Fingertips motion planning and force computation for dextrous - - PowerPoint PPT Presentation

▶
fingertips motion planning
SMART_READER_LITE
LIVE PREVIEW

Fingertips motion planning and force computation for dextrous - - PowerPoint PPT Presentation

Oct 23, 2022 123 likes •287 views

IROS 2010 IROS 2010 Workshop On Workshop On Grasp Planning and Task Grasp Planning and Task Learning by Imitation Learning by Imitation Fingertips motion planning and force computation for dextrous manipulation N.Daoud, J.P.Gazeau,

slide-1
SLIDE 1

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Fingertips motion planning and force computation for dextrous manipulation N.Daoud, J.P.Gazeau, S.Zeghloul, M.Arsicault

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

slide-2
SLIDE 2

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Outline

  • Context of research
  • Introduction & Prior Art
  • Motion strategy
  • Simulation of manipulation tasks
  • Grasp stability
  • Experimental results
  • Conclusion
slide-3
SLIDE 3

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

2006 - 1 dof underactuated hand 2002 - 16 dof exosqueleton 1991 - 1 dof underactuated gripper 1996 – A mechanical hand with 16 dof fully actuated

2009 ABILIS project

slide-4
SLIDE 4

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

A general strategy for

  • bject manipulation

including :

  • Planning method for object

manipulation;

  • Efficient algorithm to compute

fingertip forces

  • Grasp synthesis

The experimental site :

  • Hand embedded controller
  • Hand position & force control
  • Kuka KR16 6 axis robot

Mechanical Hand Electronic Interface Multi Axis Controller POWER SUPPLY PC with SMAR Software

slide-5
SLIDE 5

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

A geometrical reasoning

The manipulation task is defined by the object trajectory Pd(t)6 such as :

       ) ( ) ( ) ( t t r t P

d

 r

3 : position of the center of mass of the grasped object

 

3 : pitch, yaw and roll angles of the object, and t the time. The object trajectory is divided into a succession of small displacements Ti-1,i : Where T0,f is a homogeneous transformation that represents

  • f the object from its initial to its final configuration.

T0,f=T

0,1.T1,2..T n-1,n.Tn,f

slide-6
SLIDE 6

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

n

E P Q Xob Yob Zob G

Rob

Xd Yd Zd Rd

Assumptions

  • Objects are rigid
  • Manipulation is only carried out

with fingertips

  • Fingertip is hemispherical
  • Shape of object, dimensions

and mass are known

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

For each small displacement Ti-1,i , 2 contact modes between object and finger :

  • Fixed point mode
  • Rolling without sliding mode

Sliding mode is not considered

slide-7
SLIDE 7

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

01 Initial grasp choice 02 (Definition of contact points Pi object- finger) 03 Decomposition of the trajectory of the 04 object in N small displacements dP with 05 dP= d1 ,d2, d3, dXG, dYG, dZG 06 FOR i=1 TO N 07 (For each small object displacement N° i) 08 Update rotation of the object: 09 i= i + di with i=i+1 10 Update translation of the object: 11 XG = XG + dXG ; 12 YG = YG + dYG ; 13 ZG = ZG + dZG 14 Compute small joint displacements dqi 15 IF Solution is OK THEN 16 Update joint parameters q=q+dq 17 ELSE 18 Solution out of range: 19 Repositioning of the fingers on the object 20 Computation of the new contact point P 21 END

Find initial grasp with 3 fingers Object motion (fingers rolling without sliding on object surface) Check :

  • Collisions
  • Joint limit

Manipulation achieved with success Find a new optimal 3 fingers grasp with genetic algorithms Use finger gaiting to reach the new grasp (the fourth finger will be used to keep the object position and orientation unchanged) No Yes

Strategy for manipulation

slide-8
SLIDE 8

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation 01 Initial grasp choice 02 (Definition of contact points Pi object- finger) 03 Decomposition of the trajectory of the 04 object in N small displacements dP with 05 dP= d1 ,d2, d3, dXG, dYG, dZG 06 FOR i=1 TO N 07 (For each small object displacement N° i) 08 Update rotation of the object: 09 i= i + di with i=i+1 10 Update translation of the object: 11 XG = XG + dXG ; 12 YG = YG + dYG ; 13 ZG = ZG + dZG 14 Compute small joint displacements dqi 15 IF Solution is OK THEN 16 Update joint parameters q=q+dq 17 ELSE 18 Solution out of range: 19 Repositioning of the fingers on the object 20 Computation of the new contact point P 21 END

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

Fingers roll without sliding :

Strategy for manipulation

  • bject/Rd)

(P V = finger/Rd) (P V     We write the small displacement model :

  • bject

finger

dP dP 

slide-9
SLIDE 9

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

 

  • bjet

2 1 2 1 v

dP = dq ) q , q , (q J n R + dq ) q , q , q ( J    

  • bjet

doigt

dP dP 

n

E P Q

Xd Yd Zd

Rd

n’

P’ Q’ E’

' n R. ' P E'   

The new contact point P’ : with R radius of hemispherical fingertip. A linear system with 3 equations and 3 unknown factors dqj(j=1..3)

dPobject = dx + PG^d

qj = qj + dqj (j=1..3)

Evolution of the point of contact for a small object displacement

slide-10
SLIDE 10

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

Manipulation tasks... ... for a cylinder ... for a prism

slide-11
SLIDE 11

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

Problem :

Apply to the object the global external force Fe, necessary to ensure its stability.

  • W. Fg = Fe

with Fg = (Fg1T, Fg2T, Fg3T)T The 6x9 grasping matrix W is defined as follows :

      

3 2 1

R R R I I I W

             

i i i i i i i

x y x z y z R

where The contact force must be applied to the object without sliding or breaking contact. We write the static friction constraints :

 

. 1 .

2 2

  

i i i i

n F F F 

 

. 

i i n

F

3 , 2 , 1  i

slide-12
SLIDE 12

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

We formulate the fingertip force Fg that can be obtained from Fe with :

Fg () = WT [W.WT]-1. Fe + N. 

  • =

3 2 3 2 3 2 2 3 2 3 2 3 3 1 3 1 3 1 1 3 1 3 1 3 2 1 2 1 2 1 1 2 1 2 1 2

z z y y x x z z y y x x z z y y x x z z y y x x z z y y x x z z y y x x N T

with Find the vector that minimizes the following quadratic function

 T

3 2 1

    

F F F

T

2 1 ) (  

with the static friction constraints.

An optimization problem

slide-13
SLIDE 13

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Problem resolution :

Use of sequential unconstrained minimization techniques (SUMT). Advantages : Initialization of the process doesn’t require any feasible solution. Time cost 1.4ms / C++ Time comparison in Matlab (0.34s vs 1.4s for fmincon)

Finger Fx Fy Fz |F| [N] 1 1.1582

  • 0.864

0.5895 1.5606 2

  • 0.234

0.8617 0.5981 1.0748 3

  • 0.923

0.0025 0.7743 1.2052

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

slide-14
SLIDE 14

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

A learning based method for force evaluation

0.00 0.40 0.80 1.20 1.60 Time (s)

  • 20.00

0.00 20.00 40.00 60.00 80.00 Contact Measure

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

Force Control

NEURAL NETWORK FINGER N°i

1 2 3 Normal force evaluation N Contact on distal phalanx Contact on intermediate phalanx Contact on proximal phalanx

slide-15
SLIDE 15

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

Position Control …Manipulation … Reach and Grasp

slide-16
SLIDE 16

IROS 2010 Workshop On Grasp Planning and Task Learning by Imitation

Context

  • f research

Introduction & Prior Art Motion strategy Simulation of manipulation tasks Grasp stability Experimental results Conclusion

  • a general scheme for planning the fingers motion in the aim
  • f object manipulation;
  • object stability solved for online manipulation;

Today Tomorrow

  • Find the initial grasp ;
  • Solve regrasping problem inside the hand during a

manipulation task.