LINGO-Space：面向空间的语言条件增量 Grounding
LINGO-Space: Language-Conditioned Incremental Grounding for Space
AI总结：
针对复合空间指称表达式的定位与歧义问题，本文提出 LINGO-Space，利用可配置极坐标概率分布进行增量空间 grounding，并在桌面操作、模拟操作及真实四足导航任务中验证其有效性。
摘要：
我们旨在解决对指称空间的复合指令进行空间定位的问题：space grounding（空间 grounding）。与当前的实例 grounding 相比，空间 grounding 具有挑战性，原因在于识别离散表达式所指位置的不适定性，以及指称表达式的组合歧义。因此，我们提出一种新颖的概率空间 grounding 方法（LINGO-Space），该方法能够准确识别所指空间的概率分布，并利用可配置的极坐标分布，根据后续指称表达式对其进行增量更新。我们的评估表明，使用极坐标分布进行估计使机器人能够在 20 项桌面操作基准测试中成功完成位置 grounding。我们还表明，更新该分布有助于 grounding 方法准确缩小指称空间。最后，我们通过模拟操作和真实四足机器人导航任务展示了空间 grounding 的鲁棒性。代码和视频可在 https://lingo-space.github.io 获取。
英文摘要：
We aim to solve the problem of spatially localizing composite instructions referring to space: space grounding. Compared to current instance grounding, space grounding is challenging due to the ill-posedness of identifying locations referred to by discrete expressions and the compositional ambiguity of referring expressions. Therefore, we propose a novel probabilistic space-grounding methodology (LINGO-Space) that accurately identifies a probabilistic distribution of space being referred to and incrementally updates it, given subsequent referring expressions leveraging configurable polar distributions. Our evaluations show that the estimation using polar distributions enables a robot to ground locations successfully through $20$ table-top manipulation benchmark tests. We also show that updating the distribution helps the grounding method accurately narrow the referring space. We finally demonstrate the robustness of the space grounding with simulated manipulation and real quadruped robot navigation tasks. Code and videos are available at https://lingo-space.github.io.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
