Visual collective behaviors on spherical robots.

Area of Science:

• Robotics
• Artificial Intelligence
• Collective Behavior

Background:

• Traditional collective motion models often assume unrealistic omniscient perception.
• Individual robots possess limited sensing capabilities, which are frequently overlooked.

Purpose of the Study:

• To implement a visual flocking model using a robot-in-the-loop approach.
• To reproduce collective motion behaviors in a flock of 10 spherical robots using minimal visual information.
• To bridge the gap between simulation and physical experiments in robotic collective motion.

Main Methods:

• Utilized panoramic visual information (retinal position, optical size, optic flow) from individual robots.
• Employed a virtual anchor to confine robot movements and prevent wall interactions.
• Validated the model in a robot-in-the-loop system, comparing simulation and physical experiments.

Main Results:

• Successfully reproduced collective motion phases, including swarming and milling, with independent robots.
• Demonstrated nearly identical behaviors in both simulated and physical robot-in-the-loop environments.
• Showcased the efficacy of a minimal visual model for recreating complex collective behaviors.

Conclusions:

• A simple visual robot-in-the-loop approach can effectively reproduce collective robotic motion.
• The developed model bridges the gap between simulation and physical robotic experiments.
• Minimal visual cues are sufficient for achieving complex collective behaviors in robotic systems.