标题： 人形生成：通过大语言模型推理进行双臂灵巧操作的数据生成 显示英文标题

标题： HumanoidGen: Data Generation for Bimanual Dexterous Manipulation via LLM Reasoning

摘要： 对于机器人操作，现有的机器人数据集和仿真基准主要针对机械臂平台。 然而，对于配备双臂和灵巧手的人形机器人，仿真任务和高质量演示明显不足。 双手灵巧操作本质上更加复杂，因为它需要协调的臂部运动和手部操作，使得自主数据收集具有挑战性。 本文提出了HumanoidGen，一个自动任务创建和演示收集框架，该框架利用原子灵巧操作和大语言模型推理来生成关系约束。 具体来说，我们基于原子操作为资产和灵巧手提供空间注释，并使用大语言模型规划器根据物体功能和场景生成一系列可操作的空间约束链。 为进一步提高规划能力，我们采用蒙特卡洛树搜索的一种变体来增强大语言模型在长时域任务和注释不足情况下的推理能力。 在实验中，我们创建了一个带有增强场景的新基准来评估收集数据的质量。 结果表明，2D和3D扩散策略的性能可以随着生成的数据集进行扩展。 项目页面是https://openhumanoidgen.github.io. 显示英文摘要

摘要： For robotic manipulation, existing robotics datasets and simulation benchmarks predominantly cater to robot-arm platforms. However, for humanoid robots equipped with dual arms and dexterous hands, simulation tasks and high-quality demonstrations are notably lacking. Bimanual dexterous manipulation is inherently more complex, as it requires coordinated arm movements and hand operations, making autonomous data collection challenging. This paper presents HumanoidGen, an automated task creation and demonstration collection framework that leverages atomic dexterous operations and LLM reasoning to generate relational constraints. Specifically, we provide spatial annotations for both assets and dexterous hands based on the atomic operations, and perform an LLM planner to generate a chain of actionable spatial constraints for arm movements based on object affordances and scenes. To further improve planning ability, we employ a variant of Monte Carlo tree search to enhance LLM reasoning for long-horizon tasks and insufficient annotation. In experiments, we create a novel benchmark with augmented scenarios to evaluate the quality of the collected data. The results show that the performance of the 2D and 3D diffusion policies can scale with the generated dataset. Project page is https://openhumanoidgen.github.io.