NVIDIA 教程：用 AI 智能体训练跨实体机器人导航策略
原标题：How to Train a Cross-Embodiment Robot Navigation Policy with AI Agents
AI 摘要
NVIDIA 技术博客发布教程，介绍如何使用 AI 智能体驱动的 COMPASS 工作流训练跨实体机器人导航策略。该工作流通过编码智能体（如 Codex）自动完成依赖验证、场景准备、训练、诊断和评估，并以 Spot 机器人为参考，支持内置场景和 SAGE-10K 场景。COMPASS 框架利用预训练的 X-Mobility 策略和残差强化学习，减少重复训练成本，并支持将多个专家策略蒸馏为共享策略。
正文节选
Navigation enables a robot to turn perception and motion into purposeful autonomy. Unlike locomotion, which produces stable movement, navigation must be used to continuously localize the robot, interpret changing surroundings, select a route, and avoid obstacles to reach a goal safely. Moving this capability to a new robot or scene can require new data, simulation assets, robot interfaces, training, diagnosis, and evaluation. Repeating that work for every robot-scene pair is expensive and diffic