Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Learning to Navigate in Cities Without a Map
Piotr Mirowski, Matt Grimes, Mateusz Malinowski, Karl Moritz Hermann, Keith Anderson, Denis Teplyashin, Karen Simonyan, koray kavukcuoglu, Andrew Zisserman, Raia Hadsell
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use a dual-LSTM architecture with a locale-specific goal pathway | GoalNav on the NYU courier task · NYU courier task | average reward per episode | Improved |
| add heading prediction to GoalNav (ablation) | GoalNav · NYU courier task | average reward per episode | Improved |
| add a skip connection from the vision layers to the policy LSTM (ablation) | CityNav · NYU courier task | average reward per episode | Improved |
| remove the goal input (ablation) | CityNav · NYU courier task | average reward per episode | Hurt |
| add heading prediction to two-LSTM CityNav without a skip connection (ablation) | CityNav · NYU courier task | average reward per episode | No clear effect |
| use CityNav instead of GoalNav | NYU courier task · NYU courier task | average reward per episode | No clear effect |
| add random coin rewards (ablation) | CityNav reward training · NYU courier task | goal reward | No clear effect |
| use larger-radius early rewards (ablation) | CityNav reward training · NYU courier task | goal reward | Improved |
| reduce landmark density (ablation) | CityNav goal representation · NYU courier task | average reward per episode | Hurt |
| replace landmark goals with normalized latitude and longitude scalar coordinates (ablation) | CityNav · NYU courier task | average reward per episode | Improved |
| remove the skip connection in multi-city pre-training and transfer (ablation) | MultiCityNav · Wall Street transfer | goal reward | Improved |
| pre-train on more regions before transferring to the target city | MultiCityNav · Wall Street transfer | goal reward | Improved |
Closest papers by meaning
- Learning to Follow Directions in Street View (AAAI 2020 · similarity 0.87)
- CityWalker: Learning Embodied Urban Navigation from Web-Scale Videos (CVPR 2025 · similarity 0.85)
- CitySeeker: How Do VLMs Explore Embodied Urban Navigation with Implicit Human Needs? (ICLR 2026 · similarity 0.84)
- An Interactive Navigation Method with Effect-oriented Affordance (CVPR 2024 · similarity 0.83)
- Learning with a Mole: Transferable latent spatial representations for navigation without reconstruction (ICLR 2024 · similarity 0.83)
- NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models (AAAI 2024 · similarity 0.83)
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.
