AI Paper Atlas

Research map / Reinforcement learning / MDPs and planning

Qualitative Analysis of ω-Regular Objectives on Robust MDPs

Ali Asadi, Krishnendu Chatterjee, Ehsan Kafshdar Goharshady, Mehrdad Karrabi, Ali Shafiee

AAAI 2026 · Reinforcement learning › MDPs and planning · approach: Robust MDP policy gradient

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 oracle-based ASReach instead of long-run-average policy iteration RPPI for almost-sure reachabilityrobust MDP reachability · Frozen Lake and QComp, L1/L2/Linf uncertaintyinstances solved within 60 s: 81 -> 426Improved
use ASReach instead of RPPI on instances both solverobust MDP reachability · instances solved by bothaverage time (s): 6.3 -> 0.1Improved
ASReach applied to L2 uncertainty sets, which RPPI does not supportASReach · Frozen Lake, L2, R_max=1instances solved within 60 s: 24 of 24 solvedImproved
use ASParity instead of EffASParity despite worse worst-case complexity (ablation)parity objectives on Frozen Lake · Frozen Lake parity benchmarksinstances solved within 60 s: 120 -> 135Improved
use ASParity instead of EffASParity (ablation)parity objectives · Frozen Lake, L1 uncertainty, R_max=0.5average time (s): 7.8 -> 7.3Improved
increase uncertainty radius R_max from 0.5 to 1.5 for EffASParity on L2 Frozen Lake (ablation)EffASParity · Frozen Lake parity, L2instances solved within 60 s: 12 -> 15Improved

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