从基础到应用：在实践中改进VLA模型

摘要

尽管VLA基础模型近期取得了进展，但实验室条件与现实应用之间的差距仍阻碍其实用化。为弥合这一鸿沟，我们提出LingBot-VLA 2.0，通过三大功能领域的改进实现了对LingBot-VLA的升级：(1) 跨任务与跨本体的泛化能力。相较前代版本，我们重构了数据处理流程，并筛选了约6万小时数据用于预训练，其中包括涵盖20种机器人构型的5万小时机器人轨迹数据，以及1万小时第一人称人类视频数据。(2) 在双机械臂硬件平台基础上扩展动作空间。具体而言，本系统支持头部、腰部、移动底盘及灵巧手的自由度控制，使机器人能够应对实际场景中更复杂的任务。(3) 基于预测动态建模提升时序推理能力。我们通过视频表征模型获取语义先验、深度估计模型获取几何线索，将未来状态预测作为代理任务进行建模。在通用化设置下的GM-100基准测试中，上述改进的有效性得到了验证。此外，受益于覆盖全身自由度的扩展预训练数据，LingBot-VLA-2.0在两个机器人平台上展现出强大的跨本体长时域移动操作能力。

English

Despite recent progress of VLA foundation models, the disparity between laboratory conditions and real-world applications continues to impede their practical implementation. To bridge this gap, we present LingBot-VLA 2.0, which advances LingBot-VLA through improvements in three functional domains. (1) Generalization across tasks and embodiments. Compared to the previous version, we revamp the data processing pipeline and curate around 60,000 hours of data for pretraining, including 50,000 hours of robot trajectories spanning 20 robot configurations and 10,000 hours of egocentric human videos. (2) Expanded action space in addition to dual-arm hardware platforms. In particular, our system accommodates degrees of freedom for the heads, waists, mobile bases, and dexterous hands, thereby empowering the robots to tackle more complex tasks in practical scenarios. (3) Predictive dynamics modeling for improved temporal reasoning. Specifically, we formulate future prediction as a proxy task, facilitated by a video representation model for semantic priors and a depth estimation model for geometric cues. Evaluations on the GM-100 benchmark, conducted in a generalist setting, validate the beneficial impact of these proposed modifications. Furthermore, benefiting from the expanded pretraining data that covers whole-body degrees of freedom, LingBot-VLA-2.0 demonstrates strong cross-embodiment long-horizon mobile manipulation capability across the two robotic platforms.