外观
论文arXiv 2405.122132024-05-20
Octo开源通用机器人策略：80万轨迹预训练支持消费级GPU微调
本文由 AI 根据公开内容整理，可能有误，请以原文为准：Octo: An Open-Source Generalist Robot Policy（arXiv 2405.12213）
核心内容
研究团队推出开源通用机器人操作策略Octo，旨在解决跨平台机器人学习中的数据稀缺与适配难题。该模型基于目前最大的机器人操作数据集Open X-Embodiment中的80万条轨迹进行预训练，支持通过自然语言命令或目标图像进行任务指定。核心突破在于其高效的迁移能力，能够在标准消费级GPU上于数小时内完成针对新机器人 setups 的微调，显著降低了通用策略在特定场景落地的算力与数据门槛。
技术架构上，Octo采用Transformer作为骨干网络，专门设计以兼容多样化的传感器输入与动作空间。研究在9个不同的机器人平台上进行了广泛实验，证实其作为通用策略初始化权重的有效性，能快速适应新的观测模态与控制接口。此外，论文提供了从模型架构到训练数据配比的详细消融实验，系统分析了构建通用机器人基础模型的关键设计决策，为后续具身智能算法研发提供了可复现的工程基准与优化方向。
关键看点
- 数据规模：基于Open X-Embodiment数据集80万条轨迹预训练，覆盖多样化机器人操作场景
- 微调效率：在标准消费级GPU上数小时内即可完成新传感器与动作空间的领域适配
- 交互模态：支持自然语言指令与目标图像双重条件输入，提升任务泛化与语义对齐能力
- 跨平台验证：在9种不同机器人本体上实测，验证了策略初始化权重对新观测/动作空间的兼容性
工程与产业观察
Octo确立了开源通用操作策略的工程范式，证明了基于大规模异构数据预训练结合轻量级微调的可行性。其消费级GPU适配特性大幅降低了具身大模型研发门槛，消融实验则为算法工程师优化Transformer策略架构、数据配比及多模态对齐提供了关键实证参考，加速通用操作策略从学术验证走向产业应用。
英文摘要 (Abstract)
Large policies pretrained on diverse robot datasets have the potential to transform robotic learning: instead of training new policies from scratch, such generalist robot policies may be finetuned with only a little in-domain data, yet generalize broadly. However, to be widely applicable across a range of robotic learning scenarios, environments, and tasks, such policies need to handle diverse sensors and action spaces, accommodate a variety of commonly used robotic platforms, and finetune readily and efficiently to new domains. In this work, we aim to lay the groundwork for developing open-source, widely applicable, generalist policies for robotic manipulation. As a first step, we introduce Octo, a large transformer-based policy trained on 800k trajectories from the Open X-Embodiment dataset, the largest robot manipulation dataset to date. It can be instructed via language commands or goal images and can be effectively finetuned to robot setups with new sensory inputs and action spaces within a few hours on standard consumer GPUs. In experiments across 9 robotic platforms, we demonstrate that Octo serves as a versatile policy initialization that can be effectively finetuned to new observation and action spaces. We also perform detailed ablations of design decisions for the Octo model, from architecture to training data, to guide future research on building generalist robot models.
作者：Octo Model Team, Dibya Ghosh, Homer Walke, Karl Pertsch, Kevin Black, Oier Mees, Sudeep Dasari, Joey Hejna, Tobias Kreiman, Charles Xu, Jianlan Luo, You Liang Tan, Lawrence Yunliang Chen, Pannag Sanke…
来源
本页事实对应的公开页面，以原文为准。