Grasp-TAGS selects grasps via critic distillation for loco-manipulation
Original title: Grasp-TAGS: Task-Aware Grasp Selection via Critic Distillation for Prehensile Loco-Manipulation
Grasp-TAGS ranks grasps by predicted downstream utility before contact in robot rearrangement tasks.
Mobile manipulators must coordinate locomotion and manipulation for long-horizon tasks. Grasp-TAGS trains a PPO policy across multiple initial grasps and distills a critic into a lightweight network. The network ranks candidate grasps using only grasp configuration and goal. It achieves 89.84% success on rearrangement, outperforming fixed-grasp baselines.
- Unified PPO policy across grasps
- Distills critic into grasp utility network
- Selects highest utility grasp pre-contact
- 89.84% average success rate
- 9.97 percentage point gain over single-grasp policies
From the original article
1 min read at arxiv.orgLoco-manipulation, which requires mobile manipulators to coordinate locomotion and manipulation over long horizons, is a key capability for large-object rearrangement in applications such as service robotics and warehouse automation. Despite recent progress, reliably executing these tasks remains challenging because performance depends on complex robot-object and object-environment interaction dynamics.
By Xun Tu · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.