporated into the manipulability measures for singu-

larity avoidanceand controlled position and orientation.

There isresearches using vision system latermoving for regulation.

There is researches which control the pose of the control object that incorp orated

with mobilerobot, robot manipulator and camera.

But, we found that in the above the previous researches the use of camera and the

mobilestructuremounted therobotmanipulatordose not alwaysgivegoodperformer. It

dose not transact it information with vision system is expected target motion. When a

manipulatorencountersakinematicsingularityduringCartesianmotion,the visualservo

Copyright c

1997byTakehito Sugawara

with a robot with integrated locomotion and manipulator, it is also important that the

robotmanipulatormaintainsacongurationthatallowsmotioninalldirectionsofpossible

targetmotionwithoutrequiringextremelylargejointvelocityfromanyactuator,because

thefuturemotionofthetargetmayincludeimpreciselyknown. Thisalsorequiresthatthe

robot manipulator shouldnot be near singularities. Typically,when visual servo targets

with arobotwith integrated locomotion and manipulatoritis necessarytoconstrain the

allowablevisualservoregionsoftheworkspacewheretherenodangerthatthemanipulator

passes near kinematicsingularities.

Hence, In this thesis, we consider a visual servo control of mobile manipulator. The

mobile manipulator mounted the robot manipulator of end-eector mounted camera to

position arobotarm, and weconsidervisual servoisto use visual information tocontrol

the poseof the robot's end-eector relative toa target objector a setof target features.

We consider redundant robot manipulator with respect to kinematics from the mobile

robot mounted the robot manipulator. Hence the mobile robot, the robot manipulator

and camerashall regardone objectthat itcontrolconsider. Acontrolsubjectis behavior

at the same time with respect to the mobile robot, the robot manipulator and camera.

Because, visual servo can apply any time as always moved a target. In addition, for

make use of fully merit of the mobile robot mounted the robot manipulator, it add an

idea of the manipulability measure. An arm pose decides with valuation function used

manipulability measure so that is kept an optimal arm pose, or the robot manipulator

performs control of singularity avoidance. And, When we consider modeling robot with

integrated lo comotion and manipulator, each motionof mobileparts, robot manipulator

parts and camera part give consideration with respect to mutual motion give inuence.

And, we must verify a proper technique with respect to visual servo of the the mobile

robot mounted the robot manipulator with simulation with reference to velo city level

control. These are a purpose ofthis thesis.

We dened several functions for modeling the robot with integrated locomotion and

manipulator.

Afunctionofmobilerob otchangesthewheelvelo cityintothemobilerob otvelocity.

Afunctionofrobotmanipulatorchangesthejointvelo cityintotheCartesianvelocity

A function of camerachangesthe Cartesian velo city intothe feature velocity

A functiongiving abovefunctions changesthe jointvelo city intothe featurevelo city is a

combinationeachfunctionof frobotmanipulator+camera g,fmobilestructure+camera

g, f mobile structure+robot manipulator g. But, It function adds to terms consider a

mutual relationship.

The mobile robot, in this thesis, has 2 xed wheels(drive wheels) on the same axle

and 1 castor wheel(2DW1C), is the autonomous mobile manipulator moving in the x-y

horizontal plane. An arbitrary inertial base frame 6

w

is xed in the plane of motion,

while a frame 6

v

is attached tothe mobile rob ot. A movingdirection is y axis. A frame

of the robot manipulator is same mobile rob ot frame 6

v

. Camera mounted the robot

manipulator ofend-eector, it is acamera frame 6

c .

horizontal plane,provided atargetiscarryconveyorbelts andatargetvelo city isknown.

This thesis intro duce furthermore to idea of the manipulability measure used to aid in

visual servo,and it providesothat the robotmanipulator maintainsaconguration that

allowsmotioninalldirectionsofp ossibletargetmotionwithresp ecttounexp ectedmotion

and unknown environment,namely, itp erforms singularity avoidance.

In this thesis, a control method uses dynamic look-and-move system. This control

architecture is hierarchical and uses the vision system to provideset-point inputs to the

joint-level controller. Nearly all implemented system adopt the dynamic look-and-move

approach. For reasons, Firstly, the relatively low sampling rates available from vision

makedirectcontrolofarobotend-eectorwithcomplex,nonlineardynamicsanextremely

challengingcontrolproblem. Using internalfeedbackwith ahigh sampling rate generally

presentsthevisualcontrollerwithidealizedaxisdynamics. Secondly,manyrob otsalready

haveaninterfaceforacceptingCartesianvelo cityorincrementalpositioncommands. This

simplies the construction of the visual servo system, and also makes the methodsmore

portable. Thirdly, look-and-moveseparates the kinematicsingularities of the mechanism

fromthevisualcontroller,allowingtherobottobeconsideredasanidealCartesianmotion

device. Sincemanyresolvedratecontrollershavespecializedmechanismsfordealingwith

kinematic singularities,the system designis again greatly simplied.

Asaresultofthis thesis,rob otwithintegratedlocomotionandmanipulatorisderived

by a model froma function giving inconsiderationof eect with respect to eachmotion.

Weproposedacontrolmethodinconsiderationofmanipulabilitymeasure,andperformed

simulation of visual servo in consideration of manipulability measure for it conrm ap-

propriateness. As a consequence, robot with integrated locomotion and manipulator in

consideration of manipulability measure has vast work space and a conguration that

allows motionin all direction of p ossible target motion. Hence we prop osed visual servo

of rob ot with integrated lo comotion and manipulator considers useful than usual visual

servo. And,again, itis fully appropriate system with reference tovelo city levelcontrol.