CausGround: Causal Interaction Grounding for Adaptive Embodied Planning

📄 arXiv: 2610.29702v1 📥 PDF

作者: Yijun Shen, Diqi He, Jinbang Guo, Wanyue Zhang, Jiaming Mi, Huan Yang, Zhongyuan Wang, Zhiwei Yu, Kai Zhang

分类: cs.RO

发布日期: 2026-10-21

备注: Project Page: https://causground.github.io

💡 一句话要点

提出CausGround以解决多模态大语言模型的物理交互规划问题

🎯 匹配领域: 支柱一：机器人控制 (Robot Control) 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 因果推理 多模态大语言模型 物理交互 机器人规划 可重用知识

📋 核心要点

  1. 现有方法在处理物理交互时，往往需要昂贵的环境特定训练或依赖难以重用的非结构化交互经验。
  2. CausGround通过非参数化的因果交互基础框架，将交互经验转化为可重用的因果知识，支持高效的动作选择和规划。
  3. 在VirtualHome和RoboCasa基准测试中，CausGround显著提高了基于视觉的高层次规划成功率，验证了其有效性。

📝 摘要（中文）

多模态大语言模型（MLLMs）展现出良好的推理能力，但可靠的物理交互需要考虑动作的可执行性及其对环境的影响。这些前提条件和效果依赖于环境，现有方法往往无法有效捕捉，导致适应性挑战。为此，本文提出CausGround，一个非参数的因果交互基础框架，将交互经验提炼为可重用的以动作为中心的因果知识。CausGround识别动作的前提条件和效果，并将其组织在一个因子化的世界状态表示中，捕捉对象属性和对象间关系。在规划时，任务相关的因果依赖基于当前观察进行基础化，以支持对前提条件的反向推理和对效果的正向推理，从而实现动作选择和多步规划。

🔬 方法详解

问题定义：本文旨在解决多模态大语言模型在物理交互中的规划问题，现有方法面临环境适应性差、训练成本高等痛点。

核心思路：CausGround通过提炼交互经验为因果知识，构建可重用的因果框架，以支持高效的动作选择和多步规划。

技术框架：CausGround的整体架构包括因果知识提取、状态表示和推理模块。在规划时，基于当前观察进行因果依赖的基础化，支持反向和正向推理。

关键创新：CausGround的核心创新在于其非参数化的因果交互基础框架，能够有效捕捉动作的前提条件和效果，与现有方法相比，避免了昂贵的环境特定训练。

关键设计：在设计中，CausGround采用因子化的世界状态表示，强调对象属性和对象间关系的捕捉，确保因果知识的重用性和适应性。通过精确的参数设置和损失函数设计，提升了模型的推理能力。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在VirtualHome和RoboCasa基准测试中，CausGround将任务成功率从24.9%提升至53.3%，闭环规划成功率从43.0%提升至57.0%。在SWITCH数据集上，交互推理准确率从25.8%提升至31.9%，显示出显著的性能提升。

🎯 应用场景

CausGround的研究成果在机器人操作、智能家居和自动化系统等领域具有广泛的应用潜力。通过提供可重用的因果知识，该框架能够支持复杂环境中的长期规划和决策，推动智能体在动态环境中的适应能力和效率提升。

📄 摘要（原文）

Multimodal large language models (MLLMs) show promising reasoning capabilities, but reliable physical interaction requires planning that accounts for when actions are executable and how they change the world. These preconditions and effects are environment-dependent and often not captured by MLLMs, making adaptation challenging. Yet existing approaches either require costly environment-specific training or rely on unstructured interaction experience that is difficult to reuse. Therefore, we introduce CausGround, a non-parametric causal interaction grounding framework that distills interaction experience into reusable action-centered causal knowledge. CausGround identifies action preconditions and effects and organizes them in a factorized world-state representation capturing object attributes and inter-object relations. At planning time, task-relevant causal dependencies are grounded in current observations to support backward reasoning over preconditions and forward reasoning over effects for action selection and multi-step planning. We evaluate CausGround on simulated high-level planning, interaction reasoning, and real-world robot manipulation. On adapted VirtualHome and RoboCasa benchmarks, CausGround improves vision-based high-level planning across MLLM backbones, increasing average task success from 24.9% to 53.3% on VirtualHome and closed-loop planning success from 43.0% to 57.0% on RoboCasa. On SWITCH, average interaction reasoning accuracy increases from 25.8% to 31.9%, indicating improved action selection, recovery, and verification. Moreover, real-world experiments show that CausGround reasons backward from MLLM-interpreted goals to derive executable high-level action sequences for sequential manipulation. These results demonstrate that reusable action preconditions and effects support long-horizon embodied planning across diverse environments without parameter updates.