humanoid robotics at waseda university -...
TRANSCRIPT
Atsuo TakanishiDept. of Modern Mech. Eng./Humanoid Robotics Institute, Waseda University
Humanoid RoboticsHumanoid Roboticsat Waseda Universityat Waseda University
Tutorial on Humanoid TechnologiesTutorial on Humanoid Technologies
The 2The 2ndnd EC FP7 RoboSoM Project MeetingEC FP7 RoboSoM Project MeetingFebruary 18, 2010February 18, 2010
Prof. Ichiro Kato Prof. Ichiro Kato (1925(1925--1994)1994)
WasedaWaseda’’s Achievementss Achievementsin Humanoid Roboticsin Humanoid Robotics
led by the late Prof. Ichiro Katoled by the late Prof. Ichiro Kato
WABOT-1 (1973) WABOT-2 (1984)
Why Humanoid Robotics?Why Humanoid Robotics?To Build Human Model to Understand Human from Robotics View Point: “Robotic Human Science”To Make Robotics as One of the Social Technologies based on the Human Model to Support Elderly Dominated Society and to Support Education, Etc.Grand Challenge for Roboticists: A human is considered to be an extreme robot both in motional functionality and in intelligential functionality.
Humanoid FamilyHumanoid Familyin WABOTin WABOT--HOUSE ProjectHOUSE Project
WABIANWABIAN--2 as a Robotic Human Simulator and2 as a Robotic Human Simulator anda Measuring Device for Quantitative Evaluationa Measuring Device for Quantitative Evaluation
of Handicapped/Elderly Supporting Machines/Robotsof Handicapped/Elderly Supporting Machines/Robots
height : 1.5 [m]weight : 55 [kg]
Total DOF : 41leg : 6×2waist : 2trunk : 2arm : 7×2neck : 3
Yu Ogura TeamYu Ogura Team
2 DOF
3 DOF
1 DOF
3 DOF
22--DOF PelvisDOF Pelvis
New LegNew Leg--Body Mechanism Body Mechanism Inspired by Human WalkingInspired by Human Walking
Modeling of Pelvis Motion
2 DOF hips
•Stretch walking •Obstacle avoidance
2DOF
1DOF
3DOF
3DOF
Pelvis-Waist Mechanism:2DOF
Most Tough Parts in Design:Most Tough Parts in Design:WaistWaist--PelvisPelvis--HipHip--KneeKnee
WABIANWABIAN--2 Designed Based2 Designed Basedon the Human Dimensionson the Human Dimensions
More HumanMore Human--like Walking by like Walking by WABIANWABIAN--2R with a 50cm2R with a 50cm--StepStep
0.50[m/step], 0.96[s/step]
0.40[m/step], 0.96[s/step]
HumanHuman--Like Outdoor Walking Like Outdoor Walking
Two Peaks in Vertical Floor Two Peaks in Vertical Floor Reaction ForceReaction Force
Human WalkingHuman Walking
Knee Bended Knee Bended WalkingWalking
Knee Extension WalkingKnee Extension Walking
WABIANWABIAN--22’’s Walking Assisted by s Walking Assisted by Hitachi Walker for Elderly/HandicappedHitachi Walker for Elderly/Handicapped
Bending 0.2[m] Knee stretched 0.2[m]
Real World Simulations of a human assisted by the Walker
9th,FEB,2007 2006年度 修論発表
■ Optimal Pattern
Subject
Applications of WasedaApplications of Waseda’’s Bipedals BipedalWalking Robot TechnologyWalking Robot Technology
METIMETI HRPHRP--11ZMP PINO2ZMP PINO2 TMSUKTMSUK//Waseda WLWaseda WL--1616
SONY QRIOSONY QRIO
TMSUKTMSUK//SANYO BanryuSANYO Banryu
SHINPOSHINPO「「新歩新歩」」: A Brother of WABIAN: A Brother of WABIAN--2 2 Walking in Niigata Natural Science Walking in Niigata Natural Science Museum Resulting of Collaboration with Museum Resulting of Collaboration with TMSUKTMSUK
Samurai Humanoid Samurai Humanoid KIYOMORIKIYOMORI::22ndnd Brother of WABAINBrother of WABAIN--2 Visited Munakata Shrine 2 Visited Munakata Shrine
Prayed for the Safety of All the Robots in the WorldPrayed for the Safety of All the Robots in the World
33--Moment Compensatory Bipedal Moment Compensatory Bipedal Walking Control Extending the ZMP Walking Control Extending the ZMP CriteriaCriteria
Y
PP
RR
YY
PP
RR
m ( ) ( )
{( ) }
ii
All Particles
i ZMP i
Fk ZMPk
All Points
k k
∑
∑
− × −
− − × + =
r r r G
r r F M
&&
0
GravityGravity
FrictionFriction
March 18, 2005March 18, 2005
How to walk with balance
ZMP: a point where the total forces and moments acting on the robot are zero
ZMP trajectory: set arbitrarily within the support polygon
Stable region
Compensatory motion to keep the balance, computed using ZMP equation
Foot trajectory ZMP trajectory
Stability criterion for biped robot based on ZMP (Zero Moment Point)
March 18, 2005March 18, 2005
ZMP Equation and the Coordinate Systems
XYZO −
XYZO −
m ( ) ( )
{( ) }
ii
All Particles
i P i
Fk Pk
All Points
k k
∑
∑
− × −
− − × =
r r r G
r r F M
&&
+ + T 0
m ( ) ( )
{( ) }
ii
All Particles
i ZMP i
Fk ZMPk
All Points
k k
∑
∑
− × −
− − × + =
r r r G
r r F M
&&
0
ZMP Equation
m ( ) {
2 ( )}
{( ) }
ii
All Particles
i ZMP i
i i i
Fk ZMPk
All Points
k k
∑
∑
− × + −
+ × + × + × ×
− − × + =
r r r Q G
r r r
r r F M 0
&& &&
& &ω ω ω ω
March 18, 2005March 18, 2005
Iterative Computation Algorithmto Converge ZMP/Moment Error
Compute walking pattern using compensatory waist motion
Compute moments Mgenerated by lower-limbs
Compute compensatorytrunk and waist motion
using linear model (FFT&IFFT)
Compute moment errors (eM) between calculated ZMP and planned ZMP
eM < εM ?Yes
No
Compute walking pattern using initial waist motion
Complete walking pattern
Initial parameters;Foot trajectory, Initial trunk and waist trajectories,ZMP trajectory, and εM (a tolerance level moment error)
nLL EMM +=
moment errors (eM)
)1(1 −− ⋅+= nnn Gain eMEE
Iterative computationto derive motion pattern
WWTT yxyx ,,,
WABIANWABIAN--2R Goes to Streets, 2R Goes to Streets, but Many Problems!but Many Problems!
Special Economical Zone for Robot Test and Experiment in FukuokaSpecial Economical Zone for Robot Test and Experiment in Fukuoka
We need more efforts for higher usability in the real worldWe need more efforts for higher usability in the real world
Open Experiment at Okubo Campus of Waseda in 2006
Biped Robot That CanBiped Robot That CanCarry a Human: WLCarry a Human: WL--1616
WLWL--16 Goes to the US16 Goes to the USCarrying 170 People in NextFest 2007Carrying 170 People in NextFest 2007
Towards Vocal Humanoid Towards Vocal Humanoid Robot: WT Series (2000Robot: WT Series (2000-- ))
Human Vocal OrgansHuman Vocal Organs
Robot Vocal OrgansRobot Vocal Organs
Jump!Jump!
Kotaro Fukui TeamKotaro Fukui Team
Sound SourceSound Source
Vocal CordsVocal Cords
Out PutOut Put
LipsLips
Soft PalateSoft Palate
TongueTongue
Airflow from LungsAirflow from Lungs
Palate Palate
Nasal Cavity Nasal Cavity
WT-5’s Mechanism
xx
zz
yy
Human Vocal Cords
戸幾一郎:発声機構の面よりみた喉頭の病態生理、耳鼻臨床(戸幾一郎:発声機構の面よりみた喉頭の病態生理、耳鼻臨床(19661966))
※※
Structure of Vocal CordsStructure of Vocal Cords
Vocal Fold VibrationVocal Fold Vibration
Lower PartLower Part
Upper PartUpper Part
Cover Cover (Nonmuscle Tissue )(Nonmuscle Tissue )
BodyBody(Muscle Tissue )(Muscle Tissue )
Phase Difference in Glottal Closure Phase Difference in Glottal Closure Between Upper Part and Lower PartBetween Upper Part and Lower Part
WT-5’s Vocal Cords
35353030
5050
xxyyzz
1515 2424
zz
frontfront sideside backbacktoptop
Developed New Vocal Cords Mimicking Developed New Vocal Cords Mimicking Human Biomechanical StructureHuman Biomechanical Structure
2222
1616
yyzz
xx
Vocal Fold Vocal Fold VibrationVibration
** Thermoplastic Rubber by Kuraray CoThermoplastic Rubber by Kuraray Co
Using the Thermoplastic Rubber Using the Thermoplastic Rubber ““**SeptonSepton””Same Size as Human (Adult Male)Same Size as Human (Adult Male)
Voiced/Unvoiced Sounds Switching Mechanism
CloseClose OpenOpen
OpenOpenCloseClose
Glottal Opening/Closing Control Mechanism by Glottal Opening/Closing Control Mechanism by using Arms Mimicking Human Arytenoid Cartilagesusing Arms Mimicking Human Arytenoid Cartilages
xx
yyzz
Modeling of Elastic Tongue
Elastic Tongue ShapeElastic Tongue Shape
Rubber( EPDM )Rubber( EPDM )
Applicable to All Elastic BodyApplicable to All Elastic BodyApplicable to All Elastic Body
ControlControlPointPoint
Modeling of Tongue asModeling of Tongue asConnection ofConnection of Rigid LinksRigid Links
RigidRigidLinksLinks}}
ControlControlPointPoint
SimulateSimulate
Modeling of Elastic Tongue
( )11, yxx
1l
nq
y
Free End
α
1q
nl( )nn yx ,
Virtual Link
1k
1+nk
Virtual Link
nk
2q2q2q2k
yx FFF ,
⎩⎨⎧
−=−=
yny
xnx
PyPx
δδ
Virtual DisplacementVirtual Displacement by Force F,by Force F,( )yx δδ ,
Potential Energy (by Potential Energy (by Virtual Work PrincipleVirtual Work Principle))
yyxx
n
iii FFqkqkU δδπ
+++⎟⎠⎞
⎜⎝⎛ −= ∑
+
=
1
2
22
11 21
221
Referred by HookeReferred by Hooke’’s Laws Law
(3)(3)
(4)(4)
Kinetic Energy (Stationary)Kinetic Energy (Stationary)LagrangeLagrange‘‘s Equations of Motion ( )s Equations of Motion ( )UTL −=
0=T
(5)(5)
( )nqqq ,,, 21 …Link AnglesLink Angles
NewtonNewton--Raphson MethodRaphson Method
Towards Vocal Humanoid Towards Vocal Humanoid Robot: WT SeriesRobot: WT Series
Human Vocal OrgansHuman Vocal Organs
Robot Vocal OrgansRobot Vocal Organs
Jump!Jump!
Kotaro Fukui TeamKotaro Fukui Team
Flutist Robot for SimulatingFlutist Robot for SimulatingHuman Flute Playing:WFHuman Flute Playing:WF--44
(1990(1990-- ))
GifuGifu--Waseda WABOTWaseda WABOT--HOUSE ProjectHOUSE Project
Mr. Wakamatsu:Mr. Wakamatsu:Technical AdvisorTechnical Advisor
W a s e d a U n i v e r s i t y第25回日本ロボット学会学術講演会 2007年9月14日
ソリス ホルへ
Lips:3-DOFsy
z
x
Waseda Flutist No.4 Refined IVLips Mechanism Design
z
W a s e d a U n i v e r s i t y第25回日本ロボット学会学術講演会 2007年9月14日
ソリス ホルへ
Waseda Flutist No.4 Refined IVLips Mechanism Motion
W a s e d a U n i v e r s i t y第25回日本ロボット学会学術講演会 2007年9月14日
ソリス ホルへ
Tongue
Oral cavity
yx
x
z
Film of SEPTON
z
Waseda Flutist No.4 Refined IVTonguing Mechanism Design
Tongue:1-DOF
W a s e d a U n i v e r s i t y第25回日本ロボット学会学術講演会 2007年9月14日
ソリス ホルへ
Waseda Flutist No.4 Refined IVTong Mechanism Motion (Tonging)
Waseda University
83 [m
m]
45 [mm]
x y
z
Vibrato Mechanism
Human-like Vibrato Mechanism
Human Vocal Cords
ProfessionalIntermediate
Vibrato Mechanism
Low AmplitudeHigh Amplitude
4 [Hz], A4x
y
Mukai(1987)
W a s e d a U n i v e r s i t y第25回日本ロボット学会学術講演会 2007年9月14日
ソリス ホルへ
Lung Mechanism
ShaftShaft
AC ServomotorAC Servomotor
LinkLink
yy
zz
413 [mm]413 [mm]
398
[mm
]39
8 [m
m]
yy
zz
Takanishi Lab. 14th Oct. 2002The Robotics Society of Japan
Finger Mechanisms
DC Motor
pulleyOuter TubeTeflon-lined Wire (Chukoh Chemical Industries LTD)
Link mechanism
Supported by SokidWorks K.K
VV : Total Sound Volume: Total Sound Volume
HH : Average of Harmonics: Average of Harmonics
HsHs : Average of Semi: Average of Semi--HarmonicsHarmonics
HdHd : Even Har. : Even Har. –– Odd Har.Odd Har.
ItemsItems
W1, W2 : WeightinsW1, W2 : Weightins
FF(Note Pitch)(Note Pitch) = = VV + w+ w1 1 ×× ((HH--HsHs) + w) + w2 2 ×× HdHdRR RRR [Pitch Range] = {Low, Middle, High}R [Pitch Range] = {Low, Middle, High}
What is good flute sound?:What is good flute sound?:Evaluation IndexEvaluation Index
WFWF--44’’s Flute Play:s Flute Play:The Flight of the BumbleThe Flight of the Bumble--BBeeee
(Rimsky(Rimsky--Korsakov)Korsakov)
Duet between Human and WFDuet between Human and WF--44
Saxophonist Robot: WASSaxophonist Robot: WAS--22
Part Degrees of Freedom
Pump 1 DOFValve 1 DOFLip 3-DOFTongue 1-DOFFingers 16-DOFs
Total : 22 DOFsTotal : 22 DOFs14
0014
00
WASWAS--22
Lip Lip & Tongue& Tongue
FingersFingers
PumpPump
ValveValve
New Mouth & Hand MechanismNew Mouth & Hand Mechanism
TEAM FLUTE
Waseda universityRSJ2009New Mouth Mechanism
Overview
Part No. of Degree
Upper lip 1
Sideway (1)
Lower lip 1
xyz
WAS-2
TEAM FLUTE
Waseda universityRSJ2009Development of a Human-Like Hand
Video
Demo Play of WAS-2:Sing Sing Sing
Emotion Expression Humanoid Emotion Expression Humanoid EYEEYE--ChanChan::WEWE--4RII for Modeling Human Mind4RII for Modeling Human Mind
PartNeckEyes
EyelidsEyebrows
LipsJaw
Arms
Lung
68
34
411
18
DOF
Waist 2
Total 59Hand 12
WE-4 (Waseda Eye No.4)Expressions
Eyebrows
Eyelids
Facial Color
Lips
Voice
Neck
Waist
Lung(Breath)
Sensors
Visual(CCD Camera)
Auditory(Microphone)
Olfactory(Gas Sensor)
Tactile(FSR)
Temperature(Thermistor)
Control System and PersonalityControl System and Personality
Total Personality
Intelligence
Environment
RecognitionBehavior
EmotionSensing
PersonalityExpression Personality
SensingMotion Reflex
Equations of EmotionMt+Δt = Mt + ΔM
Mt = (at, pt, ct)ΔM = (Δa, Δp, Δc)
at : Activationpt : Pleasantnessct : Certainty
arousal
asleep
pleasant unpleasant
certain
uncertain
Mt
Mt+ΔtΔM
Equations of EmotionEquations of Emotion
Pleasantness
Activ
ation
Cer
tain
ty
Happiness
Surprise
Happiness
Disgust
Sadness
Fear
Sleep
Sleep
Neutral
Neutral Anger
*Paul Ekman, Wallace V. Friesen: Facial Action Coding System (1978 )
Emotion MappingEmotion Mappingin Mental Spacein Mental Space
E = {Happiness,Anger,Surprise, Sadness,Fear,Disgust,Neutral}
emotion ∈ E
From Eq. of MotionFrom Eq. of Motionto Eq. Emotionto Eq. Emotion
M : Emotional Inertia Matrix(情動慣性)
Γ : Emotional Viscosity Matrix(情動粘性)
K : Emotional Stiffness Matrix(情動弾性)
Emotional Coefficient Matrix(情動係数行列)
)(tFxkxxm =++ &&& γEquation of Motion(運動方程式)
Equations of Emotion(情動方程式)
EAFKEEEM =+Γ+ &&&
E: Emotion Vector, FEA: Emotional Appraisal
Emotional DynamicsEmotional Dynamics(情動力学)(情動力学)
Time
Em
otio
n V
ecto
r
0
**
*
2 ii
ii
kmγζ =
0det 2 =− MK niω
Emotional AttenuationCoefficient(情動減衰係数)
Emotional Natural AngularFrequency(情動固有周波数)
iTii φφγ Γ=*
iTii Mm φφ=*
iTii Kk φφ=* rEigenvectoi :φ
*
*
i
ini m
k=ω
Stimulus
Visual
Tactile
AuditoryTemperature
Olfactory
SensationLoose Sight of the Target
Discover the TargetDazzling LightTarget is Near
PushedPushed Strongly
StrokedHit
Loud SoundHeat
AlcoholAmmonia
Cigarette SmokeNo Sense
No Stimulus* “→0” means to converge at “0”
Δa ΔcΔp
++++++
++++
+++
-
-
-
--
--
--
-
-
--0
0000000000
00
0
0
00
→0*→0*
Sensing Personality TableSensing Personality Table
11
11
11
10
0=
Eo NeutralEo SurprisedEo HappinessEo SadnessEo AngerEo FearEo Disgust
Ei NeutralEi SurprisedEi HappinessEi SadnessEi AngerEi FearEi Disgust
Eo
ExpressionPersonality
MatrixEi=
Expression PersonalityExpression Personality
Ei: Input Emotion, Eo: Output Emotion
Demos of EYEDemos of EYE--ChanChan
Additional DemosAdditional Demos
More Behaviors Addition of Consciousness
WABIANWABIAN--2 + EYE2 + EYE--ChanChan= KOBIAN= KOBIAN
Emotional ExpressionsEmotional Expressionsof KOBIANof KOBIAN
The EndThe End
Thank you!Thank you!