面向机器人抓取的障碍物推理
Obstruction Reasoning for Robotic Grasping
把「障碍物推理」形式化为以目标物体为起点的障碍路径多步推理，并用障碍感知视觉线索锚定每一步推理，再以可验证奖励做监督+强化微调；做具身操作或 VLA 的研究者可复用该推理链构造与奖励设计，UNOBench 也可作为遮挡场景抓取的评测基准。
原文依据 We devise a novel multi-step reasoning process based on obstruction paths originated by the target object.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论在杂乱环境中成功完成机器人抓取，不仅要求模型在视觉上定位目标物体，还要求其推理出必须事先清除的障碍物。当前的视觉语言具身推理模型虽已展现出涌现的空间理解能力，但在障碍物推理与可达性规划方面仍存在局限。为弥补这一差距，我们提出 UNOGrasp，一种基于学习的视觉语言模型，能够执行视觉接地的障碍物推理，以推断出清除路径障碍并抓取目标物体所需的动作序列。我们设计了一种新颖的多步推理过程，其基于由目标物体产生的障碍路径。我们用障碍感知的视觉线索锚定每一步推理，以激励推理能力。UNOGrasp 通过可验证的推理奖励，将监督微调与强化微调相结合。此外，我们基于 MetaGraspNetV2 构建了 UNOBench，这是一个用于训练与评测的大规模数据集，包含超过 10 万条由人工标注的障碍路径，标注内容包括障碍比例、接触点与自然语言指令。大量实验与真实机器人评估表明，UNOGrasp 在合成环境与真实环境中均显著提升了障碍物推理能力与抓取成功率，优于通用模型与专有模型。
英文原文摘要 1381 字符
Successful robotic grasping in cluttered environments not only requires a model to visually ground a target object but also to reason about obstructions that must be cleared beforehand. While current vision-language embodied reasoning models show emergent spatial understanding, they remain limited in terms of obstruction reasoning and accessibility planning. To bridge this gap, we present UNOGrasp, a learning-based vision-language model capable of performing visually-grounded obstruction reasoning to infer the sequence of actions needed to unobstruct the path and grasp the target object. We devise a novel multi-step reasoning process based on obstruction paths originated by the target object. We anchor each reasoning step with obstruction-aware visual cues to incentivize reasoning capability. UNOGrasp combines supervised and reinforcement finetuning through verifiable reasoning rewards. Moreover, we construct UNOBench, a large-scale dataset for both training and benchmarking, based on MetaGraspNetV2, with over 100k obstruction paths annotated by humans with obstruction ratios, contact points, and natural-language instructions. Extensive experiments and real-robot evaluations show that UNOGrasp significantly improves obstruction reasoning and grasp success across both synthetic and real-world environments, outperforming generalist and proprietary alternatives.