A framework for coordinating 2–3 robotic arms on shared manipulation tasks, using reinforcement learning for task allocation and ROS2 + MoveIt2 + Open3D for the full execution stack.
Status: Work in progress.
- Multi-arm speedup — does adding arms reduce batch completion time?
- RL vs. baseline allocators — does PPO beat simpler scheduling heuristics?
- Motion planning latency — does trajectory caching reduce planning time?
Results will be reported here once measured.