RoboPace: Contact-Aware Time-Optimal Retiming for Action-Chunk Policies

Key points

• Robot manipulation data collection has been shifting from teleoperation toward robot-free demonstrations, through interfaces such as the Universal Manipulation Interface (UMI) or directly from human hands.1
• Yet human timing does not directly transfer to robots: compliant hands tolerate fast contact, whereas robots may overshoot due to actuator and tracking limitations; conversely, robots can move faster in free space.1
• It adapts execution speed based on predicted contact, jointly accounting for contact-dependent speed limits and the robot's motion constraints.1
• Across three contact-rich tasks on a dual-arm robot, faster uniform execution and physical-limit-only retiming largely fail.1

Sources (1)

• [1]RoboPace: Contact-Aware Time-Optimal Retiming for Action-Chunk Policies
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 08:55 AM

  We present RoboPace, an online retiming layer that preserves the policy's geometric path while adapting its timing, respecting the target robot's kinematic and dynamic constraints.

  Robot manipulation data collection has been shifting from teleoperation toward robot-free demonstrations, through interfaces such as the Universal Manipulation Interface (UMI) or directly from human hands.

Extractive summary: sentences quoted from the sources.