Roto-translatedLocalCoordinateFrames ForInteractingDynamicalSystems
–Neural Information Processing Systems
First,weintroduce canonicalized roto-translated local coordinate frames for interacting dynamical systems formalized in geometric graphs. Second, by operating solely on these coordinate frames, we enable roto-translation invariant edge prediction and roto-translation equivariant trajectory forecasting. Third, we present anovelmethodology for natural anisotropic continuous filters based onrelativelinear and angular positions ofneighboring objects in the canonicalized local coordinate frames.
Neural Information Processing Systems
Feb-8-2026, 04:05:35 GMT
- Country:
- North America > United States > Pennsylvania > Allegheny County > Pittsburgh (0.04)
- Technology: