Manifold Generative Annealing for Contact Constraint-aware Impedance Control with Reinforcement Learning Prior

Key points

• Contact-rich robotic tasks require coordinated motion and impedance under uncertain interaction dynamics and physical constraints.1
• Model-based generative methods show promising receding-horizon trajectories optimization results, but often lack structured behavioral guidance and refinement that jointly accounts for constraint geometry and actual contact responses.1
• MGA first organizes pretrained reinforcement learning policies into structured sequence proposals, retaining task-specific interaction strategies without committing to their execution.1
• Crucially, realization-aware manifold refinement aligns updates from the refinement bank with both task constraints and locally achievable physical responses, allowing learned experience to guide online adaptation under changing contact conditions.1

Sources (1)

• [1]Manifold Generative Annealing for Contact Constraint-aware Impedance Control with Reinforcement Learning Prior
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 15, 02:51 AM

  To this end, we propose Manifold Generative Annealing (MGA), a joint model-free and model-based generative framework for motion--impedance optimization.

  Contact-rich robotic tasks require coordinated motion and impedance under uncertain interaction dynamics and physical constraints.

Extractive summary: sentences quoted from the sources.