Grasp-TAGS: Task-Aware Grasp Selection via Critic Distillation for Prehensile Loco-Manipulation
arXiv cs.AI · 7 Nov 2026, 04:44 AEST
Grasp-TAGS is a task-aware grasp selection framework designed for large-object rearrangement in mobile manipulators. It improves rearrangement success rates by selecting the grasp with the highest predicted downstream utility before making contact.
Key facts
- Loco-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.
- Existing approaches address this challenge either through model-based methods, which often require accurate dynamics models, or through learning-based methods, which reduce this requirement but typically assume a fixed or predefined initial grasp.
- We propose Grasp-TAGS, a task-aware grasp selection framework for goal-conditioned large-object rearrangement.
- Experiments demonstrate that Grasp-TAGS improves rearrangement success over both fixed single-grasp policies and random grasp selection, achieving an average success rate of 89.84% compared with 79.87% for the best-performing single-grasp policies, an improvement of 9.97 percentage points.
Read aloud
Grasp-TAGS is a task-aware grasp selection framework designed for large-object rearrangement in mobile manipulators. It improves rearrangement success rates by selecting the grasp with the highest predicted downstream utility before making contact. Loco-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. Existing approaches address this challenge either through model-based methods, which often require accurate dynamics models, or through learning-based methods, which reduce this requirement but typically assume a fixed or predefined initial grasp. We propose Grasp-TAGS, a task-aware grasp selection framework for goal-conditioned large-object rearrangement. Experiments demonstrate that Grasp-TAGS improves rearrangement success over both fixed single-grasp policies and random grasp selection, achieving an average success rate of eighty nine point eight four percent compared with seventy nine point eight seven percent for the best-performing single-grasp policies, an improvement of nine point nine seven percentage points.
Sources in this cluster
- Grasp-TAGS: Task-Aware Grasp Selection via Critic Distillation for Prehensile Loco-Manipulation
arXiv cs.AI · 7 Nov 2026, 04:44 AEST