[论文] CoDA: Co-Evolution of Data Engines and Action Policies

论文概要

研究领域: ML
作者: Tianjiao Yu, Zhilin Ou, Yifan Shen, Xinzhuo Li, Yuanzhe Liu, Onkar Kishor Susladkar, Fan Wang, Ismini Lourentzou
发布时间: 2026-10-15
arXiv: 2610.22362

中文摘要

生成式仿真可以自动化机器人环境构建，但物理上有效的环境不一定对训练有用。挑战在于识别能解决策略失败的环境和有助于克服失败的监督。我们引入 CoDA，这是一个仿真数据引擎和动作策略通过共享异构记忆共同进化的框架。记忆将视觉-语言评论者的已验证诊断编译为重现困难的生成程序和帮助策略克服困难的训练程序。我们进一步提出 Text2potential，将诊断转换为分阶段势能用于动作块价值学习，为策略更新提供密集监督，同时在标准塑造假设下保持任务的最优策略。在四个操作任务族中，CoDA 将监督微调后的分布外成功率从 5.3% 提高到 27.6%，在相同生成预算下超过最强闭环基线 9.5 个百分点。在 LIBERO-Pro 上，15 轮生成和训练将位置扰动下的成功率从 25% 提高到 84%，指令扰动下从 1% 提高到 76%。

原文摘要

Generative simulation can automate robot environment construction, but physically valid environments are not necessarily useful for training. The challenge is to identify environments that address a policy's failures and supervision that helps it overcome them. We introduce CoDA, a framework in which a simulation data engine and an action policy co-evolve through a shared heterogeneous memory of environment construction, policy failures, and learning progress. The memory compiles validated diagnoses from a vision-language critic into generation programs that recreate difficulties and training programs that help the policy overcome them. We further propose Text2potential to convert these diagnoses into staged potentials for action-chunk value learning, providing dense supervision for policy...

自动采集于 2026-10-17

#论文 #arXiv #ML #小凯