yoonsang lee sungeun kim jehee lee seoul national university

51
Yoonsang Lee Sungeun Kim Jehee Lee Seoul National University Data-Driven Biped Control

Upload: meriel

Post on 24-Feb-2016

62 views

Category:

Documents


0 download

DESCRIPTION

Data-Driven Biped Control. Yoonsang Lee Sungeun Kim Jehee Lee Seoul National University. Biped Control. Human. Biped character. ?. Biped Control is Difficult. Balance, Robustness, Looking natural Various stylistic gaits. ASIMO Honda. HUBO KAIST. PETMAN Boston Dynamics. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Yoonsang LeeSungeun Kim

Jehee Lee

Seoul National University

Data-Driven Biped Control

Page 2: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Biped Control

?Human Biped character

Page 3: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Biped Control is Difficult

• Balance, Robustness, Looking natural• Various stylistic gaits

PETMANBoston Dynamics

ASIMOHonda

HUBOKAIST

Page 4: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Issues in Biped Control

Naturalness

Robustness

Richness

Interactivity

human-like natural result

maintaining balance

variety of motor skills

interactive control via user interface

Page 5: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Goal

As realistic as motion capture data

Robust under various conditions

Equipped with a variety of motor skills

Controlled interactively

Naturalness

Robustness

Richness

Interactivity

Page 6: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Related Work• Manually designed controller

– [Hodgins et al. 1995] [Yin et al. 2007]

• Non-linear optimization– [Sok 2007] [da Silva 2008] [Yin 2008] [Muico 2009] [Wang 2009] [Lasa 2010] [Wang 2010] [Wu

2010]

• Advanced control methodologies– [da Silva 2008] [Muico 2009] [Ye 2010] [Coros 2010] [Mordatch 2010]

• Data-driven approach– [Sok 2007] [da Silva 2008] [Muico 2009] [Tsai 2010] [Ye 2010] [Liu 2010]

Page 7: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Our Approach

• Control methods have been main focus– Machine learning, optimization, LQR/NQR

• We focus on reference data– Tracking control while modulating reference data

Page 8: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Our Approach

• Modulation of reference data

• Balancing behavior of human

• Importance of ground contact timings

Page 9: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Advantages

• Do not require– Non-linear optimization solver– Derivatives of equations of motion – Optimal control– Precomputation

Easy to implement & Computationally efficient

Page 10: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Advantages

• Reference trajectory generated on-the-fly can be used

Any existing data-driven techniques can be used to actuate physically simulated bipeds

Page 11: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University
Page 12: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Overview

forward dynamics simulation

animation engine

user interaction

data-driven control tracking control

Page 13: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Overview

forward dynamics simulation

user interaction

data-driven control tracking control

animation engine

Page 14: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

• High-level control through user interfaces• Generate a stream of movement patterns

Animation Engine

motion fragments

query

motion DBpattern generator

user interaction

stream of movement patterns

Page 15: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

• Collection of half-cycle motion fragments

• Maintain fragments in a directed graph

Motion Database

motion capture data motion fragments

Page 16: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Overview

forward dynamics simulation

user interaction

tracking control

animation engine

data-driven control

Page 17: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Data-Driven Control

• Continuous modulation of reference motion

• Spatial deviation– SIMBICON-style feedback balance control

• Temporal deviation– Synchronization reference to simulation

Page 18: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Balancing

...reference motion

simulation

frame n frame n+1 frame n+2

...

...

Page 19: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

frame n frame n+1 frame n+2

Balancing

target pose

...reference motion

simulation

...

...

Page 20: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

frame n frame n+1 frame n+2

Balancing

tracking

...reference motion

simulation

...

...

Page 21: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

frame n+1 frame n+2frame n

Balancing

tracking

...reference motion

simulation

...

...

Page 22: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Balance Feedback

• Near-passive knees in human walking

• Three-step feedback– stance hip– swing hip & stance ankle– swing foot height

Page 23: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Balance Feedback

• Biped is leaning backward

?

reference motionat current frame

reference motionat next frame

simulation

Page 24: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

• Stance Hip

Balance Feedback

target poseat next frame

reference frame simulation

Page 25: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

• Swing Hip & Stance Ankle

Balance Feedback

target poseat next frame

reference frame simulation

Page 26: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Balance Feedback

• Swing Foot Height

target poseat next frame

reference frame simulation

Page 27: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Feedback Equations

Stance hip

Swing hip

Stance ankle

Swing foot height

reference frametarget pose

Page 28: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Feedback Equations

desired states current states

Stance hip

Swing hip

Stance ankle

Swing foot height

Page 29: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Feedback Equations

parameterstransition function

Stance hip

Swing hip

Stance ankle

Swing foot height

Page 30: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Synchronization

referencemotion

swing foot contacts the ground

Page 31: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Synchronization

current time

referencemotion

simulation

Page 32: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Early Landing

referencemotion

contact occurs!simulation

Page 33: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Early Landing

referencemotion

simulation

dequed

Page 34: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Early Landing

referencemotion

simulation

Page 35: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Early Landing

referencemotion

simulation

warped

Page 36: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Early Landing

referencemotion

simulation

Page 37: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Delayed Landing

referencemotion

not contact yet!simulation

Page 38: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Delayed Landing

referencemotion

simulation

expand byintegration

Page 39: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Delayed Landing

referencemotion

simulation contact occurs!

expand byintegration

Page 40: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Delayed Landing

referencemotion

simulation

warped

Page 41: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Delayed Landing

referencemotion

simulation

Page 42: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Overview

forward dynamics simulation

user interaction

animation engine

data-driven control tracking control

Page 43: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

• Compute torques that attempts to follow reference trajectory (ex. PD control)

• We use floating-base hybrid inverse dynamics

Tracking Control

inverse dynamicsdesiredjoint accelerations joint torques

external forces

Page 44: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University
Page 45: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University
Page 46: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University
Page 47: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University
Page 48: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Why does this simple approach work?

• Human locomotion is inherently robust

• Mimicking human behavior– Distinctive gait serves as a reference trajectory– We do modulate the reference trajectory

Page 49: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Discussion

• We do not need optimization, optimal control, machine learning, or any precomputation

• Physically feasible reference motion data

• Future work– Wider spectrum of human motions

Page 50: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Acknowledgements

• Thank– All the members of SNU Movement Research

Laboratory– Anonymous reviewers

• Support– MKE & MCST of Korea

Page 51: Yoonsang  Lee Sungeun  Kim Jehee  Lee Seoul National University

Data-Driven Biped ControlYoonsang Lee, Sungeun Kim, Jehee Lee