Experimental Robotics - Stanford Computer Science

19
4/13/2021 1 CS225A Experimental Robotics Lecture 5 Oussama Khatib Project Proposals Experimental Robotics 1 2

Transcript of Experimental Robotics - Stanford Computer Science

Page 1: Experimental Robotics - Stanford Computer Science

4/13/2021

1

CS225A

Experimental Robotics

Lecture 5 

Oussama Khatib

Project ProposalsExperimental Robotics

1

2

Page 2: Experimental Robotics - Stanford Computer Science

4/13/2021

2

Spring2021 Projects

Sports EnvironmentCookingGroceries

Human Robot InteractionMedicalService

Kinematics

Dynamics

Jacobians

Inverses

Task

Representations

Equations of Motion

Operational Space Control

Dynamic

Models

Compliance

Force Control

Control

Modalities

Redundant

Robots

Posture

Null Space

Dynamic Behavior

Whole-Body Control

Menu

3

4

Page 3: Experimental Robotics - Stanford Computer Science

4/13/2021

3

Project ProposalsExperimental Robotics

.. motion in contact

Operational Space Framework

5

6

Page 4: Experimental Robotics - Stanford Computer Science

4/13/2021

4

Joint Space Control

Joint Space Control

7

8

Page 5: Experimental Robotics - Stanford Computer Science

4/13/2021

5

Joint Space Control

F

( )GoalV xF

( )GoalV x

T FJ

Task‐Oriented Control

9

10

Page 6: Experimental Robotics - Stanford Computer Science

4/13/2021

6

F

dynamics( )F F

x

x Fp

Task‐Oriented Control

Unified Motion & Force Control

motion contactF F F contactF

motionF

11

12

Page 7: Experimental Robotics - Stanford Computer Science

4/13/2021

7

Effector Equations of Motion

Non‐Redundant Manipulator ; 0n m

01 2 T

mx x x x

1 2 T

nq q q q x G q

0R1nR

10n

Lagrange Equations in Operational Space

( ) Fd L L

dt x x

L VT

( ) GravityVF

d T T

dt x x x

Inertial forces Gravity vector

( )GravityV V q

Kinetic Energy

Potential Energy (Gravity)

Lagrangian

Since

13

14

Page 8: Experimental Robotics - Stanford Computer Science

4/13/2021

8

( ) ( , ) ( )x x x x p x F

Operational Space Dynamics

End‐Effector Centrifugal and Coriolis forces

( , ) :x x

( ) :p x End‐Effector Gravity forces

:F End‐Effector Generalized forces

( ) : x End‐Effector Kinetic Energy Matrix

:x End‐Effector Position andOrientation

Operational Space Control

( )TJ q F

F

1

2

T

Goal p Goal GoalV k x x x x

( )GoalV xF

15

16

Page 9: Experimental Robotics - Stanford Computer Science

4/13/2021

9

1

2

T

Goal p Goal GoalV k x x x x

System GravityT V

Fd T

dt x x

ˆGoal GravityF V V

X

Passive Systems

0GoalTd T

dt x x

V Stable

Conservative Forces

Asymptotic Stability

is asymptotically stable if

0 ; 0TsF x for x

0s v vF k x k

ˆp Goal vF k x x k x p Control

s

GoalTd T

dt x xF

V a system

sFx

p̂ Estimate of p

17

18

Page 10: Experimental Robotics - Stanford Computer Science

4/13/2021

10

Artificial Potential Field

Example: 2‐d.o.f arm

ˆ ( )p g vF k x x k x p x

( ) ( , ) ( )x x x x p x F

1q1l

2q2l

19

20

Page 11: Experimental Robotics - Stanford Computer Science

4/13/2021

11

Closed loop behavior

111*

1( ) v p gm q x k x k mx yx

122*

2( ) v p gm q y k y k my xy

* 2 *1 2 1 112 p g vm c m x m y k x x k x

* 2 *1 2 1 212 p g vm c m y m x k y y k y

Joint Space/Operational SpaceRelationships

( , ) ( , ) x qT x x T q q

1 1( ) ( )

2 2T Tx X x q A q q

1 1

2 2T TTq qJ J Aq q

Using  ( )x J q q

21

22

Page 12: Experimental Robotics - Stanford Computer Science

4/13/2021

12

where ( , ) ( )h q q J q q

1( ) ( ) ( ) ( )Tx J q A q J q

( , ) ( ) ( , ) ( ) ( , )Tx x J q b q q q h q q

( ) ( ) ( )Tp x J q g q

Joint Space/Operational SpaceRelationships

Example

2

2

1

1

d cx

d s

20

2

1 1

1 1

d s cJ

d c s

2 2q d

g

l1y

CI1

CI2

m1

m2

d2

x

23

24

Page 13: Experimental Robotics - Stanford Computer Science

4/13/2021

13

20

2

1 1

1 1

d s cJ

d c s

21 1 0 1;

1 0

dJ

11 21

2 22

0 1 0 0 1

1 0 0 1 0

m d

d m

0

2

0 11 1

01 1

c sJ

ds c

1 J

21

2 2

0

0

m

m m

2221 222 1 1

2 22

I I m l

md

2 2m m

1x1y

1m

2m

2m

25

26

Page 14: Experimental Robotics - Stanford Computer Science

4/13/2021

14

20

2

01 1 1 1

1 1 1 10

mc s c s

s c s cm

2 2 2m m m

20 2 2 2

22 2 2

1 1

1 1

m m s m sc

m sc m m c

2m2 2m m

20 2 2 2

22 2 2

1 1

1 1

m m s m sc

m sc m m c

27

28

Page 15: Experimental Robotics - Stanford Computer Science

4/13/2021

15

Nonlinear Dynamic Decoupling

with TJ F

( ) ( , ) ( )x x x x p x F Model

*( ) ( ,ˆ ˆ) ( )ˆF px x x xF Control Structure

*I x FDecoupled System

Dynamic Decoupling

( ) ( , ) ( )x x x x p x F

*ˆ ˆ ˆ( , ) ( )F F x x p x

0

1 * 1 1ˆ ˆˆmI X F P P

( )G x ( , )x x ( )P x

29

30

Page 16: Experimental Robotics - Stanford Computer Science

4/13/2021

16

0

*) ( , ) ( )(mI x x dx tGx F

1 ˆ( )G x I

1( , )x x P

( ) :d t unmodeled disturbances

Dynamic Decoupling: Closed Loop

Perfect Estimates

0

*mI x F

input of decoupled end‐effector*F

Goal Position Control

*v p gF k x k x x

Closed Loop

0m v p p gI x k x k x k x

31

32

Page 17: Experimental Robotics - Stanford Computer Science

4/13/2021

17

Closed Loop0m v p p gI x k x k x k x

t

x

max

2gx

1gx

PD Control

*v p gF k x k x x

Velocity‐Like Control

* pv g

v

kF k x x x

k

33

34

Page 18: Experimental Robotics - Stanford Computer Science

4/13/2021

18

dx

pd g

v

kx x x

k

dx

* pv g

v

kF k x x x

k

* v dF k x x

withmax

d

Vsat

x

1

0 1max

dx

1

( ) 1

x if xsat x

sign x if x

35

36

Page 19: Experimental Robotics - Stanford Computer Science

4/13/2021

19

Trajectory Tracking

Trajectory: , , d d dx x x

0

* ( ) ( ) m d v d p dF I x k x x k x x

( ) ( ) ( ) d v d p dx x k x x k x x

with

or

X dx x

0 X v X p Xk k

In joint space

0 q v q p qk k

with q dq q

37

38