ATRON Walker Learning

Hand-coding locomotion controllers for modular robots is difficult due to their polymorphic nature. Instead, we propose to use a simple and distributed reinforcement learning strategy. ATRON modules with identical controllers can be assembled in any configuration. To optimize the robot's locomotion speed its modules independently and in parallel adjust their behavior based on a single global reward signal.

In the video a nine modules robot first learns to move, this process takes on average 15 minutes. Then the robot disassembles and is manually assembled into different robots, which continue to learn. This demonstrates that the learning strategy is morphology independent.

Category:

Science & Technology

Tags:

• Learning
• Locomotion
• Reconfigurable
• Modular
• Robot

[RineMan215]

Protoss dragoon yo!

~RineMan215

[roboman2444]

But I wouldnt want to be in the room with them!

[DesmondInnovation]

See my YouTube channel for more on self reconfiguring technology.

[piepersongg]

I see... Intelligence is very hard to program. Good job.

[MeebleMeeble]

I would love to see 100 of these things in the same room!