Dynamic Scoop-And-Flick Manipulation for Rapid Non-Prehensile High-Arc Object Transfer
Gijae Ahn, Junwoo Lee, Seung Hwa Oh, Mujin Shin, Seung-Joon Yi, Jungwon Seo
AI summary
Problem
Conventional pick-and-place relies on complex grasping strategies that struggle with thin, low-profile objects, creating a need for faster, non-prehensile manipulation methods.
Approach
The method uses a custom two-degree-of-freedom finger to scoop objects and launch them via elastic energy, guided by a neural network that predicts optimal flicking angles from object mass and target location.
Key results
- Successfully scooped and flicked low-profile objects into repeatable projectile trajectories
- Achieved up to 10/10 success rates in throwing objects over a 50 cm barrier into target buckets
- Demonstrated that adding arm horizontal velocity reliably extends range with under 5% error
- Validated a data-driven controller that predicts optimal finger parameters using only ~440 training samples
Why it matters
Provides a fast, planning-light alternative for high-volume automation tasks involving thin or hard-to-grasp items.
Abstract
This study presents dynamic scoop-and-flick ma- nipulation, a robotic technique that achieves desired projec- tile motions of target objects through rapid, non-prehensile physical interactions. The method allows a robot to scoop objects resting on a surface and quickly launch them into projectile trajectories. We formulate a theoretical model of the technique and realize it through a hybrid approach that combines model-based reasoning and data-driven learning. The advantages—namely, rapid and accurate pick-and-place with reduced planning complexity—are validated in experiments conducted with a particularly challenging class of objects: low- profile items with small thickness.