Hierarchical Policy Learning via Spectral Decomposition
About
In this paper, we identify a semantic decomposition in robot action sequences, separating task-level motion intent from execution-level refinements. By analyzing actions in the spectral domain using the discrete cosine transform (DCT), we observe that low-frequency components capture global motion trajectories, while high-frequency components encode precise timing, alignment, and contact behaviors. Motivated by this structure, we propose Causal Spectral Policy (CSP), which models action generation as a causal coarse-to-fine process: coarse motion is predicted from observation and language, and fine corrections are generated conditionally on the realized trajectory. Across simulation and real-world evaluations, CSP consistently outperforms strong baselines on precision-sensitive manipulation tasks. Additionally, we propose human-inspired teleoperation noise injection as a data augmentation method, under which our approach demonstrates strong robustness to noisy demonstrations.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Robot Manipulation | LIBERO | -- | 1025 | |
| Robot Manipulation | MimicGen | Coffee Success Rate83.8 | 25 | |
| Block-stacking | Real-robot Franka Emika Panda (real-world) | Success Rate90 | 11 | |
| Close candle lid | Real Robot Franka Emika Panda (Real-world evaluation) | Success Rate7 | 4 | |
| Press C | Real Robot Franka Emika Panda (Real-world evaluation) | Success Rate8 | 4 | |
| Stack thin block | Real Robot Franka Emika Panda (Real-world evaluation) | Success Rate90 | 4 | |
| Press Enter Key | Real Robot Franka Emika Panda (Real-world evaluation) | Success Rate100 | 4 |