legged locomotion planning kang zhao b659 intelligent robotics spring 2013 1

30
Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

Upload: lawrence-wiggins

Post on 17-Jan-2016

220 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

1

Legged Locomotion Planning

Kang ZhaoB659 Intelligent RoboticsSpring 2013

Page 2: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

2

Planning Biped Navigation Strategies in Complex Environmentsโ€ข Joel Chestnutt, James

Kuffner, Koichi Nishiwaki, Satoshi Kagami

Page 3: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

3

O Global terrain map MO GoalO Primitive set {Trans}O Search algorithm

Page 4: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

4

Algorithm - Biped Robot ModelO State:

O ฮธ: position and orientation relative to {U}

O One-step motion destination

Page 5: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

5

Algorithm- State transitionsO Footstep transition

โ€ฆ

0

1

2 34

5 6

7

A 16-transitions set

Branching factor

Page 6: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

6

Algorithm- EnvironmentO Terrain map

Page 7: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

7

Algorithm- State Evaluation

๐‘‰= ๐‘“ (๐‘„ ,๐‘‡ ,๐‘„๐‘ ,๐‘„๐‘”)Location metric to

evaluate a locationโ€™s cost

๐ฟ๐‘–={ ๐ฟ๐‘– (๐‘„ )โˆž ๐‘–๐‘“ ๐ฟ๐‘– (๐‘„ )>๐ฟ๐‘–

๐‘™๐‘–๐‘š๐‘–๐‘ก, ๐‘–=1โ€ฆ5

๐ฟ (๐‘„ )=โˆ‘๐‘ค๐‘–๐ฟ๐‘–

Slope angle

Roughness

Stability

Largest bump

Safety

Page 8: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

8

Slope angle

Roughness

Stability

Largest bump

Safety

The slope angle of the surface at the candidate location. Perfectly horizontal surfaces are desired. The slope angle is computed by fitting a plane h(x, y) to the cells in the location.

1๐‘ โˆ‘

๐‘โˆˆ๐ถ

ยฟ๐‘ . h h๐‘’๐‘–๐‘” ๐‘กโˆ’h (๐‘ . ๐‘ฅ ,๐‘ . ๐‘ฆ )โˆจยฟยฟ

max {๐‘ . h h๐‘’๐‘–๐‘” ๐‘กโˆ’h (๐‘ . ๐‘ฅ ,๐‘ . ๐‘ฆ )โˆจ๐‘โˆˆ๐ถ }

Itโ€™s purpose is to take into account the possible inaccuracy of foot positioning. This can be computed using the roughness and largest bump metrics, using the cells around the foot location

Page 9: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

9

Algorithm- State Evaluation

Step metric to evaluate cost

of taking a step

๐‘† (๐‘„ ,๐‘‡ ,๐‘„๐‘ )=๐‘‡ .๐‘๐‘œ๐‘ ๐‘ก+๐‘คhโˆจ๐ป (๐‘„ ,๐‘„๐‘ )โˆจยฟ

Cost of transition

โ€ข Penalty for height changeโ€ข Collision check

๐‘„๐‘=๐‘‡ (๐‘„ )๐ป={ ๐ป (๐‘„ ,๐‘„๐‘ )

โˆž ๐‘–๐‘“ ๐ป (๐‘„ ,๐‘„๐‘ )>๐ปโ‘๐‘™๐‘–๐‘š๐‘–๐‘ก

๐‘‰= ๐‘“ (๐‘„ ,๐‘‡ ,๐‘„๐‘ ,๐‘„๐‘”)

Page 10: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

10

Algorithm- State Evaluation

Heuristic metric to evaluate

remaining cost

๐‘… (๐‘„ ,๐‘„๐‘”)=๐‘ค๐‘‘๐ท (๐‘„ ,๐‘„๐‘” )+๐‘ค๐œƒ|ฦŸ (๐‘„ ,๐‘„๐‘” )|+๐‘คhโˆจ๐ป (๐‘„ ,๐‘„๐‘” )โˆจยฟ

Euclidean distance Relative angle Height

difference

๐‘‰= ๐‘“ (๐‘„ ,๐‘‡ ,๐‘„๐‘ ,๐‘„๐‘”) The heuristic function estimates the cost to go from to a goal state

Its value is independent of the current search tree; it depends only on and the goal

Page 11: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

11

Best First SearchO It exploits state description to estimate how

โ€œgoodโ€ each search node isO An evaluation function maps each node of

the search tree to a real number

O Greedy BFS

๐‘… (๐‘„ ,๐‘„๐‘”)

h (๐‘ )=๐‘… (๐‘„ ,๐‘„๐‘”)

Page 12: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

12

A* Search

h (๐‘ )=๐‘… (๐‘„ ,๐‘„๐‘”)

๐‘… (๐‘„ ,๐‘„๐‘”)

๐ฟ (๐‘„๐‘ )+โˆ‘ ๐‘† (๐‘„๐‘– ,๐‘‡ ,๐‘„๐‘)

Page 13: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

Search tree

Searching the State SpaceA schematic view

Q s

Q g

Page 14: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

Search tree

Searching the State SpaceA schematic view

Q s

Q g

T 1

T 2

Page 15: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

Search tree

Searching the State SpaceA schematic view

Q s

Q g

Page 16: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

Search tree

Searching the State SpaceA schematic view

Q s

Q g

Page 17: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

Search tree

Searching the State SpaceA schematic view

Q s

Q g

Page 18: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

Search tree

Searching the State SpaceA schematic view

Q s

Q g

Page 19: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

19

ResultsO Cluttered terrain

Page 20: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

20

ResultsO Multi-level terrain

Page 21: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

21

ResultsO Uneven ground with obstacles

Page 22: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

22

Comparisons

O Distance to goalO Transitions and obstacle effectsO Metric weights

Page 23: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

23A 26-transitions set

A 40-transitions set BFS

Page 24: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

24

Performance comparison of A* and BFS for increasing numbers of stairs along the path

Page 25: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

25

Page 26: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

26

๐‘… (๐‘„ ,๐‘„๐‘”)=๐‘ค๐‘‘๐ท (๐‘„ ,๐‘„๐‘” )+๐‘ค๐œƒ|ฦŸ (๐‘„ ,๐‘„๐‘” )|+๐‘คhโˆจ๐ป (๐‘„ ,๐‘„๐‘” )โˆจยฟ

Page 27: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

27

Local-minimum problem

Page 28: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

28

Online Experiments

Stereo vision system

PlannerFootstep sequence

Trajectory generator

Walking area map

Page 29: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

29

Following work

O A tired planning Strategy for biped navigation, 2004O Biped navigation in rough environments using

on-board sensing, 2009

Page 30: Legged Locomotion Planning Kang Zhao B659 Intelligent Robotics Spring 2013 1

30

Multi-Step Motion Planning for Free-climbing Robotsโ€ข Tim Bretl, Sanjay Lall,

Jean-Claude Latombe, Stephen Rock