Scholarlink

This paper explores the use of reinforcement learning (RL) to optimize urban layouts under the n-minute city model, a planning paradigm that aims to ensure that all essential services are accessible within a short walk distance. The paper presents two RL agents: Qtopia, a Q-learning agent that builds cities from scratch through episodic reward maximization, and city_planner, a Deep Q-Network (DQN) agent capable of classifying and redesigning preexisting city grids. Both agents aim to generate livable, modular and accessible urban environments by balancing proximity, diversity, and spatial compatibility. Both agents were tested and validated on artificial grids and a real-world layout of Rome, integrating natural constraints such as the Tiber River. A new metric for urban efficiency—residential service reachability rate—was introduced to measure how many residences are within the target service radius. The results demonstrate the potential of RL as a scalable tool for intelligent urbanism.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing Urban Layouts Using Reinforcement Learning (Q-Learning Agent and Deep Q Network Agents)

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Alan Crespo Murillo,
  • Alejandro Villafaña Anaya,
  • Bruno Galli Hambleton,
  • Daniel Adrián Contreras Olivas,
  • Humberto Ángel Plata Durán,
  • Oscar Cueto Farley Rivas,
  • Rodrigo Naranjo Hernández

摘要

This paper explores the use of reinforcement learning (RL) to optimize urban layouts under the n-minute city model, a planning paradigm that aims to ensure that all essential services are accessible within a short walk distance. The paper presents two RL agents: Qtopia, a Q-learning agent that builds cities from scratch through episodic reward maximization, and city_planner, a Deep Q-Network (DQN) agent capable of classifying and redesigning preexisting city grids. Both agents aim to generate livable, modular and accessible urban environments by balancing proximity, diversity, and spatial compatibility. Both agents were tested and validated on artificial grids and a real-world layout of Rome, integrating natural constraints such as the Tiber River. A new metric for urban efficiency—residential service reachability rate—was introduced to measure how many residences are within the target service radius. The results demonstrate the potential of RL as a scalable tool for intelligent urbanism.