Riemannian Manifolds for Robot Learning and Optimization
Geometry-aware Bayesian Optimization (GaBO)Our idea is to leverage the prior geometric structure of the parameter space via geometry-aware kernels and Riemannian optimization, which often outperforms Euclidean approaches. For details:
Orientation Probabilistic Movement Primitives |
Learning Riemannian Manifolds for Robot Motion GenerationWe propose to see the problem of robot motion skills learning through the lens of Riemannian geometry. Our idea is to consider that a motion skill is described by a smooth surface containing the main motion patterns (i.e. demonstrated trajectories). For details:
Learning Riemannian Stable Dynamical Systems |
Geometry-aware Manipulability Tracking, Learning and ControlTo edit, click on the text to start adding your own words.
Hyperbolic Spaces for Robotic Taxonomies |