Composes pre-trained unimanual robotic policies into complex bimanual tasks without requiring bimanual demonstration data.
March 24, 2026
Original Paper
EnergyAction: Unimanual to Bimanual Composition with Energy-Based Models
arXiv · 2603.20236
The Takeaway
Bimanual data is notoriously scarce; EnergyAction uses Energy-Based Models to fuse existing single-arm policies while enforcing spatial-temporal constraints, effectively doubling the utility of existing unimanual datasets.
From the abstract
Recent advances in unimanual manipulation policies have achieved remarkable success across diverse robotic tasks through abundant training data and well-established model architectures. However, extending these capabilities to bimanual manipulation remains challenging due to the lack of bimanual demonstration data and the complexity of coordinating dual-arm actions. Existing approaches either rely on extensive bimanual datasets or fail to effectively leverage pre-trained unimanual policies. To a