LLM作为机器人大脑：由大型语言模型生成的云驱动机器人动作序列 | KC23

LLM as a Robotic Brain: Cloud-Driven Robot Action Sequences Generated by Large... - Sitong Mao

LLM作为机器人大脑：由大型语言模型生成的云驱动机器人动作序列 | LLM as a Robotic Brain: Cloud-Driven Robot Action Sequences Generated by Large Language Models - Sitong Mao, Huawei Cloud

具身人工智能在机器人领域至关重要，需要为机器人在物理环境中执行任务生成连贯的动作序列。通过处理自然语言指令和环境观察，大型基础模型已经引起了生成计划的兴趣。高性能的大型语言模型，如斯坦福Alpaca，为相关技术领域开辟了新的可能性。在大型模型和边缘机器人之间实现统一行为定义和平衡资源冲突仍然存在挑战。为了解决这些问题，我们的云原生技术兴趣小组提出了RoboPilot，这是一个在KubeEdge社区中的开源项目。它将大型语言模型与边缘云协作相结合，实现了在开放环境中灵活部署和执行真实世界机器人的能力。这项研究突出了边缘云协作在实现先进机器人能力和推进云原生机器人研究方面的潜力。

Embodied AI is vital in robotics, requiring the generation of coherent action sequences for robots to perform tasks in physical environments. Large foundation models have gained interest in generating plans by processing natural language instructions and environmental observations. High-performance large language models, such as Stanford Alpaca, have opened up possibilities in related technical fields. Challenges remain in achieving unified behavior definition and balancing resource conflicts between large models and edge robots. To address these, our cloud-native technology interest group proposes RoboPilot, an open-source project in the KubeEdge community. It combines large language models with edge-cloud collaboration, enabling flexible deployment and execution of real-world robots in open environments. This research highlights edge-cloud collaboration's potential in enabling advanced robot capabilities and advancing cloud-native robotics research.

CNCF概况（幻灯片）

扫描二维码联系我们！

CNCF (Cloud Native Computing Foundation)成立于2015年12月，隶属于Linux  Foundation，是非营利性组织。 

CNCF（云原生计算基金会）致力于培育和维护一个厂商中立的开源生态系统，来推广云原生技术。我们通过将最前沿的模式民主化，让这些创新为大众所用。请关注CNCF微信公众号。