02_leggedlocomotion
TRANSCRIPT
-
7/29/2019 02_leggedLocomotion
1/28
Zrich
ETH Master Course: 151-0854-00L
Autonomous Mobile Robots
Legged Locomotion
Roland Siegwart
Margarita Chli
Martin RufliDavide Scaramuzza
-
7/29/2019 02_leggedLocomotion
2/28
gg
Zrich
Lecture OverviewMobile Robot Control Scheme
2
raw data
positionglobal map
Sensing Acting
InformationExtraction
PathExecution
CognitionPath Planning
knowledge,data base
missioncommands
Real World
Environment
LocalizationMap Building
MotionControl
Perce
ption
actuator
commands
environment modellocal map
path
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
3/28
Zrich
Locomotion PrinciplesPrinciples Found in Nature
3
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
4/28
Zrich
Locomotion PrinciplesBiped Walking and Pure Rolling
Biped walking mechanism approximates purerolling via polygonal motion
The smaller the step size, the more
the polygon tends to a disk (wheel)
Work against gravity is required
Allows to overcome larger ob-
stacles (compared to rolling)
4
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
5/28
Zrich
Locomotion PrinciplesSelection of Locomotion Concept
Selection depends on terrain properties
robot weight and complexity
desired operating speed
maximal energy expenditure
required energy efficiency etc.
5
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
6/28
Zrich
Legged LocomotionNumber of Legs vs. Control Complexity
The number of legs influences Mechanical complexity
Control complexity
Insects can walk directly upon birth
Most mammals require several minutes to stand
Humans require more than a year to walk on two legs
6
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
7/28
Zrich
Legged LocomotionNumber of Distinct Gait Sequences
The gait is characterized as a distinct sequence oflift and release events of the individual legs
The number of possible events Nfor a walking machine with klegs is
For a biped walker, the number of possible events becomes
For a robot with 6 legs (hexapod)
7
!12
kN
6!3!12 kN
800'916'39!11 N
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
8/28
Zrich
CoG
Legged LocomotionStatic Locomotion
Static Locomotion
Characteristics
Body weight supported by atleast three legs
CoG withing support triangle
Safe, slow and inefficient
8
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
9/28
Zrich
Legged LocomotionStatic Locomotion
9
Most widespread static sequence with 6 legs
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
10/28Zrich
Legged LocomotionStatic Locomotion on ALoF
10
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
11/28Zrich
Legged LocomotionDynamic Locomotion
11
CoG
3
2 - Legged Locomotion
Dynamic Locomotion
Characteristics
The robot falls unless it moves
legs can be in ground contact
Fast, efficient, but demanding for
actuation and control
-
7/29/2019 02_leggedLocomotion
12/28Zrich
Legged LocomotionDynamic Locomotion with 4 Legs
12
Changeover walking
Galloping
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
13/28Zrich
Legged LocomotionDynamic Locomotion with 2 Legs
In principle only dynamic walking feasible Large feet allow for static walking, however
Two legs (biped) allow for four different states
1. Both legs down
2. Right leg down, left leg up
3. Right leg up, left leg down
4. Both leg up
13
1 2 1
1 3 1
1
4
1
2 3 2
2 4 2
3
4
3
turning
on right leg, orlimping
hopping
with two legs
hopping on
left leg
walking,
running
hopping onright leg
turningon left leg, orlimping
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
14/28Zrich
Legged Locomotion(Dynamic) Locomotion on ASIMO
14
Courtesy K. Moriyama
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
15/28Zrich
Legged LocomotionEnergy Optimization of Gaits
15
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
16/28Zrich
Legged LocomotionEnergy Opt. via Series Elastic Actuation
The optimal actuator should be back-drivable
be able to perform negative work
have a low inertia and gear ratio
be highly efficient
Series Elastic Actuators can emulate someof these properties
16
Series Elastic Actuator
x
Fground
u
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
17/28Zrich
Legged LocomotionDynamic Locomotion on StarlETH
17
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
18/28Zrich
Legged LocomotionFrom Legs to Links and Joints
A minimum of two DOF required to move leg a lift and a swing motion.
Sliding-free motion in more than one direction not possible
Three DOF for each leg required in most cases
Additional joints (and thus DOF) increase the complexity of the designand especially of the locomotion control
18
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
19/28Zrich
Forward KinematicsDefinition and Introduction
Forward KinematicsGiven is a set of joint angles
Determine resulting end-effector position
19
0y
0
x
),(gg yx
g
321
321321211
321321211
)sin()sin(sin
)cos()cos(cos
g
g
g
aaay
aaax
1a
2a
3a
2 - Legged Locomotion
-
7/29/2019 02_leggedLocomotion
20/28Zrich
0y
0x
Forward KinematicsChain of Coordinate Frames
20
2 - Legged Locomotion
21
-
7/29/2019 02_leggedLocomotion
21/28Zrich
Forward KinematicsHomogeneous Transformation Matrix
21
0y
0x
),( gg yx
1
10
0
11
g
g
g
g
y
x
Ty
x
: any two frames can be connected viaat most 1 rotation and 2 translation parameter
1y
1
x
2
2 - Legged Locomotion
a
22
-
7/29/2019 02_leggedLocomotion
22/28Zrich
Forward KinematicsHomogeneous Transformation Matrix
22
: any two frames can be connected viaat most 3 rotation and 3 translation parameter
i
g
g
g
i
i
i
i
i
i
i
ii
g
g
g
z
y
x
zy
x
z
y
x
110001
1
1
111
ii
R1
cossin0
sincos0001
1 ii
xR
cos0sin
010sin0cos
1 ii
yR
100
0cossin0sincos
1
ii
zR
3
2 - Legged Locomotion
23
-
7/29/2019 02_leggedLocomotion
23/28Zrich
Forward KinematicsDenavit-Hartenberg Convention
23
Denavit-Hartenberg (DH) reference frame layout Adds structure into kinematic chains
Involves 4 parameter only (instead of 6 for the general case)
Procedure
C
ourtesyE.
Tira-Th
ompson
2 - Legged Locomotion
24
-
7/29/2019 02_leggedLocomotion
24/28Zrich
Forward KinematicsDenavit-Hartenberg Convention
24
Resulting Denavit-Hartenberg transform. matrix
(joint angle): rotation about . Angle of w.r.t.
(twist angle): rotation about . Angle of w.r.t.
(link length): distance between axis and axis (i.e., and )(link offset ): offset along axis
1000
cossin0
sinsincoscoscossin
cossinsincossincos
11
111
111
1
iii
iiiiiii
iiiiiii
ii
d
a
a
T
i 1iz 1ix
1i ix
1ia 1i i
id 1iz
1iz iz
ix
iz 1iz
2 - Legged Locomotion
25
-
7/29/2019 02_leggedLocomotion
25/28Zrich
Forward KinematicsDH Coordinate Frames on a PUMA Arm
25
2 - Legged Locomotion
26
-
7/29/2019 02_leggedLocomotion
26/28Zrich
Inverse KinematicsDefinition and Introduction
26
Inverse KinematicsGiven is a desired end-effector position
Determine corresponding joint angles
Problem is non-trivial and generally not well-posed
No Solution One Solution >1 Solution
2 - Legged Locomotion
27
-
7/29/2019 02_leggedLocomotion
27/28Zrich
For 2-linkmanipulators employ e.g. cosine law(and solve for )
Inverse KinematicsSolution via Closed-form Approach
27
0x
0y
),( gg yx
221
2
2
2
1
22cos2 aaaayx gg ),( 21
2a
1a
2 - Legged Locomotion
28
-
7/29/2019 02_leggedLocomotion
28/28
Inverse KinematicsSolution via Iterative Search
28
From forward kinematics we know
his often not easily invertible in closed form
Approach: iteratively perform the following steps
1. Start from a known forward-kinematic solution (e.g., viasampling).
2. Linearize around , resulting in the Jacobian
....................................................................................................................................................................................................
..................................................................................................
........
3. Invert the Jacobian to obtain
4 Move by in direction
),,( 1 nT
ggg hzyx
h ),,( 1 n
TiiiT
zyxJ 121
),,( 1 nT
iii hzyx
T
igigig zzyyxx
i
n
mm
n
hh
hh
J
1
1
1
1