Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Learning to Navigate in Complex Environments
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 |
|---|---|---|---|
| add depth prediction from the last LSTM layer (ablation) | Nav A3C · I-Maze | AUC: 143.3 -> 203.5 | Helped |
| add depth prediction from the last LSTM layer (ablation) | Nav A3C · Static 1 | AUC: 60.1 -> 104.3 | Helped |
| add depth prediction from the last LSTM layer (ablation) | Nav A3C · Static 2 | AUC: 59.9 -> 157.6 | Helped |
| add depth prediction from the last LSTM layer (ablation) | Nav A3C · Random Goal 1 | AUC: 45.5 -> 71.1 | Helped |
| add depth prediction from the last LSTM layer (ablation) | Nav A3C · Random Goal 2 | AUC: 37.0 -> 82.1 | Helped |
| combine depth prediction from convolutional and last LSTM layers instead of using last-LSTM depth prediction alone (ablation) | Nav A3C · Static 1 | AUC: 104.3 -> 100.3 | Hurt |
| combine depth prediction from convolutional and last LSTM layers instead of using last-LSTM depth prediction alone (ablation) | Nav A3C · Random Goal 2 | AUC: 82.1 -> 75.1 | Hurt |
| add reward prediction (ablation) | Nav A3C* · I-Maze | AUC: 143.3 -> 128.2 | Hurt |
| add reward prediction (ablation) | Nav A3C* · Static 1 | AUC: 60.1 -> 86.9 | Helped |
| add reward prediction alongside last-LSTM depth prediction (ablation) | Nav A3C+D2 · Random Goal 1 | AUC: 71.1 -> 72.3 | No clear effect |
| use depth regression rather than depth classification (ablation) | Nav A3C+D1 · multiple maze settings | task performance | Hurt |
| use RGBD as input rather than predicting depth as an auxiliary target (ablation) | Nav A3C · 3D maze environments | task performance | Hurt |
Closest papers by meaning
- Learning Exploration Policies for Navigation (ICLR 2019 · similarity 0.90)
- Reasoning in Visual Navigation of End-to-end Trained Agents: A Dynamical Systems Approach (CVPR 2025 · similarity 0.87)
- Exploration by Learning Diverse Skills through Successor State Representations (NeurIPS 2024 · similarity 0.86)
- GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- Efficient Exploration and Discriminative World Model Learning with an Object-Centric Abstraction (ICLR 2025 · similarity 0.86)
- Exploring the Edges of Latent State Clusters for Goal-Conditioned Reinforcement Learning (NeurIPS 2024 · similarity 0.86)
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.