Research map / Reinforcement learning / Visual and representation RL
Learning Exploration Policies for Navigation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use coverage gain as the intrinsic reward instead of curiosity prediction error (ablation) | Exploration policy · House3D unseen environments | coverage | Helped |
| Pre-train with imitation learning before RL (ablation) | Exploration policy · House3D unseen environments, no estimation noise | coverage | Helped |
| Pre-train with imitation learning (ablation) | Exploration policy · House3D unseen environments | coverage variance | Helped |
| Add RGB image observations (ablation) | Exploration policy · House3D unseen environments, no estimation noise | coverage | Helped |
| Add occupancy map inputs (ablation) | Exploration policy · House3D unseen environments, no estimation noise | coverage | Helped |
| Use map and RGB inputs rather than RGB alone (ablation) | IL + RL exploration policy · House3D unseen environments, no estimation noise | coverage | Helped |
| Use RGB observations and coverage reward rather than curiosity reward | IL + RL exploration policy · House3D unseen environments, no estimation noise | coverage | Helped |
| Use the full learning-based policy rather than frontier-based exploration | Exploration policy · House3D unseen environments with estimation noise | coverage | Helped |
| Use the full learning-based policy rather than frontier-based exploration | Exploration policy · House3D unseen environments without estimation noise | coverage | Hurt |
| Use learning-based exploration rather than frontier-based exploration | Exploration policy · House3D with geometry and affordance mismatch | coverage | Helped |
| Train the policy with 10% estimation noise in intrinsic reward and map construction rather than only using that noise at test time (ablation) | IL + RL with Map + RGB · House3D test houses | coverage area (m^2): 117.4m^2 -> 98.9m^2 | Hurt |
| Use exploration-policy experience to build a map before goal-driven navigation | Downstream navigation policy · House3D goal-image navigation | SPL | Helped |
Closest papers by meaning
- Learning to Navigate in Complex Environments (ICLR 2017 · similarity 0.90)
- Learning with a Mole: Transferable latent spatial representations for navigation without reconstruction (ICLR 2024 · similarity 0.88)
- Decoupling Exploration and Policy Optimization: Uncertainty Guided Tree Search for Hard Exploration (NeurIPS 2026 · similarity 0.87)
- Learning to Explore in POMDPs with Informational Rewards (ICML 2024 · similarity 0.87)
- Efficient Exploration and Discriminative World Model Learning with an Object-Centric Abstraction (ICLR 2025 · similarity 0.87)
- Reasoning in Visual Navigation of End-to-end Trained Agents: A Dynamical Systems Approach (CVPR 2025 · similarity 0.87)
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.