Spectrally Routed Diffusion Policy for Multi-Task Robotic Manipulation
Spectrally Routed Diffusion Policy for Multi-Task Robotic Manipulation
This research presents SR-DP (Spectrally Routed Diffusion Policy), a method designed to improve multi-task robotic manipulation by addressing limitations in convolutional encoders and gradient interference during joint training. The approach uses a frozen VGGT encoder with spatial-compression adapters and decomposes linear layers via SVD into orthogonal experts adapted through spectral routing. Evaluated on six RoboTwin tasks, SR-DP achieves an average success rate of 68.4%, outperforming multi-task VO-DP by 21.8 points and the strongest per-task baseline by 3.8 points. On four real-world bimanual tasks, Stage 2 reaches a 43.0% average success rate, improving over Stage 1 by 19.0 points.
The paper introduces SR-DP, a diffusion policy architecture using spectral routing and SVD-decomposed experts to mitigate interference in multi-task robotic manipulation, reporting success rates on RoboTwin and real-world bimanual tasks.
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.
Reported numbers
Reported success rate
68.4%
Reported trials
View original evidence
visual condition for tasks that share a workspace. We first train the denoiser as a standard diffusion policy, then decompose selected linear layers by SVD into a frozen basis of orthogonal rank-one experts, adapted through spectral routing and Householder input rotations. On six RoboTwin tasks, SR-DP averages a 68.4% success rate, surpassing multi-task VO-DP by 21.8 points and the strongest per-task baseline by 3.8
Open source S4
- paper: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
visual condition for tasks that share a workspace. We first train the denoiser as a standard diffusion policy, then decompose selected linear layers by SVD into a frozen basis of orthogonal rank-one experts, adapted through spectral routing and Householder input rotations. On six RoboTwin tasks, SR-DP averages a 68.4% success rate, surpassing multi-task VO-DP by 21.8 points and the strongest per-task baseline by 3.8
Open source S4
points. On four real-world bimanual tasks, Stage 2 reaches a 43.0% average success rate and improves over Stage 1 by 19.0 points.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org