AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

An Imitation from Observation Approach to Transfer Learning with Dynamics Mismatch

Siddharth Desai, Ishan Durugkar, Haresh Karnan, Garrett Warnell, Josiah Hanna, Peter Stone

NeurIPS 2020 · Reinforcement learning › Imitation learning · approach: Imitation combined with RL · 20 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use adversarial imitation-from-observation action transformation instead of GATTransfer from source to target with dynamics mismatch · Minitaur target environmenttarget performance relative to optimal: more than 80% vs at most 50%Improved
Use GARAT rather than GAT for source-environment groundingInvertedPendulumHeavy · InvertedPendulum domainper-step transition errorImproved
Use GARAT rather than GAT for transferTransfer methods under dynamics mismatch · MuJoCo domains except HopperHighFrictiontarget-environment returnImproved
Use GARAT rather than related transfer methodsTransfer methods under dynamics mismatch · HopperHighFrictiontarget-environment returnNo clear effect
Use GARAT with a few thousand target transitions rather than training directly in the target environmentPolicy transfer · Most evaluated domainstarget-environment return: 1 million transitionsNo clear effect
Use GARAT rather than GAT for grounding with limited target dataAction transformation · InvertedPendulum domainper-step transition error: a single trajectory vs 100 trajectoriesImproved
Use GARAT rather than related methods for transferPolicy transfer · Ant in PyBullettarget-environment returnImproved
Use GARAT rather than related methods for transferPolicy transfer · Walkertarget-environment returnImproved
Use alternate imitation-from-observation backend algorithms instead of the selected backend (ablation)GARAT · Sim-to-real performancetarget-environment returnNo clear effect
Use other divergence measures instead of the losses detailed in the algorithm (ablation)GARAT · Transfer experimentstarget-environment returnNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy