Event
RoboSTD: Zero-Shot Construction of Dual-Arm Training Data from Single-Arm Data
1 reports · 1 sources
Hongyu Wu and co-authors propose RoboSTD, a zero-shot single-arm-to-dual-arm data construction method for dual-arm robot learning: it performs sagittal-plane mirroring on observations, proprioception, and actions to generate executable contralateral pseudo-demonstrations, then uses an LLM to generate structured spatiotemporal coordination constraints that guide pseudo dual-arm supervision reconstruction across workspace regions and task stages. The authors state that this method requires no additional dual-arm demonstrations; evaluated in the RoboTwin simulation and real-world environment on the AgileX ALOHA platform, the average number of cups collected in the cup-collection task reached 2.11 times that of the single-arm baseline.
Report timeline
- RoboSTD: A Bimanual Learning Method for Zero-Shot Single-to-Dual Transfer via Sagittal Mirroring
RoboSTD is a zero-shot single-to-dual data construction method for bimanual robot learning. It applies sagittal-plane mirroring to observations, proprioception, and actions to generate executable contralateral pseudo-demonstrations, then uses an LLM to generate structured spatio-temporal coordination constraints that guide the reconstruction of pseudo-bimanual supervision across workspace regions and task stages. The study evaluates RoboSTD on the AgileX ALOHA platform in RoboTwin simulation and real-world settings. Without additional bimanual demonstrations, it achieves an average of 2.11 times the number of cups collected by the single-arm baseline in the cup-collection task.
arXiv cs.RO ·