Traj-LeWM: Path-Aware World-Model Planning via Latent Trajectory Cost
作者: Xiaodi Huang, Ziyi Ding, Jingtian Wan, Yuchen Liu, Yuan Zhang, Xiao-Ping Zhang, Jiayu Chen, Zhang Zhang, Tao Huang
分类: cs.AI
发布日期: 2026-08-14
💡 一句话要点
提出Traj-LeWM以解决LeWM在路径规划中的局限性
🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture)
关键词: 路径规划 潜在轨迹成本 视觉世界模型 机器人导航 动态学习
📋 核心要点
- 现有方法LeWM在训练时仅关注局部下一步转移,未能全面评估完整轨迹的表现,导致规划效果不佳。
- Traj-LeWM通过引入目标条件的潜在轨迹成本(LTC),在保留LeWM局部动态目标的同时,增强了对完整轨迹的评估。
- 实验结果显示,Traj-LeWM在Push-T、OGBench-Cube、Reacher和Two-Room任务上分别提升了3%、14%、7%和7%的性能,验证了其有效性。
📝 摘要(中文)
LeWM是一种轻量级视觉世界模型,通过从像素中学习潜在动态并根据预测端点与目标之间的距离对候选动作序列进行排名。然而,LeWM存在两个主要局限性:首先,在训练过程中,它仅学习局部的下一步转移,而未评估完整轨迹相对于任务目标的表现;其次,在规划阶段,它仅通过预测端点距离来排名候选者。为了解决这些问题,本文提出了Traj-LeWM,引入了目标条件的潜在轨迹成本(LTC),以聚合轨迹级信息作为补充信号。实验结果表明,Traj-LeWM在多个任务上均优于LeWM。
🔬 方法详解
问题定义:本文旨在解决LeWM在路径规划中的局限性,特别是其在训练和规划阶段对完整轨迹信息的忽视,导致候选动作序列的评估不够全面。
核心思路:Traj-LeWM通过引入目标条件的潜在轨迹成本(LTC),在训练中补充下一步预测,并在规划中结合LTC与端点距离,增强了对路径信息的利用。
技术框架:Traj-LeWM的整体架构包括两个主要模块:一是基于LTC的轨迹偏好监督,用于训练阶段的共享表示塑造;二是结合LTC和端点距离的候选排名机制,用于规划阶段。
关键创新:Traj-LeWM的核心创新在于引入LTC作为补充信号,使得模型不仅关注端点距离,还考虑完整轨迹的演变,从而提高了路径规划的准确性。
关键设计:在模型设计中,LTC的计算方式与损失函数的设置至关重要,确保了轨迹级信息的有效聚合与利用,同时保持了LeWM的局部动态目标。
🖼️ 关键图片
📊 实验亮点
Traj-LeWM在多个任务上均优于LeWM,具体表现为在Push-T任务上提升3%,OGBench-Cube提升14%,Reacher提升7%,Two-Room提升7%。这些结果通过控制实验和消融实验进一步验证了轨迹级表示塑造和路径感知候选排名的互补作用。
🎯 应用场景
该研究的潜在应用领域包括机器人路径规划、自动驾驶车辆导航以及智能制造等场景。通过提高路径规划的准确性,Traj-LeWM能够在复杂环境中实现更高效的任务执行,具有重要的实际价值和未来影响。
📄 摘要(原文)
LeWM is a lightweight visual world model that learns latent dynamics end-to-end from pixels and ranks candidate action sequences by the distance between their predicted endpoints and the goal. However, LeWM has two limitations. First, during training, it learns local next-step transitions without evaluating complete trajectories relative to the task goal. Second, during planning, it ranks candidates solely by predicted endpoint distance. Because model predictions may differ from actual execution outcomes, the candidate whose predicted endpoint is closest to the goal may not perform best when executed in the environment. The evolution of the complete predicted trajectory can therefore provide complementary information beyond endpoint distance. To address these limitations, we propose Traj-LeWM, which retains LeWM's local-dynamics objective and endpoint score while introducing a goal-conditioned latent trajectory cost (LTC) that aggregates trajectory-level information as a complementary signal. During training, LTC-based trajectory-preference supervision complements next-step prediction in shaping the shared representation. During planning, LTC is combined with endpoint distance to incorporate intermediate-path information into candidate ranking. With joint endpoint-plus-LTC scoring, Traj-LeWM outperforms LeWM on Push-T, OGBench-Cube, Reacher, and Two-Room by $3$, $14$, $7$, and $7$ percentage points, respectively. Controlled experiments and ablations further verify the complementary roles of trajectory-level representation shaping and path-aware candidate ranking.