DuoMind: Enabling Distributed Multi-Robot Coordination with Semantic Communication
DuoMind splits multi-robot control between local action models and VLM orchestrators that exchange semantic messages.
Each robot reasons over the task, its own observations, and messages from peers before issuing low-level instructions and new messages. The paper also introduces RoboPoly, a benchmark for long-horizon manipulation tasks under distributed control. Experiments on RoboPoly and RoboTwin report better multi-robot task performance, with ablations crediting both hierarchical orchestration and semantic communication. ArXiv · AI/CL/LG's note
Each robot reasons over the task, its own observations, and messages from peers before issuing low-level instructions and new messages. The paper also introduces RoboPoly, a benchmark for long-horizon manipulation tasks under distributed control. Experiments on RoboPoly and RoboTwin report better multi-robot task performance, with ablations crediting both hierarchical orchestration and semantic communication. ArXiv · AI/CL/LG's note
score 4