assignment 1 on the web page: jkh/anim ...jkh/anim_class/kinematics.pdf · forward and inverse...
TRANSCRIPT
Ann
ounc
emen
ts
Ass
ignm
ent 1
on
the
web
pag
e:
ww
w.c
s.cm
u.ed
u/~
jkh/
anim
_cla
ss.h
tml
Tes
t log
in p
roce
dure
NO
W!
02/2
2/02
For
war
d an
d In
vers
e K
inem
atic
s CO
MP
UT
ER
AN
IMA
TIO
N
15-4
97/1
5-86
1
02/2
6/02
Par
ent:
Cha
pter
4.2
Gira
rd a
nd
Mac
ieje
wsk
i 19
85
Zha
o an
d B
adle
r 199
4
Kin
emat
ics
•La
st tw
o cl
asse
s: h
ow to
inte
rpol
ate
posi
tions
/tran
slat
ions
and
orie
ntat
ions
•B
ut w
e al
so n
eed
to s
et th
e keyf
ram
es—
mak
e th
e fo
ot c
ome
into
con
tact
with
the
ball,
for
exam
ple
Hie
rarc
hica
l Mod
els
•W
ant s
truc
ture
for
body
to b
e m
aint
aine
d
–P
ivot
join
t
–P
rism
atic
join
t
–B
all a
nd s
ocke
t joi
nt
Hum
an M
odel
s
Art
icul
ated
figu
res
are
a ho
rrib
le a
ppro
xim
atio
n
http
://ov
rt.n
ist.g
ov/p
roje
cts/
vrm
l/h-ani
m/jo
intIn
fo.h
tml
Hie
rarc
hica
l Mod
els
in S
imul
atio
n
Way
ne W
oote
n, Gat
ech
Kin
emat
ics
•T
he s
tudy
of m
otio
n w
ithou
t reg
ard
to th
e fo
rces
that
ca
use
it.
For
war
d:
B
ackw
ard:
)
,(
βα
fA
=)
(,
1A
f−
=β
αD
raw
gra
phic
sS
peci
fy fe
wer
deg
rees
of f
reed
omM
ore
intu
itive
con
trol
of d
ofM
aint
ain
cont
act w
ith th
e en
viro
nmen
tC
alcu
late
des
ired
join
t ang
les
for
cont
rol
αβ
A
For
war
d K
inem
atic
s
)si
n(si
n
)co
s(co
s
21
21
1
21
21
1
θθ
θθ
θθ
++
=+
+=
LL
y
LL
x
=
1000
1zyx
[][
][][
]1
12
2θ
θro
ttr
ansL
rot
tran
sL=
2L2θ 1θ
1L
Why
Inve
rse
Kin
emat
ics?
2L2θ 1θ
1L
•P
ick
up a
n ob
ject
or
plac
e fe
et o
n th
e gr
ound
Har
d to
do
with
forw
ard
kine
mat
ics
•A
llow
ani
mat
or to
set
few
er
para
met
ers
or a
t lea
st g
et a
goo
d fir
st
appr
oxim
atio
n
Use
r Con
trol
via
Inve
rse
Kin
emat
ics
•Jo
int s
pace
–P
ositi
on a
ll jo
ints—
fine
leve
l of c
ontr
ol
•C
arte
sian
spa
ce
–S
peci
fy e
nviro
nmen
tal i
nter
actio
ns e
asily
–M
ost d
egre
es o
f fre
edom
com
pute
d au
tom
atic
ally
Wha
t do
we
need
?
•N
atur
al lo
okin
g pa
th o
r ju
st a
goa
l pos
ition
? (t
ime
cohe
renc
y)
•Lo
cal o
r gl
obal
sol
utio
n?
Why
Inve
rse
Kin
emat
ics
—C
ontr
ol
•B
alan
ce--
-kee
p ce
nter
of m
ass
over
su
ppor
t pol
ygon
•C
ontr
ol--
-pos
ition
vaul
ter’s
han
ds o
n lin
e be
twee
n sh
ould
er a
nd v
ault
•C
ontr
ol--
-com
pute
kne
e an
gles
that
will
gi
ve th
e ru
nner
the
right
leg
leng
th
Wha
t do
we
need
?
2L2θ 1θ
1L
•S
kele
ton
with
1,2
,3 d
egre
e of
free
dom
join
ts
•S
olve
from
roo
t pos
ition
to e
nd e
ffect
or
posi
tion
•A
rbitr
ary
posi
tion
cons
trai
nts
•D
irect
ion/
orie
ntat
ion
cons
trai
nts
•Jo
int l
imits
•T
echn
ique
s fo
r ha
ndlin
g un
cons
trai
ned
syst
em
–A
ddin
g co
nstr
aint
s
–H
euris
tics
to p
ush
solu
tion
into
rig
ht p
art o
f sp
ace
–O
ptim
izat
ion
base
d on
som
e cr
iterio
n
Inve
rse
Kin
emat
ics
—C
lose
d F
orm
Sol
utio
n
2L
218
0θ−
1θ1L
++
−−
=
++
−−
=−
++
−−
++
=
+−
−+
+=
−
+=
+=
21
2 22 1
22
2
21
2 22 1
22
2
22
1
2 22
22 1
1
22
1
2 22
22 1
1
22
22
2
)(
cos
rule
co
sine
2
)(
)18
0co
s(
2
cos
rule
co
sine
2)
cos(
cos
)co
s(
LL
LL
yx
a
LL
LL
yx
yx
L
Ly
xL
a
yx
L
Ly
xL
yx
xa
yx
x
T
T
T
T
θ
θ
θθ
θθ
θ
θ
Tθ
Met
hods
•C
lose
d fo
rm—
only
for
fairl
y si
mpl
e m
echa
nism
s
•Ite
rativ
e so
lutio
ns
•S
olut
ions
•N
o so
lutio
n (o
utsi
de w
orks
pace
, too
few
do
f)
•M
ultip
le s
olut
ions
(re
dund
ancy
)
•S
ingl
e so
lutio
n
Wha
t mak
es IK
har
d?--
Red
unda
ncie
s
{}
X)
(
if )
...,
,(
by
de
fined
subs
pace
a
21
x
=∈
θθ
θθ
θθ
θf
xn
Add
con
stra
ints
to r
educ
e re
dund
anci
es
Cho
ose
solu
tion
that
is
•“c
lose
st”
to c
urre
nt c
onfig
urat
ion
•M
ove
oute
rmos
t lin
ks th
e m
ost
•E
nerg
y m
inim
izat
ion
•M
inim
um ti
me
•N
atur
al lo
okin
g m
otio
n???
Wha
t mak
es IK
har
d?--
sing
ular
ities
•Ill
-con
ditio
ned
near
sin
gula
ritie
s
•H
igh
stat
e sp
ace
velo
citie
s fo
r lo
w
cart
esia
n vel
ociti
es Goa
l
Rea
chab
le W
orks
pace
21
22
21
LL
yx
LL
+≤
+≤
−A
gain
, not
so
sim
ple
for
mor
e co
mpl
ex m
echa
nism
s
Itera
tive
Met
hod
Use
d in
Gira
rd a
nd Mac
ieje
wsk
i 198
5D
escr
ibed
in P
aren
t
Use
s in
vers
e of
Jaco
bian
to it
erat
ivel
y st
ep a
ll th
ejo
int a
ngle
s to
war
ds th
e go
al
The
Jac
obia
n
jiij
xfJ
Jij
xJ
xx
nxm
xx
f
∂∂=
∂=
∂∂
∂
=
is
ofel
emen
t th
the
whe
re)
(
)(
ch
ange
s
aldi
ffere
nti
to)
(
of ch
ange
s
aldi
ffere
nti
re
latin
gm
atrix
th
eis
Jaco
bian
dof)
of
(# m
dim
ensi
on
of
is
6) (g
ener
ally
n di
men
sion
of is
)
(
θθ
θθ
θθ
Jaco
bian
map
s ve
loci
ties
in s
tate
spa
ce to
vel
ociti
es
in c
arte
sian
spac
e. Ja
cobi
an is d
epen
dent
on
stat
e (a
nd m
ust b
e re
com
pute
d fr
eque
ntly
).
The
Jac
obia
n—
usin
g P
aren
t’s n
otat
ion
[]
∂∂∂∂∂∂
=
∂∂===
nznxx
ij
T
n
Tz
yx
zy
x
vv
J
VJ
vv
vV
θωθθθ
θθ
θθ
ωω
ω
ΜΟΚ
1
.
2
.
1
..
...,
,,
,,
,
IK a
nd th
e Ja
cobi
an
k
kk
xtJ
xJJ
x
xf
mnx
xf
θθ
θθθ
θθ
θ
abou
t
linea
rize
)(
dof)
of
(# di
men
sion
of is
6) (g
ener
ally
di
men
sion
of is
)
(
11
1
1
∂∆
+=
∂=
∂∂
=∂
=
=
−+
−
−
x
xx
∂+
goal
x
An
itera
tive
solu
tion
Inve
rtin
g th
e Ja
cobi
an
()
()
()
()1
1
.
.1
1
.
. w
here
in
vers
e-
pseu
do
com
pute
gene
ral
in
sq
uare
not
is
−−
+
+
−−
+
==
=
=
=
=
TT
TT
TT
TT
TT
JJJ
JJ
JJ
VJ
JJ
JJ
VJ
JJ
JJ
VJ
JV
J
nxm
J
θ
θ
θ
θ
Usi
ng th
e Ja
cobi
an
for I
K
•Sin
gula
ritie
s ca
use
the
rank
of t
he
Jaco
bian
to
chan
ge
•Jac
obia
n is v
alid
onl
y fo
r th
e co
nfig
urat
ion
for
whi
ch it
was
com
pute
d
•Pse
udo
inve
rse
min
imiz
es jo
int a
ngle
rat
es (
loca
lly)
•Cou
ld m
inim
ize
othe
r qu
antit
ies
Non
-Lin
ear O
ptim
izat
ion
Zha
o an
d B
adle
r, T
OG
199
4 (poi
nter
from
syl
labu
s)
Non
-line
ar p
rogr
amm
ing—
num
eric
al m
etho
d fo
r fin
ding
the
(loca
l) m
inim
um o
f a n
on-lin
ear
func
tion
Obj
ectiv
e fu
nctio
nC
onst
rain
tsN
on-li
near
opt
imiz
atio
n ro
utin
e
Obj
ectiv
e F
unct
ion
•Pos
ition
of e
nd e
ffect
or
•Or
posi
tion
and
orie
ntat
ion
•Or
aim
ing
at•F
or e
xam
ple,
you
mig
ht w
ant t
he h
and
to s
lide
alon
g th
e ta
blet
op o
r to
kee
p th
e gl
ass
uprig
ht.
)(2
)(
)(
)(
2
px
xP
xp
xP
x−
=∇
−=
)(2
),
(
)(2
),
(
vec
tors
lor
thon
orm
a
ofpa
ir a
as go
aln
orie
ntat
io
the
is ,
w
here
)(
)(
)(
22
,
eg
ee
y
eg
ee
x
gg
eg
eg
ee
yy
yx
P
xx
yx
P
yx
yy
xx
yx
P
ee
−=
∇
−=
∇
−+
−=
For
mul
atio
n
ii
ii
iT i
iT i
u
l
ba
ba
≤−
≤−
<=
θθθθθ
es
ineq
ualit
i
are
lim
itsjo
int
su
bjec
t to
)G
(
min
imiz
e
tech
niqu
enu
mer
ical
st
anda
rd a
use
)(
)(
an
d )
(e
Pe
GG
x
T
∇∂∂
=∇
θθ
θθ
Sol
utio
n
Sum
mar
y of
Kin
emat
ics
•F
orw
ard
is s
trai
ghtfo
rwar
d
•In
vers
e us
ually
req
uire
s a
num
eric
al s
olut
ion
•M
ay n
ot a
lway
s ge
t the
“rig
ht”
answ
er
–F
ram
e-to-
fram
e co
here
nce
(fix
long
seg
men
ts)
–N
atur
al lo
okin
g m
otio
n (h
ow d
efin
ed?)
–If
you
don’
t lik
e th
e so
lutio
n ad
d m
ore
cons
trai
nts
—ha
rdly
an
ele
gant
sol
utio
n…
The
stu
dy o
f mot
ion
with
out r
egar
d to
the
forc
es th
at c
ause
it--
-so
, wha
t abo
ut th
ose
forc
es?
Sim
ulat
ion-
--T
echn
ique
#3