Sparse Compositional Flow Matching by geometric assembly from motion primitives
The article discusses a new framework for matching embodied trajectories in robotics using motion primitives. This approach aims to improve the efficiency of generative models by directly composing in physical trajectory space. The framework achieves state-of-the-art accuracy, significantly reducing error ratios compared to existing methods.
- ▪The framework utilizes a compositional latent structure to enhance the efficiency of embodied AI tasks.
- ▪It incorporates a Motion-Primitive Dictionary Learning method that allows for the reuse of motion fragments.
- ▪The Structural Sparse Flow Matching with Geometric Constraints generates a binary placement matrix to ensure spatial and temporal continuity.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Robotics arXiv:2605.23341 (cs) [Submitted on 22 May 2026] Title:Sparse Compositional Flow Matching by geometric assembly from motion primitives Authors:Yan Tang, Yuanbo Tang, Tingyu Cao, Shaolun Huang, Yang Li View a PDF of the paper titled Sparse Compositional Flow Matching by geometric assembly from motion primitives, by Yan Tang and 4 other authors View PDF HTML (experimental) Abstract:Embodied trajectories, such as the executable motion sequences of robotic manipulators, underwater vehicles, and mobile robots, are a fundamental output of embodied AI.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
