Planning Biped Locomotion using Motion Capture Data and Probabilistic Roadmaps

1
Planning Biped Locomotion using Motion Capture
Data and Probabilistic Roadmaps

• MIN GYU CHOI, JEHEE LEE, SUNG YONG SHIN
• Korea Advanced Institute of Science and Technology

ACM Transactions on Graphics 2002
2
Outline

• Introduction
• Roadmap Construction
• Roadmap Search
• Motion Generation
• Experimental Results

3
Introduction

• Motivation
• Interactive games and simulations
• Off-line production animation
• Goal
• Produces a natural-looking motion for a
  human-like biped figure to move from a given
  start position to a goal position using a set of
  prescribed motions

4
Outline

• Introduction
• Roadmap Construction
• Node Generation
• Local Planner
• Cost Function
• Roadmap Search
• Motion Generation
• Experimental Results

5
Node Generation

• Node
• Configuration of stance foot (p, q)
• Edge
• Motion clip
• Randomly generate valid configurations
• Newly generated node using local planner add to
  roadmap

6
Local Planner
7
Cost Function

• Distance cd
• Adaptation cr
• Users preference cp
• Cost function

8
Roadmap Construction
9
Outline

• Introduction
• Roadmap Construction
• Roadmap Search
• Augmented Roadmap
• Path search
• Footprint Refining
• Motion Generation
• Experimental Results

10
Augmented Roadmap
11
Path Search

• Minimum-cost path algorithm Dijkstra
• Complexity
• O(kVlogkVE)
• k is node numbers of Posture transition graph

12
Footprint Refining

• Path straightening

13
Footprint Refining (cont.)

• Difference smoothing

14
Roadmap Search
15
Outline

• Introduction
• Roadmap Construction
• Roadmap Search
• Motion Generation
• Experimental Results

16
Motion Generation

• Motion retargetting problem
• Each footprints as a variational constraint over
  a time interval
• Initial body trajectory is very important for the
  convergence of numerical optimization and the
  quality of the result

17
Body Trajectory Estimation
18
Motion Generation
19
Outline

• Introduction
• Roadmap Construction
• Roadmap Search
• Motion Generation
• Experimental Results

20
(No Transcript)
21
(No Transcript)