industrial robots challenging environmentsmicrosoft powerpoint - italy.ppt author: ok created date:...

Post on 05-Oct-2020

3 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

1

Robots in Environments Co-habited by Human Beings

Oussama KhatibRobotics Laboratory

Department of Computer ScienceStanford University

r o b o t sa 50 year journey

ICRA 2000, San Francisco

Industrial Robots Challenging Environments

2

Service & Assistance Haptic Interaction

Robotically Aided Surgery

The Latest

Honda: P2, P3, & Asimo

The Latest

Sony: SDR-4X

3

The Latest

AIST: HRP2

4

At Stanford

Stanford Robotic Platforms (1993)

Human Guided MotionAssistance

• Greater Freedom• Branching Structures• Difficult Coordination• Compliance: Multi-Contact• Complex Motion Planning

Exploring Human-Like Structures

5

Joint Space Control Joint Space Control

Joint Space Control

F

( )GoalV xF = −∇

( )GoalV x

T FJΓ =

Task-Oriented Control

Γ

F

dynamics( )F F=

x

xΛ F=pµ+ +

Task-Oriented Dynamics Task Dynamics – Posture Space

Λ

Fx

x pµ+ + F=

TJT

J

Task Space: TJ

Posture Space: TN

= Γ

Joint-Space Dynamics

Task-relatedPosture

TN

6

Task Dynamics – Posture Space

Fx

Λ x pµ+ + F=

TJT

J

Task Space: TJ

Posture Space: TN

= Γ

Joint-Space Dynamics

Task-relatedPosture

TN

Task Dynamics – Human-Like Structure

x J⇒

1

2

.

m

x

xx

x

=

T TJ N⇒ ⇒

Task & Posture Decomposition

Whole-body Control Task/Posture Control

Task T skT

aJ=Γ F

PostureTask ΓΓΓ +=

0=x⇒

Posture Desired Po uT

st reN −=Γ Γ

Task/Posture Consistency Task and Posture Control

no joint trajectories⇒

Whole-body Control

Task Field

Posture Field

Dynamically Decoupled

7

Posture Energy

( )Posture Desired Po t rT

s u eN V −= −∇Γ( )T

Task TaskJ V= −∇Γ

PostureTask ΓΓΓ +=

Posture Energy

Total center-of-mass

Human Natural Motion

Motion Capture Specific Robot MotionsMotion CharacteristicsNatural Energies

Posture EnergyHuman Natural Motion

• Load Distribution• Balance - CoM• Joint Efforts• Joint Limits• Joint Stiffness• Kinetic Energy• Environment Awareness• Kinematic Symmetry• Etc. Motion Capture

HumanMotionCapture

HumanDynamicModel

Learning from Human Motion

Natural Energies Animation of motion capture data

Simulation 79 DOF and 136 MusclesBiometric Data & Bone Geometry (Scott Delp)

8

HumanMotionCapture

RobotDynamic

Model

HumanDynamicModel

Learning from Human Motion

Human Energies Robot Energies

9

10

General Motion

Task 1 = joint limits avoidance.

Task 2 = balancing.

Task 3 = left hand positioning.

Task 4 = right hand positioning.

Task 5 = wrist alignment.

Task 6 = body symmetry.

1

2

3

3

4

5

PriorityTasks

11

In motion, humansminimize fatigue & maximize social posture, under physical constraints

Compliant MotionMulti-point Contact

DecentralizedCooperation

Safety

Human-Friendly Robots

• Dependable & Safe• Soft Actuators• Light Structures• Impact-reduction Skin• Low Reflected Inertia• Distributed Sensing & Control• Advanced Capabilities

Towards Human-Friendly Robots

Safety

Performance

Competing?

Requirements

12

Why Are Robotic Arms Unsafe?

Equivalent Mass-Spring Model

Robot Collision

Head Injury Criteria (HIC)

Injury Prediction

• Uncontrolled collision biggest danger

• robot characteristics impact on injury

Risk Of Serious Injury!

>20 cm compliant covering

DMM Actuation

Parallel Actuation

Elastic Coupling

Large Base Actuator

Small JointActuator

Distributed Macro Mini Acutation

Safety Through Inertia Reduction

>20 cm compliant covering

Revisited: Risk Of Serious Injury!

DMM: 10x reduction in effective inertia

Human-Friendly Robot ConceptElastic Planning

Real-time collision-free path modification

ConnectingReactive Local AvoidancewithGlobal Motion Planning

13

Artificial Potential Field -Video

Free-Space Representation Free-Space Tunnel

14

Elastic Strip Task Consistent

Humanoid Elastic Planning - Video

Haptic Interaction

• Physical Interaction with Virtual Environments

• Touch and Manipulation• Interactive Dynamic

Simulation - inter-object interactions

15

Virtual PrototypingCAD Assembly/Serviceability

Animation

Many Applications...

Medicine

Education

Robotics

Interfaces for the Impaired

Sculpting

Many Devices...

Exos Exoskeleton

Immersion LaparoscopicImpulse Engine

Virtual Technologies CyperGrasp Glove

CMU MagLev Wrist

Iwata 6dof HapticDisplay

McGill PantographSensAblePhantom

Interactive Haptic Simulation

Contact Resolution

– Handle Multi-point contact

– Handle Multi-body Articulated Systems

– Needs to be interactive

Fast Dynamic AlgorithmsSimulation, Contact Resolution, and Control

• Efficient algorithms for contact dynamics O(n)

• Avoid constraints elimination• Cost-free effective mass at

arbitrary contacts

M

16

Contact Space

S

=

7

6

5

4

3

2

1

6

4

1

0100000

0001000

0000001

x

x

x

x

x

x

x

x

x

x

( )X⊕Augmented Contact Space–Full set of contact points

–Non Independent

( )XActive Contact Space– subset of contact points

– independent, but unknown

X XS ⊕=Active Contact Matrix

Contact Mass Properties

• Augmented Space

• Active Contact Space

1 1 TJ A J− −⊕ ⊕ ⊕Λ =

1 1 TS S− −⊕Λ = Λ

:J⊕

where

augmented space jacobian

Contact Mass Properties

• Augmented Space

• Active Contact Space

1 1 TJ A J− −⊕ ⊕ ⊕Λ =

1 1 TS S− −⊕Λ = Λ

:J⊕

where

augmented space jacobian

17

SAI Environment

Thank You!

top related