基于强化学习和城市感知的碳排评价方法——以都江堰为例
A Carbon Emission Evaluation Method Based on Reinforcement Learning and Urban Perception——Taking Dujiangyan as an Example
张雨舟 1邓颖睿2
作者信息
- 1. 成都市建筑设计研究院有限公司
- 2. 四川大学计算机学院
- 折叠
摘要
文章以碳排评价方法为研究对象,以都江堰城市街景图为实证落脚点,鉴于人们对城市碳排水平的感知水平主要来源于天空、建筑、车辆、区域位置等,通过问卷调查和运用城市感知获取部分数据,应用强化学习的方法构建城市碳评新模型.区别于传统研究方式,本文提出基于弱监督学习的方法,结合深度强化学习与按照马尔科夫决策过程特点,模拟人类水平的感知能力,兼顾反映主观性与客观性,更为准确的评价居民对碳排放感知,使CIM建设、城市更新、城市管理工作更具有针对性,进而提高人民满意度.
Abstract
This article focuses on the research of carbon emission evaluation methods,with the city street view of Dujiangyan as the empirical basis.Considering that people's perception of urban carbon emissions mainly comes from the sky,buildings,vehicles,and regional locations,this study utilizes questionnaire surveys and urban perception to obtain partial data,and applies the method of reinforcement learning to construct a new model for urban carbon evaluation.Different from traditional research methods,this article proposes a weakly supervised learning method,combining deep reinforcement learning with the characteristics of Markov decision processes to simulate human-level perception abilities,and to comprehensively reflect subjectivity and objectivity.This approach provides a more accurate evaluation of residents'perception of carbon emissions,making the construction of CIM(Carbon Emission Evaluation Method),urban renewal,and urban management work more targeted,thereby improving people's satisfaction.
关键词
城市感知/强化学习/碳排放/计算机视觉Key words
urban perception/reinforcement learning/carbon emissions/computer vision引用本文复制引用
出版年
2024
NETL
NSTL
万方数据