Morphological computation and decentralized learning in a swarm of sterically interacting robots

Area of Science:

• Robotics
• Collective Behavior
• Morphological Computation

Background:

• Robotic swarms typically avoid physical interactions, limiting operational density.
• Naturally occurring swarms thrive in crowded conditions, unlike current robotic systems.

Purpose of the Study:

• To introduce a mechanical design rule enabling robotic swarms to function in collision-dominated environments.
• To develop Morphobots, a platform for embodied computation through morpho-functional design.

Main Methods:

• Engineered a 3D-printed exoskeleton to encode a reorientation response to external forces (gravity, collision).
• Tested the generic force orientation response on various robotic platforms (e.g., Kilobots) and custom robots.
• Investigated individual robot behavior and swarm-level collective phototaxis in dilute to crowded conditions.

Main Results:

• The exoskeleton enhances individual robot motility and stability, enabling distinct responses to forces and collisions.
• Force orientation response facilitates collective phototaxis and information flow in crowded swarms.
• Morphological computation's effect increases with swarm size, demonstrated in physical (64 robots) and simulated (8192 agents) experiments.

Conclusions:

• A mechanical design rule allows robotic swarms to leverage collisions for enhanced performance and collective intelligence.
• Morphobots demonstrate a novel approach to embodied computation, enabling dense swarm operation.
• This work opens new avenues for swarm robotics in complex, crowded environments.