Toward Autonomous Free-Climbing Robots

1
Toward AutonomousFree-Climbing Robots

• Tim Bretl
• Jean-Claude Latombe
• Stephen Rock

Special thanks to Eric Baumgartner, Brett
Kennedy, and Hrand Aghazarian at the Planetary
Robotics Lab, NASA-JPL
2
Goal

• Develop integrated control, planning, and
  sensing capabilities to enable a wide class of
  multi-limbed robots to climb steep natural
  terrain.

3
Generic vs. Specific Robot
Specific

• Generic

LEMUR IIb, Planetary Robotics Lab, NASA-JPL
Sitti and Fearing, UC Berkeley
4
Previous Multi-Limbed Climbing Robots
Each exploits a specific surface property
Neubauer, 1994
NINJA II Hirose et al, 1991
Yim, PARC, 2002
5
Free Rock Climbing is about Natural Friction
6
and Non-Gaited Motion
Non-Gaited
Gaited
7
Overall, rock climbing is about how to apply
strength, not about strength itself it is a
problem-solving activity
8
Example System
9
Equilibrium Constraint
Feasible positions of robots center of mass
10
Configuration Space

• For each combination of knee bends
• Position (xP,yP) of pelvis
• Joint angles (q1,q2) of free limb

11
Feasible Space
12
Feasible Space

1. Simple test for the feasibility of (xp,yp)

where
13
Feasible Space

1. Simple test for the feasibility of (xp,yp)
2. Feasible (?1,?2) varying with (xp,yp), in one
   half of ?f

Qf
where
14
Feasible Space

1. Simple test for the feasibility of (xp,yp)
2. Feasible (?1,?2), varying with (xp,yp), in one
   half of ?f
3. Switching between halves of ?f

15
Motion Planning

• Basic Approach (Probabilistic Roadmap)
• Sample 4D configuration space
• Check equilibrium condition
• Check (self-)collision
• Check torque limit
• Refined approach
• Sample 2D pelvis space, lift to full 4D paths
• Narrow passages are found in the 4D space

16

1. Achieve q20
2. Move with q20
3. Switch between halves of Qf
4. Move with q20
5. Move to goal

17
backstep
highstep
lieback
18
JPLs LEMUR robot
19
(No Transcript)
20
Current Work

• Terrain sensing and hold detection
• Force control and slippage sensing
• Uncertainty (hold location, limb positioning)
• Motion optimization
• Extension of feasible space analysis

21
Whats Next?
Xtreme ironing
?
gtgtgt
X
22
Xtreme ironing is one of the fastest-growingsport
s in the world
23
(No Transcript)