Frequency-Aware Flow Matching for Continuous and Consistent Robotic Action Generation
Frequency-Aware Flow Matching for Continuous and Consistent Robotic Action Generation
This research introduces Frequency-Aware Flow Matching (FAFM), a method designed to generate continuous and temporally consistent robotic actions. By transforming discrete action sequences into the frequency domain using the discrete cosine transform (DCT) and applying flow matching over coefficients, FAFM addresses brittleness in existing methods that rely on discretized chunks. The approach also regularizes the first-order temporal derivative to suppress high-frequency errors and discourage abrupt changes. Evaluated across synthetic benchmarks (toy, obstacle avoidance, LapGym, LIBERO) and deployed on a real-world Franka robot, FAFM demonstrates improvements in success rates, multimodal expressivity, motion smoothness, convergence speed, and robustness to mechanical bias and mixed-frequency input without introducing additional network parameters.
The paper introduces Frequency-Aware Flow Matching (FAFM), a method that transforms discrete action sequences into the frequency domain using DCT to handle heterogeneous control frequencies and regularizes temporal derivatives for smoothness, demonstrating improved success rates and motion smoothness on synthetic benchmarks and a real-world Franka robot.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- code: Available Artifact link Open source S5
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ion models. Across synthetic toy benchmark, obstacle avoidance, LapGym, and LIBERO, FAFM improves success rates, multimodal expressivity, motion smoothness, convergence speed, robustness to mechanical bias and mixed-frequency input. These gains are consistent when deployed on a real-world Franka robot. Code available at https://github.com/gakakulicc/FAFM.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org