SARSwarm
A multi-agent reinforcement learning example project built on Metis-Core that simulates multiple drones conducting cooperative search and rescue, including communication modeling and a visualization interface.
Timeline (last 90 days)
SARSwarm is an example project bundled with Metis-Core that simulates 4 radio-coordinated drones searching for maritime disaster survivors.
The graphical interface of the SARSwarm example is developed based on the Raylib library.
Examples such as SARSwarm typically adopt the centralized training with decentralized execution (CTDE) paradigm: global information is allowed during training, while at execution time each agent relies only on local observations and limited communication.
In SARSwarm, a break in radio communication prevents drones from sharing maps with one another.
In SARSwarm, over time previously scanned areas become "colder" to indicate that a person in the water may drift back into that area due to ocean currents.
All Facts (5)
SARSwarm is an example project bundled with Metis-Core that simulates 4 radio-coordinated drones searching for maritime disaster survivors.
50%UnverifiedThe graphical interface of the SARSwarm example is developed based on the Raylib library.
50%UnverifiedExamples such as SARSwarm typically adopt the centralized training with decentralized execution (CTDE) paradigm: global information is allowed during training, while at execution time each agent relies only on local observations and limited communication.
50%UnverifiedIn SARSwarm, a break in radio communication prevents drones from sharing maps with one another.
50%UnverifiedIn SARSwarm, over time previously scanned areas become "colder" to indicate that a person in the water may drift back into that area due to ocean currents.
50%