论文

任务收敛流匹配:物理仿真推演提升机器人操作成功率

Task Convergent Flow Matching for Generative Robotics

精选理由

生成式机器人策略常出现动作块之间互相抵消、原地打转的问题。这篇论文用物理仿真 rollout 加 CMA-ES 搜索在推理阶段精炼流匹配策略,PushT 成功率从 31.3% 拉到 88.5%,同时用径向带约束把分布外比例压到与演示数据相当。做机器人操作和策略部署的研究者值得复现看看。

论文提出一种面向流匹配机器人策略的推理时精炼框架,核心思路是用并行物理仿真 rollout 评估候选动作块,只保留满足任务收敛充分条件的运动。PushT 任务成功率从 31.3% 提升到 88.5%,三立方体堆叠从 77.0% 提升到 86.0%。

正文

技术机制上,框架用任务级性能漏斗(funnel)约束每个动作块的净进展,允许临时回退但要求整体推进,并通过区间 Lipschitz 界验证连续时间漏斗条件。针对接触丰富的混合动力学,采用无导数的 CMA-ES 进化策略精炼流的源种子,并限制在源分布的径向分位数带内以保留已学到的运动先验。方法同时支持欧氏空间和位姿值 Riemannian Flow Matching(RMFM)策略。

对比效果方面,在 PushT 上一条学习到的漏斗将已达成进展的平均最大损失降低 19%,且成功率与仅约束动作块终点的方式相当。带径向带约束时执行窗口的分布外比例为 11.3%,与 held-out 演示数据范围接近,而无约束搜索高达 34.1%,说明精炼过程没有破坏策略的运动先验。

原文 · arXiv cs.LG

Task Convergent Flow Matching for Generative Robotics

Successful manipulation requires the manipulated objects to progress toward a desired task outcome. Generative policies can produce plausible robot motions without ensuring this progress, allowing successive action chunks to stall or undo earlier changes to the scene. We introduce a task-centric inference-time refinement framework for flow matching policies that evaluates candidate action chunks through parallel physics-based rollouts and searches for motions satisfying sufficient conditions for task convergence. Task-level performance funnels bound temporary regressions while requiring net progress over each chunk. We establish sufficient conditions under which repeated funnel satisfaction yields convergence throughout execution, and use interval-wise Lipschitz bounds to verify continuous-time funnel satisfaction from sampled rollouts. To accommodate contact-rich hybrid dynamics, we refine the flow's source seeds using derivative-free Covariance Matrix Adaptation Evolution Strategy (CMA-ES), constrained to radial quantile bands of the source distribution to help retain the learned motion prior. This formulation supports both Euclidean and pose-valued Riemannian Flow Matching (RMFM) policies. On PushT and three-cube stacking, refinement raises success from 31.3% to 88.5% and from 77.0% to 86.0%. On PushT, whose coverage score responds continuously to the motion of the block, a learned funnel lowers the average maximum loss of achieved progress by 19% relative to constraining the endpoint of each chunk alone, at a similar success rate. With the radial band, 11.3% of the executed motion windows are out of distribution, a rate within the range of held-out demonstrations, compared with 34.1% for an unconstrained search.

产业坐标与信源核验

本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:Flow Matching、CMA-ES、Riemannian Flow Matching、机器人操作,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。

信源核验说明:本文原始线索采编自「arXiv cs.LG」(@Lorenzo Paiola, Maximilian Stölzle, Giorgio Grioli, Daniela Rus, Cosimo Della Santina),经 AITOP 多维评分管线结构化提炼。