Research map / Reinforcement learning / Visual and representation RL
Using Reinforcement Learning to Iteratively Construct Road Networks from Satellite Images and GPS Data
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes (without experiments) an RL agent that iteratively builds road networks from a combined tensor of satellite-image GTE encoding and a new GPS trajectory tensor encoding.
Problem
Manual construction and maintenance of road network maps is time-consuming, labor-intensive and costly; existing iterative graph-based construction relies on hand-designed algorithms.
Approach
Combine Sat2Graph's Graph-Tensor Encoding of satellite imagery with a new trajectory-preserving tensor encoding of GPS data, and train a reinforcement learning agent that iteratively adds vertices/edges, connects existing vertices, or marks vertices finished, replacing hand-designed expansion rules.
Key results
- No experimental results are reported; the paper is a proposal of a planned research topic.
Limitations
- No implementation or experiments; the expected benefits are unverified.
- Evaluation plan only names TOPO and APLS without reported numbers.
- Details of the RL agent architecture, reward, and training are not specified.
Experimental setup
Metrics: TOPO APLS
Baselines: RoadRunner Sat2Graph
Builds on: Sat2Graph RoadRunner
Closest papers by meaning
- Patched Line Segment Learning for Vector Road Mapping (AAAI 2024 · similarity 0.84)
- Road Network Representation Learning with the Third Law of Geography (NeurIPS 2024 · similarity 0.83)
- Efficient Reinforcement Learning by Discovering Neural Pathways (NeurIPS 2024 · similarity 0.83)
- Vision-Language Reasoning for Geolocalization: A Reinforcement Learning Approach (AAAI 2026 · similarity 0.82)
- Graph-Based Prediction and Planning Policy Network (GP3Net) for Scalable Self-Driving in Dynamic Environments Using Deep Reinforcement Learning (AAAI 2024 · similarity 0.82)
- I Open at the Close: A Deep Reinforcement Learning Evaluation of Open Streets Initiatives (AAAI 2024 · similarity 0.82)
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.