AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Learning Dynamics Feature Representation via Policy Attention for Dynamic Path Planning in Urban Road Networks

Kai Zhang, Jingjing Gu, Qiuhong Wang

ICLR 2026 · Reinforcement learning › Visual and representation RL · approach: RL for combinatorial routing

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 DFR (policy attention plus n-hop neighborhoods) instead of all-dynamics input (ablation)DQN · Nanjing Subgraph 1Mean GAP: 0.170 -> 0.095Improved
add DFR instead of all-dynamics input (ablation)DQN · Nanjing Subgraph 1Success rate: 0.884 -> 0.908Improved
increase n-hop neighborhood from 1 to 2 (k=0.6) (ablation)DQN+DFR · Nanjing Subgraph 1Success rate: 0.723 -> 0.867Improved
increase n from 2 to 4 (k=0.6) (ablation)DQN+DFR · Nanjing Subgraph 1Success rate: 0.867 -> 0.892No clear effect
increase policy-attention proportion k from 0.4 to 0.6 (n=4) (ablation)DQN+DFR · Nanjing Subgraph 1Mean GAP: 0.095 -> 0.113Hurt
add DFR to RL path planners for planning speedDQN, GCN+DQN, PPO · 3 urban road networksplanning time: -85.59%, -46.08%, -79.32%Improved

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