CubeDAgger upgrades EnsembleDAgger with threshold regularization, optimal consensus switching, and colored noise injection to enable stable interactive imitation learning in dynamic systems, validated in simulation and real-robot scooping with 30 minutes of expert interaction.
An enhancement of the bisection method average performance preserving minmax optimality,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
CubeDAgger: Interactive Imitation Learning for Dynamic Systems with Efficient yet Low-risk Interaction
CubeDAgger upgrades EnsembleDAgger with threshold regularization, optimal consensus switching, and colored noise injection to enable stable interactive imitation learning in dynamic systems, validated in simulation and real-robot scooping with 30 minutes of expert interaction.