通过状态门控机制,让机器人任务执行更高效稳定。
StreamVLA: Breaking the Reason-Act Cycle via Completion-State Gating
- 采用双系统架构,仅在任务切换时触发深度推理
- 72%时间跳过复杂推理,实现98.5%任务成功率
- 适合需要长程规划与低延迟响应的机器人场景
长时序机器人操作需弥合高层规划(系统2)与底层控制(系统1)的差距。现有视觉-语言-动作(VLA)模型常在每一步重复多模态推理,导致延迟高、目标不稳定。为此,我们提出StreamVLA,一种双系统架构,将文本任务分解、视觉目标想象与连续动作生成统一于单一高效骨干网络中。引入“锁-门控”机制:仅当检测到子任务切换时,模型才触发慢思考,生成文本指令并想象具体完成状态,而非泛化未来帧。关键的是,该完成状态作为时间不变的目标锚点,使策略对执行速度变化具有鲁棒性。在稳定执行阶段,这些高层意图被锁定以条件化流匹配动作头,使模型跳过昂贵的自回归解码达72%时间。该分层抽象确保子目标聚焦的同时显著降低推理延迟。大量评估显示,StreamVLA在LIBERO基准上达到98.5%成功率,并在真实干扰场景中具备强恢复能力,相比全推理基线降低48%延迟。
原文摘要 · Abstract (English)
Long-horizon robotic manipulation requires bridging the gap between high-level planning (System 2) and low-level control (System 1). Current Vision-Language-Action (VLA) models often entangle these processes, performing redundant multimodal reasoning at every timestep, which leads to high latency and goal instability. To address this, we present StreamVLA, a dual-system architecture that unifies textual task decomposition, visual goal imagination, and continuous action generation within a single parameter-efficient backbone. We introduce a "Lock-and-Gated" mechanism to intelligently modulate computation: only when a sub-task transition is detected, the model triggers slow thinking to generate a textual instruction and imagines the specific visual completion state, rather than generic future frames. Crucially, this completion state serves as a time-invariant goal anchor, making the policy robust to execution speed variations. During steady execution, these high-level intents are locked to condition a Flow Matching action head, allowing the model to bypass expensive autoregressive decoding for 72% of timesteps. This hierarchical abstraction ensures sub-goal focus while significantly reducing inference latency. Extensive evaluations demonstrate that StreamVLA achieves state-of-the-art performance, with a 98.5% success rate on the LIBERO benchmark and robust recovery in real-world interference scenarios, achieving a 48% reduction in latency compared to full-reasoning baselines.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。