Score and Distribution Matching Policy: Advanced Accelerated Visuomotor Policies via Matched Distillation
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916532183891968 |
|---|---|
| author | Jia, Bofang Ding, Pengxiang Cui, Can Sun, Mingyang Qian, Pengfang Huang, Siteng Fan, Zhaoxin Wang, Donglin |
| author_facet | Jia, Bofang Ding, Pengxiang Cui, Can Sun, Mingyang Qian, Pengfang Huang, Siteng Fan, Zhaoxin Wang, Donglin |
| contents | Visual-motor policy learning has advanced with architectures like diffusion-based policies, known for modeling complex robotic trajectories. However, their prolonged inference times hinder high-frequency control tasks requiring real-time feedback. While consistency distillation (CD) accelerates inference, it introduces errors that compromise action quality. To address these limitations, we propose the Score and Distribution Matching Policy (SDM Policy), which transforms diffusion-based policies into single-step generators through a two-stage optimization process: score matching ensures alignment with true action distributions, and distribution matching minimizes KL divergence for consistency. A dual-teacher mechanism integrates a frozen teacher for stability and an unfrozen teacher for adversarial training, enhancing robustness and alignment with target distributions. Evaluated on a 57-task simulation benchmark, SDM Policy achieves a 6x inference speedup while having state-of-the-art action quality, providing an efficient and reliable framework for high-frequency robotic tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_09265 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Score and Distribution Matching Policy: Advanced Accelerated Visuomotor Policies via Matched Distillation Jia, Bofang Ding, Pengxiang Cui, Can Sun, Mingyang Qian, Pengfang Huang, Siteng Fan, Zhaoxin Wang, Donglin Robotics Machine Learning Visual-motor policy learning has advanced with architectures like diffusion-based policies, known for modeling complex robotic trajectories. However, their prolonged inference times hinder high-frequency control tasks requiring real-time feedback. While consistency distillation (CD) accelerates inference, it introduces errors that compromise action quality. To address these limitations, we propose the Score and Distribution Matching Policy (SDM Policy), which transforms diffusion-based policies into single-step generators through a two-stage optimization process: score matching ensures alignment with true action distributions, and distribution matching minimizes KL divergence for consistency. A dual-teacher mechanism integrates a frozen teacher for stability and an unfrozen teacher for adversarial training, enhancing robustness and alignment with target distributions. Evaluated on a 57-task simulation benchmark, SDM Policy achieves a 6x inference speedup while having state-of-the-art action quality, providing an efficient and reliable framework for high-frequency robotic tasks. |
| title | Score and Distribution Matching Policy: Advanced Accelerated Visuomotor Policies via Matched Distillation |
| topic | Robotics Machine Learning |
| url | https://arxiv.org/abs/2412.09265 |