Skip to content

Event

SPARK: A Sequential Planning Method Anchored to Robot Keypoints

1 reports · 1 sources

Bryce Grant and other authors propose a sequential planning method, SPARK: it describes tasks as a symbolic behavior tree, anchoring to the current 3D observation only when each action is invoked; it verifies the plan at runtime, re-anchors objects upon local failure, and requests a new plan only if this step fails. The authors state that the test tasks involve no practice and that simulator state is not accessed during runtime; a 4-bit quantized open-weight 27B planner is used. When objects are moved by 10 centimeters, with the help of scene-difference verification and hierarchical recovery, it still completes 66.7% of trials, and without needing to change the plan.

Report timeline

  1. SPARK: Sequential Planning via Anchored Robotic Keypoints

    The authors propose SPARK (Sequential Planning via Anchored Robotic Keypoints), which describes tasks as a symbolic behavior tree, anchoring to the current 3D observation only when each action is invoked; it verifies the plan at runtime and re-anchors objects on local failure, requesting a new plan only if this step fails.

    arXiv cs.RO ·

Related topics

Planning and decision makingGrasping and manipulationRobot models and learningResearch papers

Interest trend

Current
3
Peak
4Nov 4 01:00
Past 24 hours
–

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.