Multi-agent framework enables parts to self-assemble tangram puzzles
What happened
Paul Knaußt and co-authors propose an object-centric reinforcement learning framework that treats each part as an agent and learns a shared policy to accomplish continuous assembly control. The framework uses a graph neural network to represent parts, target poses, and their spatial relationships, predicts the translational and angular velocities of each component, and employs per-agent PPO with a centralized graph critic to learn motion and temporal coordination within a single feedback policy. The study validates this in planar tangram assembly: parts must slide and rotate on the assembly surface to a given configuration, with target configurations and initial layouts generated procedurally. The authors report demonstrating parallel assembly in simulation and transferring the policy to a real robot, where execution was changed to sequential.
AI 根据报道生成 · 2 days ago更新
Reporting timeline
Follow the reporting to explore different perspectives.
- arXiv cs.ROIt Takes Two to Tangram: Multi-Agent Object-Centric Reinforcement Learning for Assembly
The authors propose an object-centric reinforcement learning framework that treats each part as an agent and learns a shared policy for continuous assembly control. A graph neural network represents the parts, their target poses, and their spatial relationships, and predicts translational and angular velocities for each component; per-agent PPO with a centralized graph critic learns motion and temporal coordination within a single feedback policy. The study validates the approach on planar Tangram assembly, using procedurally generated target configurations and randomized initial layouts to support training and evaluation, and demonstrates parallel assembly in simulation and transfer to sequential execution on a real robot.
Activity for this event
- Current
- 3
- Peak
- 10Nov 9 14:00
- Past 24 hours
- -50%
The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.