DeViGrasp: Robust Visual Mobile Grasping for Quadruped Manipulators under Degraded Perception

This paper addresses the vulnerability of whole-body control policies for quadruped manipulators under unreliable onboard visual perception. The authors systematically examine the effects of occlusion, segmentation-mask dropout, depth noise, and target-localization jitter on grasp reasoning and target tracking. The proposed DeViGrasp method aims to develop robust control strategies for mobile grasping in complex environments with degraded visual inputs.