发表机构
George Mason University; Kiel University; Ludwig Maximilian University Munich; Ewha Womans University（乔治梅森大学; 基尔大学; 慕尼黑大学; 梨花女子大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出Hydra模型，通过将规划器移入模型内部消除生成模型与规划器的表示不对齐问题，结合离散潜在规划与条件流匹配，在物理机器人平台上的目标导向规划与闭环执行能力表现优异。
AI 中文摘要
世界模型使机器人能够想象可能的未来，但将这一能力用于实时控制受到表示不对齐的瓶颈限制：生成模型与规划器在解耦的流形上运行，因此规划器没有可搜索的共享结构，必须将每个候选解码回高维像素空间才能评估。这一步解码是物理硬件实时控制的主要障碍。本文提出Hydra，一种离散世界动作模型，通过将规划器（采样器与评估器）移入模型内部来消除这一差距。Hydra在视觉状态、物理位姿与控制动作上建立统一潜在流形，再通过特定模态的向量量化瓶颈将该流形压缩为运动学动力学意图与视觉状态的离散词汇表。由于候选现在直接从该共享流形中抽取，采样由模型自身对观测的理解驱动而非盲目提议，评估则在离散空间内原生进行：候选按运动学感知成本排序，无需解码为像素。我们将此称为离散潜在规划（Discrete Latent Planning，DLP）。由于仅在离散意图上规划无法提供物理致动所需的平滑连续指令，Hydra将DLP与条件流匹配（conditional Flow Matching）结合，后者将每个选定意图映射为执行用的连续轨迹。在两个物理机器人平台上评估显示，Hydra在目标导向规划方面优于现有最优世界模型，同时达到或超过领先反应式基础策略的闭环执行能力。
英文摘要
World models let robots imagine possible futures, but exploiting this capability for real-time planning is bottlenecked by a representation misalignment: generative models and planners operate on decoupled manifolds, requiring computationally expensive decoding of every candidate back to the high-dimensional observation space for evaluation. In this paper, we present Hydra, a discrete World Action Model that tackles this by establishing a unified latent manifold over visual states, physical poses, and control actions. By compressing this manifold through modality-specific Vector-Quantized bottlenecks, Hydra yields discrete vocabularies of kinodynamic intents and visual states. This enables Discrete Latent Planning (DLP), where candidates are sampled directly from the shared manifold and ranked by a Kinematic-Perceptual Cost within the discrete latent space. To bridge discrete planning with the continuous commands required for physical actuation, Hydra pairs DLP with conditional Flow Matching to map selected intents to smooth execution trajectories. Evaluated on two physical robotic platforms, Hydra outperforms state-of-the-art navigation world models in goal-directed planning, while matching or exceeding the closed-loop execution capabilities of leading reactive navigation policies.
Comments28 pages, 12 figures. https://robotixx.github.io/hydra
