analiza aiqbench
This research advances mobile manipulation by enabling legged robots to perform language-guided grasping under partial observation, a critical capability for safe field inspections. By combining open-vocabulary detection with point-cloud completion and execution-aware grasp selection, the system significantly improves success rates in cluttered environments compared to view-dependent baselines. The work demonstrates how integrating advanced perception algorithms directly into robotic control pipelines can reduce operator exposure in hazardous settings.
Language-Guided Grasping under Partial Observation for Mobile Manipulation in Field Inspection and Maintenance
arXiv:2603.07866v4 Announce Type: replace Abstract: Offshore inspection and maintenance have increasingly been using legged robots for routine sensing, yet many useful interventions still require physical interaction with tools, containers, and task-relevant objects. Employing robots for these tasks can reduce operators' exposure in confined, elevated, or potentially explosive areas. This paper presents a language-guided grasping pipeline for a legged mobile manipulator operating under partial observation. An operator defines the target, the system grounds it in RGB with open-vocabulary detection and promptable segmentation, extracts an object-centric RGB-D point cloud, improves sparse geometry through depth compensation and point-cloud completion, and selects a 6-DoF grasp using collision, clearance, reachability, and approach constraints. The system is implemented on a quadruped robot with an arm and evaluated in two cluttered tabletop scenes motivated by small-object retrieval during inspection and maintenance. Across paired trials, the proposed pipeline achieved 9/10 successful grasps, compared with 3/10 for a view-dependent deployment baseline. In this controlled setting, object-centric completion and execution-aware selection reduced approach collisions and improved the reliability of language-guided grasping for supervised field manipulation.
Abstract:Offshore inspection and maintenance have increasingly been using legged robots for routine sensing, yet many useful interventions still require physical interaction with tools, containers, and task-relevant objects. Employing robots for these tasks can reduce operators' exposure in confined, elevated, or potentially explosive areas. This paper presents a language-guided grasping pipeline for a legged mobile manipulator operating under partial observation. An operator defines the target, the system grounds it in RGB with open-vocabulary detection and promptable segmentation, extracts an object-centric RGB-D point cloud, improves sparse geometry through depth compensation and point-cloud completion, and selects a 6-DoF grasp using collision, clearance, reachability, and approach constraints. The system is implemented on a quadruped robot with an arm and evaluated in two cluttered tabletop scenes motivated by small-object retrieval during inspection and maintenance. Across paired trials, the proposed pipeline achieved 9/10 successful grasps, compared with 3/10 for a view-dependent deployment baseline. In this controlled setting, object-centric completion and execution-aware selection reduced approach collisions and improved the reliability of language-guided grasping for supervised field manipulation.
Robotics (cs.RO); Machine Learning (cs.LG); Systems and Control (eess.SY)
arXiv:2603.07866 [cs.RO]
(or arXiv:2603.07866v4 [cs.RO] for this version)
From: Ricardo Vilela De Godoy Dr. [view email] [v1] Mon, 9 Mar 2026 00:42:32 UTC (8,738 KB) [v2] Mon, 4 May 2026 19:58:29 UTC (8,738 KB) [v3] Sun, 5 Jul 2026 23:46:19 UTC (8,738 KB) [v4] Tue, 15 Sep 2026 22:14:41 UTC (8,738 KB)
Text integral sub licența deschisă a publicației · 231 cuvinte
Subiecte conexe
[AINews] not much happened today
Latent Space · 10 sept. 2026
The Machines that Make the Machines
NVIDIA · 7 oct. 2026
MobiGuide: Vocabulary-Guided Online Recovery for Mobile Manipulation
arXiv · 15 oct. 2026
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · 5 oct. 2026