Robotics
Researchers build scalable decentralized system for coordinating robot swarms
A new framework enables teams of autonomous robots to collaborate across large outdoor spaces without centralized control or synchronized timing.
Researchers have developed a decentralized coordination system that allows robot teams to work together at scale in outdoor environments without relying on a central hub or synchronized timing between all agents.
How the system works#
The framework, called Perception-Action-Communication (PAC), gives each robot the ability to sense its immediate surroundings, act independently, and share information with nearby teammates. Rather than forcing all robots to operate on the same clock, the system's core modules—perception, communication, message processing, and action—can run at different speeds. Information flows between these modules through independent buffers, enabling asynchronous operation.
The key to scaling across many robots is a graph neural network that propagates information hop-by-hop through the fleet's communication network. This approach allows global coordination to emerge from robots that can only sense and communicate locally, since data is relayed across the swarm.
Implementation and testing#
The researchers built PAC using ROS2, a widely used robotics middleware, structured as modular, extensible nodes designed to serve as foundational software for deployable swarm systems.
The system was validated in two ways:
- Real-world tests with up to 20 outdoor quadrotor robots
- Simulation studies based on real-world data with teams of up to 100 robots
What makes it field-ready#
The experiments demonstrated three critical properties for practical deployment:
- Scalability: the system maintains performance as team size grows
- Resiliency: it continues functioning if individual robots or communication links fail
- Repeatability: results remain consistent across multiple runs
This combination of decentralized design, asynchronous operation, and proven real-world performance positions PAC as practical infrastructure for large-scale autonomous robot teams operating in challenging outdoor environments.
