LoMAP projects each guided diffusion sample onto a PCA subspace of nearby offline trajectories, reducing infeasible plans and improving returns in Maze2D, MuJoCo locomotion, and AntMaze.
AntMaze-* State spaceS R29 Action spaceA R8 Episode length 1000 E.2
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Local Manifold Approximation and Projection for Manifold-Aware Diffusion Planning
LoMAP projects each guided diffusion sample onto a PCA subspace of nearby offline trajectories, reducing infeasible plans and improving returns in Maze2D, MuJoCo locomotion, and AntMaze.