Manipulation planning for documented objectshomepages.laas.fr/~jmirabel/defense/jmirabel-defense...PhD defense Joseph Mirabel Institut National Polytechnique Toulouse February 21,

Manipulation planning for documented objectsPhD defense

Joseph Mirabel

Institut National Polytechnique Toulouse

February 21, 2017

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

Robots in factories

J. Mirabel 2/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

European project: Factory in a Day

Best research teams in Europe in various fields.

J. Mirabel 3/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

European project: Factory in a Day

Best research teams in Europe in various fields.

Goal: make robots affordable to small industries.

J. Mirabel 3/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

European project: Factory in a Day

Best research teams in Europe in various fields.

Goal: make robots affordable to small industries.

13 researchers for 2 days.

J. Mirabel 3/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

European project: Factory in a Day

Best research teams in Europe in various fields.

Goal: make robots affordable to small industries.

13 researchers for 2 days.

J. Mirabel 4/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

Manipulation planning for documented objects

J. Mirabel 5/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

Example scenario

⇒

J. Mirabel 6/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Reduction property

Solution to a manipulation problem

A sequence of:

transit paths: the robot moves alone,

transfer paths: the robot manipulates an

object.

J. Mirabel 7/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Reduction property

Solution to a manipulation problem

A sequence of:

transit paths: the robot moves alone,

transfer paths: the robot manipulates an

object.

Paths in Placement ∩Grasp can be

approximated by transit and tranfer paths.

J. Mirabel 7/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Reduction property

Solution to a manipulation problem

A sequence of:

transit paths: the robot moves alone,

transfer paths: the robot manipulates an

object.

Paths in Placement ∩Grasp can be

approximated by transit and tranfer paths.

J. Mirabel 7/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

J. Mirabel 8/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Find sequence of tasks to achieve the

goal:

Grasp red box,

Move red box,

Grasp green box,

Move green box,

Put down red box,

. . .

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Find sequence of tasks to achieve the

goal:

Grasp red box,

Move red box,

Grasp green box,

Move green box,

Put down red box,

. . .

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Find sequence of tasks to achieve the

goal:

Grasp red box,

Move red box,

Grasp green box,

Move green box,

Put down red box,

. . .

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Find sequence of tasks to achieve the

goal:

Grasp red box,

Move red box,

Grasp green box,

Move green box,

Put down red box,

. . .

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Find sequence of tasks to achieve the

goal:

Grasp red box,

Move red box,

Grasp green box,

Move green box,

Put down red box,

. . .

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Find sequence of tasks to achieve the

goal:

Grasp red box,

Move red box,

Grasp green box,

Move green box,

Put down red box,

. . .

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Find sequence of tasks to achieve the

goal:

Grasp red box,

Move red box,

Grasp green box,

Move green box,

Put down red box,

. . .

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Task “Grasp red box”

Compute grasping pose

Motion planning query

Solution found.

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Task “Grasp red box”

Compute grasping pose

Motion planning query

Solution found.

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Task “Grasp red box”

Compute grasping pose

Motion planning query

Solution found.

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Task “Grasp red box”

Compute grasping pose

Motion planning queryNo solution found.

◮ Return failure ?◮ State infeasibility ?◮ Use a different grasping pose ?

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Task “Grasp red box”

Compute grasping pose

Motion planning queryNo solution found.

◮ Return failure ?◮ State infeasibility ?◮ Use a different grasping pose ?

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Task “Grasp red box”

Compute grasping pose

Motion planning queryNo solution found.

◮ Return failure ?◮ State infeasibility ?◮ Use a different grasping pose ?

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art: Combined task and motion planning

Task planner

Motion planner

Start, goal configurationsPath, if found

Needs an interface between the

symbolic and geometric layer:

HPN,

conditional reachability graph,

logical predicates,

. . .

J. Mirabel 9/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

Manipulation Planning

J. Mirabel 10/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

Manipulation Planning

J. Mirabel 10/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art

Subclasses of problems:

rearrangement planning,

J. Mirabel 11/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art

Subclasses of problems:

rearrangement planning,

navigation among movable obstacles,

J. Mirabel 11/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art

Subclasses of problems:

rearrangement planning,

navigation among movable obstacles,

Dual-arm manipulation,

J. Mirabel 11/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art

Subclasses of problems:

rearrangement planning,

navigation among movable obstacles,

Dual-arm manipulation,

Regrasping,

J. Mirabel 11/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

State of the art

Subclasses of problems:

rearrangement planning,

navigation among movable obstacles,

Dual-arm manipulation,

Regrasping,

. . .

J. Mirabel 11/47 Manipulation planning for documented objects

What problems ?

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

Main contributions

Constraint graph: a model of admissible motions,

Manipulation-RRT: an algorithm addressing manipulation problems,

Algorithms to validate the continuity of constrained motions,

Humanoid Path Planner: an open-source motion planning library.

J. Mirabel 13/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

Main contributions

Constraint graph: a model of admissible motions,

Manipulation-RRT: an algorithm addressing manipulation problems,

Algorithms to validate the continuity of constrained motions,

Humanoid Path Planner: an open-source motion planning library.

J. Mirabel 13/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

Main contributions

Constraint graph: a model of admissible motions,

Manipulation-RRT: an algorithm addressing manipulation problems,

Algorithms to validate the continuity of constrained motions,

Humanoid Path Planner: an open-source motion planning library.

J. Mirabel 13/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Introduction

Main contributions

Constraint graph: a model of admissible motions,

Manipulation-RRT: an algorithm addressing manipulation problems,

Algorithms to validate the continuity of constrained motions,

Humanoid Path Planner: an open-source motion planning library.

J. Mirabel 13/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Table of Contents

1 Constraint graph

Problem modelling

Constraint graph2 Manipulation planning

Foliation

Manipulation planner

Crossed foliation issue3 Continuity in constrained motion planning

Newton Raphson method

The problem

Continuous path projection4 Conclusion

Contributions

PerspectivesJ. Mirabel 14/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Composite robot

Composite robot

Kinematic chain composed of all robots and

objects kinematic chains.

J. Mirabel 15/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Composite robot

Composite robot

Kinematic chain composed of all robots and

objects kinematic chains.

Configuration space

CS = CSPR2 × CSBox × CSDrawer

Robot configuration: q = (qPR2, qbox , qdrawer ) ∈ CS

J. Mirabel 15/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

J. Mirabel 16/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

J. Mirabel 16/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

“Object in the hand”

J. Mirabel 16/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

“Object in the hand”

Explicit formulation

qbox ← HandPosition(qPR2)

J. Mirabel 16/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

“Object in the hand”

Explicit formulation

qbox ← HandPosition(qPR2)

Implicit formulation

HandPosition(qPR2)−ObjectPosition(qbox) = 0

J. Mirabel 16/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

“Object in the hand”

Explicit formulation

qbox ← HandPosition(qPR2)

Implicit formulation

HandPosition(qPR2)−ObjectPosition(qbox) = 0

f (q) = 0

J. Mirabel 16/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

“Drawer in the hand”

J. Mirabel 17/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

“Drawer in the hand”

Explicit formulation

qdrawer ← HandPosition(qPR2) ?

J. Mirabel 17/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

“Drawer in the hand”

Explicit formulation

qdrawer ← HandPosition(qPR2)

J. Mirabel 17/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

“Drawer in the hand”

Explicit formulation

qdrawer ← HandPosition(qPR2)

Implicit formulation

HandPosition(qPR2)− DrawerPosition(qdrawer ) = 0

J. Mirabel 17/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

Geometrical problem

depends only

q = (qPR2, qbox , qdrawer )

“Drawer in the hand”

Explicit formulation

qdrawer ← HandPosition(qPR2)

Implicit formulation

HandPosition(qPR2)− DrawerPosition(qdrawer ) = 0

f (q) = 0

J. Mirabel 17/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

“Object in the hand”

“Drawer in the hand”

J. Mirabel 18/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

“Object in the hand”

“Drawer in the hand”

“Object in right hand” and “drawer in left hand”

J. Mirabel 18/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

“Object in the hand”

“Drawer in the hand”

“Object in right hand” and “drawer in left hand”

Equilibrium ?

J. Mirabel 18/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

“Object in the hand”

“Drawer in the hand”

“Object in right hand” and “drawer in left hand”

Equilibrium ?Numerical constraint

f (q) = 0

J. Mirabel 18/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

“Object in the hand”

“Drawer in the hand”

“Object in right hand” and “drawer in left hand”

Equilibrium ?Numerical constraint

f (q) = 0

Constraint solver

Given q such that f (q) 6= 0

J. Mirabel 18/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Constraints

“Object in the hand”

“Drawer in the hand”

“Object in right hand” and “drawer in left hand”

Equilibrium ?Numerical constraint

f (q) = 0

Constraint solver

Given q such that f (q) 6= 0

→ Find q∗ ∈ CS such that f (q∗) = 0.

J. Mirabel 18/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Equilibrium constraint

Motion planning for humanoid robot

J. Mirabel 19/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Equilibrium constraint

Motion planning for humanoid robot

J. Mirabel 19/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Object documentation

J. Mirabel 20/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Object documentation

Gripper and handle frames

X-axis: grasp axis approach,

Z-axis: possible allowed rotation.

J. Mirabel 20/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Object documentation

Contact and support surfaces

Z-axis: orthogonal to the surface.

J. Mirabel 20/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Object documentation

Numerical constraints: Gripper + Handle

Validation of grasp.

Parametrization of the grasp space.

J. Mirabel 20/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Problem modelling

Object documentation

Numerical constraints: Contact + Support surface

Validation of placement.

Parametrization of the placement space.

J. Mirabel 20/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Manipulation problem

Definition

Robots

Objects

Environment

Initial and goal configurations for robots and objects

J. Mirabel 21/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Manipulation problem

Definition

Robots with end-effectors

Objects

Environment

Initial and goal configurations for robots and objects

J. Mirabel 21/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Manipulation problem

Definition

Robots with end-effectors

Objects with handles and contact surfaces

Environment

Initial and goal configurations for robots and objects

J. Mirabel 21/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Manipulation problem

Definition

Robots with end-effectors

Objects with handles and contact surfaces

Environment with support surfaces

Initial and goal configurations for robots and objects

J. Mirabel 21/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Manipulation problem

Definition

Robots with end-effectors

Objects with handles and contact surfaces

Environment with support surfaces

Constraint graph

Initial and goal configurations for robots and objects

J. Mirabel 21/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

State

J. Mirabel 22/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

State

represents a set of configuration satisfying some constraints,

uses validation constraints.

J. Mirabel 22/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

State

represents a set of configuration satisfying some constraints,

uses validation constraints.

Transition

J. Mirabel 22/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

State

represents a set of configuration satisfying some constraints,

uses validation constraints.

Transition

represents a set of motions from one state to another,

uses parametrization constraints.

J. Mirabel 22/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

placement graspP

P

G

G

placement: Object is on the table.

J. Mirabel 23/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

placement graspP

P

G

G

placement: Object is on the table.

grasp: Robot holds the object.

J. Mirabel 23/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

placement graspP

P

G

G

placement: Object is on the table.

grasp: Robot holds the object.

J. Mirabel 23/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

placement graspP

P

G

G

placement: Object is on the table.

grasp: Robot holds the object.

J. Mirabel 23/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

placement graspP

P

G

G

placement: Object is on the table.

grasp: Robot holds the object.

P: Placement parameter (yp, zp, θp) is constant.

J. Mirabel 23/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph

placement graspP

P

G

G

placement: Object is on the table.

grasp: Robot holds the object.

P: Placement parameter (yp, zp, θp) is constant.

G: Grasp parameter θg is constant.

J. Mirabel 23/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

Available objects

Object A

Object B

Available grippers

Gripper 1

Gripper 2

Constraints

Placement of A,B

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1Available objects

Object B

Available grippers

Gripper 2

Constraints

Placement of B, Grasp A-1

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1Available objects

Object B

Available grippers

Gripper 2

Constraints

Placement of B, Grasp A-1

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1

A-1, B-2

Available objects

Available grippers

Constraints

Grasp A-1, B-2

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1

A-1, B-2

Available objects

Available grippers

Constraints

Grasp A-1, B-2

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1

A-1, B-2

Available objects

Object B

Available grippers

Gripper 2

Constraints

Placement of B, Grasp A-1

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1

A-1, B-2

Available objects

Object A

Object B

Available grippers

Gripper 1

Gripper 2

Constraints

Placement of A,B

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1

A-1, B-2

B-2

Available objects

Object A

Available grippers

Gripper 1

Constraints

Placement of A, Grasp B-2

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1

A-1, B-2

B-2

Available objects

Object A

Available grippers

Gripper 1

Constraints

Placement of A, Grasp B-2

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1

A-1, B-2

B-2

Available objects

Available grippers

Constraints

Grasp A-1, B-2

J. Mirabel 24/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Constraint graph generation

Placement

A-1

A-1, B-2

B-2

A-2

A-2, B-1

B-1

J. Mirabel 25/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Example

J. Mirabel 26/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Constraint graph

Example

J. Mirabel 26/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Table of Contents

1 Constraint graph

Problem modelling

Constraint graph2 Manipulation planning

Foliation

Manipulation planner

Crossed foliation issue3 Continuity in constrained motion planning

Newton Raphson method

The problem

Continuous path projection4 Conclusion

Contributions

PerspectivesJ. Mirabel 27/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Foliation

Constraint graph and configuration space

2 constraints on motion

f : position of the object.

g: grasp of the object.

J. Mirabel 28/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Foliation

Constraint graph and configuration space

2 constraints on motion

f : position of the object.

g: grasp of the object.

placement grasp

J. Mirabel 28/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Foliation

Constraint graph and configuration space

2 constraints on motion

f : position of the object.

g: grasp of the object.

placementf grasp

f

f

g

J. Mirabel 28/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

Rapidly exploring Random Treeusing the constraint graph

placement grasp

Manipulation RRT

qrand = shoot random config()

for each connected components

qnear = nearest neighbor(qrand , cc)

fe, fp = select next state(qnear )

qproj = project(qrand , fe)

qnew = extend(qnear , qproj , fp)

tree.insert node( (qnear , qnew , fp) )

J. Mirabel 29/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

J. Mirabel 30/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Manipulation planner

J. Mirabel 30/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Crossed foliation issue

What about this case ?

⇒

J. Mirabel 31/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Crossed foliation issue

Crossed foliation issue

placement grasp

J. Mirabel 32/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Crossed foliation issue

Crossed foliation issue

placement grasp

The probability that q1proj and q2

proj are on the same

Bgiis zero.

J. Mirabel 32/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Crossed foliation issue

Crossed foliation issue

placement grasp

The probability that q1proj and q2

proj are on the same

Bgiis zero.

Crossed foliation transition

Keep track of the reached leaves.

J. Mirabel 32/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Crossed foliation issue

Results

Joseph Mirabel and Florent Lamiraux, Manipulation planning: addressing the crossed foliation issue, ICRA 2017.

J. Mirabel 33/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Crossed foliation issue

Results

Joseph Mirabel and Florent Lamiraux, Manipulation planning: addressing the crossed foliation issue, ICRA 2017.

J. Mirabel 33/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Table of Contents

1 Constraint graph

Problem modelling

Constraint graph2 Manipulation planning

Foliation

Manipulation planner

Crossed foliation issue3 Continuity in constrained motion planning

Newton Raphson method

The problem

Continuous path projection4 Conclusion

Contributions

PerspectivesJ. Mirabel 34/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Discontinuities

J. Mirabel 35/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Discontinuities

J. Mirabel 35/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Solving abstract constraints

Find x such that f (x) = 0:

f(x)

x

J. Mirabel 36/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Solving abstract constraints

Find x such that f (x) = 0:

f(x)

x

x1

J. Mirabel 36/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Solving abstract constraints

Find x such that f (x) = 0:

f(x)

x

x1

f(x1)

J. Mirabel 36/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Solving abstract constraints

Find x such that f (x) = 0:

f(x)

x

x1

f(x1)

x2

J. Mirabel 36/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Solving abstract constraints

Find x such that f (x) = 0:

f(x)

xx2

f(x2)

J. Mirabel 36/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Solving abstract constraints

Find x such that f (x) = 0:

f(x)

xx2

f(x2)

x3

J. Mirabel 36/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Solving abstract constraints

Find x such that f (x) = 0:

f(x)

xx3

J. Mirabel 36/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Newton Raphson method

Solving abstract constraints

Find x such that f (x) = 0:

f(x)

x

x1

x2

x3

J. Mirabel 36/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

The problem

Point-wise path projection

− Constraint:

f (x , y) = x2 − 1 = 0

J. Mirabel 37/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

The problem

Point-wise path projection

− Constraint:

f (x , y) = x2 − 1 = 0

− Discretize

J. Mirabel 37/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

The problem

Point-wise path projection

− Constraint:

f (x , y) = x2 − 1 = 0

− Discretize

− Project each sample

J. Mirabel 37/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

The problem

Point-wise path projection

− Constraint:

f (x , y) = x2 − 1 = 0

− Discretize

− Project each sample

J. Mirabel 37/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Newton-Raphson algorithm

Intuition

Discontinuity arises when the constraint Jacobian becomes singular.

J. Mirabel 38/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Newton-Raphson algorithm

Intuition

Discontinuity arises when the constraint Jacobian becomes singular.

Theorem

Let q ∈ CS. If the constraint is not singular in q, then the iteration function is continuous

on a ball of radiusσ(q)

M

where

σ(q) is the smallest singular value of the constraint Jacobian,

M is an upper bound of the norm of the constraint Hessian.

J. Mirabel 38/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

Initial path to project

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

Continuity ball

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

Add sample within continuity ball

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

Project sample

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

After one iteration

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

Continuity ball

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

Add sample within continuity ball

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

Project sample

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

After two iterations

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Progressive path projection algorithm

J. Mirabel 39/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Example

J. Mirabel 40/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Continuous path projection

Example

J. Mirabel 40/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Table of Contents

1 Constraint graph

Problem modelling

Constraint graph2 Manipulation planning

Foliation

Manipulation planner

Crossed foliation issue3 Continuity in constrained motion planning

Newton Raphson method

The problem

Continuous path projection4 Conclusion

Contributions

PerspectivesJ. Mirabel 41/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Result

J. Mirabel 42/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Result

J. Mirabel 42/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Main contributions

Constraint graph: a model of admissible motions,

Manipulation-RRT: an algorithm adressing manipulation problems,

Algorithms to validate the continuity of constrained motions,

Humanoid Path Planner: an open-source motion planning library.

J. Mirabel 43/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Main contributions

Constraint graph: a model of admissible motions,

Manipulation-RRT: an algorithm adressing manipulation problems,

Algorithms to validate the continuity of constrained motions,

Humanoid Path Planner: an open-source motion planning library.

J. Mirabel 43/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Main contributions

Constraint graph: a model of admissible motions,

Manipulation-RRT: an algorithm adressing manipulation problems,

Algorithms to validate the continuity of constrained motions,

Humanoid Path Planner: an open-source motion planning library.

J. Mirabel 43/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Main contributions

Constraint graph: a model of admissible motions,

Manipulation-RRT: an algorithm adressing manipulation problems,

Algorithms to validate the continuity of constrained motions,

Humanoid Path Planner: an open-source motion planning library.

J. Mirabel 43/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Humanoid Path Planner

General purpose motion and manipulation planning library.

J. Mirabel 44/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Humanoid Path Planner

General purpose motion and manipulation planning library.

Main contributors: Florent Lamiraux, Joseph Mirabel.

J. Mirabel 44/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Humanoid Path Planner

General purpose motion and manipulation planning library.

Main contributors: Florent Lamiraux, Joseph Mirabel.

Published in IROS 2016

HPP: a new software for constrained motion planning, Joseph Mirabel et al..

J. Mirabel 44/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Contributions

Humanoid Path Planner

General purpose motion and manipulation planning library.

Main contributors: Florent Lamiraux, Joseph Mirabel.

Published in IROS 2016

HPP: a new software for constrained motion planning, Joseph Mirabel et al..

Used in many other works:◮ Interactive Motion Planning with Contact, Blin et al, IROS 2016.◮ Ballistic motion planning, Campana et al., MIG 2016 and IROS 2016.◮ A gradient-based path optimization method for motion planning, Campana et al., Advanced Robotics 2016.◮ A Versatile and Efficient Pattern Generator for Generalized Legged Locomotion, Carpentier et al., ICRA 2016.◮ Motion Generation for Pulling a Fire Hose by a Humanoid Robot, Ramirez-Alpizar et al., Humanoids 2016.◮ Tuning Interaction in Motion Planning with Contact, Blin et al, RoMan 2016.◮ Manipulation planning: addressing the crossed foliation issue, Mirabel et al., ICRA 2017.

◮ A fast and efficient acyclic contact planner for multiped robots, Tonneau et al., Submitted to IJRR 2016.

◮ Exploiting Structure in Humanoid Motion Planning, Andreas Orthey, PhD Thesis, 2015

J. Mirabel 44/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Perspectives

Robot programing

Using a graphical interface, the

user:

provides models with

documentation,

J. Mirabel 45/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Perspectives

Robot programing

Using a graphical interface, the

user:

provides models with

documentation,

specifies the task in a

supervised process,

J. Mirabel 45/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Perspectives

Robot programing

Using a graphical interface, the

user:

provides models with

documentation,

specifies the task in a

supervised process,

click on “solve” button.

J. Mirabel 45/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Perspectives

Robot programing

Offline

Build a 3D model Set-up planner

Online

Localize objects

(re) Plan path

Execute plan

J. Mirabel 46/47 Manipulation planning for documented objects

Introduction Constraint graph Manipulation planning Continuity in constrained motion planning Conclusion

Perspectives

Thank you !

J. Mirabel 47/47 Manipulation planning for documented objects