AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

Robust Inverse Reinforcement Learning under Transition Dynamics Mismatch

Luca Viano, Yu-Ting Huang, Parameswaran Kamalaruban, Adrian Weller, Volkan Cevher

NeurIPS 2021 · Reinforcement learning › Imitation learning · approach: Reward learning from demonstrations · 3 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
add robust MCE IRL (two-player Markov game, alpha<1) instead of standard MCE IRLMCE IRL · GridWorld / ObjectWorld under expert-learner dynamics mismatchlearner return V under learner MDPImproved
add robust training (alpha<1) in the low expert-noise regime (little mismatch)MCE IRL · GridWorld / ObjectWorld, low expert noiselearner return V under learner MDPHurt
underestimate opponent strength 1-alpha relative to true mismatch (ablation)Robust MCE IRL · GridWorld / ObjectWorldlearner return VHurt
overestimate opponent strength 1-alpha relative to true mismatch (ablation)Robust MCE IRL · GridWorld / ObjectWorldlearner return VHurt
set 1-alpha approx d_dyn(T^L, estimated T^E)/2 using a rough estimate of expert dynamics (ablation)Robust MCE IRL · GridWorld / ObjectWorldlearner return VImproved
add robust RE IRL (robust variant of Relative Entropy IRL) in continuous MDPRE IRL · GaussianGridlearner returnImproved
add robust MCE IRL with low-dimensional feature mapMCE IRL · low-dimensional feature GridWorld (Appendix)learner returnImproved
choose alpha = 1 - d_dyn/2 in constructive exampleRobust MCE IRL · three-state constructive MDP (theory)performance gap to no-mismatch MCE IRL: gamma/(1-gamma) * d_dyn -> 0Improved

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