居住外迁个体通勤活动链中的多维出行决策机理

Residential relocation as consequence of city reform and construction during urbanization, is responsible for some commuting travel problems. By looking at that in a more microscopic way, effect of residential relocation on commuting travel can be captured by studying relocated commuter's activity-travel behavior. This research follows a technical guideline of "Collection of multi-source data - Identification of influencing factors - Analysis of travel mechanism - Verification of mixed simulation". It aims to better capture the complex evolution of multi-dimensional commuting chain features including travel mode, time, destination and path, which is caused by residential relocation. More specifically, based on data collected by carefully organized survey about residential relocation and individuals' activity-travel behaviors, influencing factors of multi-dimensional decisions in relocated commuters' activity chains will be targeted. Analysis models about mechanism of making those decisions will be established using methods like hierarchical bayesian networks and agent-based reinforcement learning. Then, using Multi-Agent and ArcGIS as a mixed simulated method and based on case study of typical city, evolution of relocated individual's commuting behavior as well as some dynamic features of macroscopic travel mode structure and its space-time distribution can be predicted. The expected research results can help to provide some theoretic basis as well as analyzing method to forecast potential commuting travel problems cause by residential relocation in a microscopic way.

立足当前城镇化进程中老城改造和新区建设并举背景下居住外迁所衍生的通勤交通出行问题，从微观个体活动-出行行为源头探索居住外迁对通勤交通出行的影响规律。针对居住外迁影响下个体通勤活动链中方式、时间、起讫点、路径多维出行特征演变的复杂性，采用"多源数据采集-影响因素识别-行为机理解析-混合仿真实证"的总体技术路线展开研究。基于个体居住外迁及活动-出行多源数据的同步调查分析深入挖掘居住外迁个体通勤链中的多维出行决策影响因素，扩展运用多层贝叶斯网络建模和智能体强化学习方法构建居住外迁个体通勤链中的多维出行决策机理解析模型。结合典型城市案例，运用Multi-Agent 和ArcGIS 混合仿真预测居住外迁个体微观通勤行为演变及宏观通勤交通结构与时空分布的动态变化特征。预期研究成果可为从微观尺度预见居住外迁引发的潜在通勤交通问题提供科学的理论依据和分析方法。

城市居住外迁引发城市通勤交通问题的同时，也给城市迁居个体出行带来一系列微观变化，导致城市交通需求时空的重分布。居住外迁个体通勤方式、时间、起讫点、路径等多方面出行特征相互关联、相互制约，呈现出复杂的多维决策行为。已有研究虽已开始关注居住外迁引发的通勤交通问题，但居住外迁衍生交通出行演变规律的理论方法和分析手段一直缺乏科学的系统研究，难以提前预测感知居住外迁可能引发的潜在交通矛盾。因此，本项目围绕居住外迁衍生交通出行演变规律，以预防和控制该交通问题的二次演化和蔓延为研究目的，采用“多源数据采集-影响因素识别-行为机理解析-混合仿真实证”总体技术路线，针对居住外迁个体通勤活动-出行多源数据获取与出行特征分析方法、居住外迁个体通勤链的多维出行决策影响因素、多维出行决策模型以及居住外迁下的交通特征演变仿真展开研究。研究结合个体居住外迁及活动-出行数据调查设计和API、POI网络数据，以南京城市内部迁居人群为例，建立多源迁居人群出行数据库，基于Sequential Mixed Logit、Cox、贝叶斯网络以及Rasch模型，模拟南京内部迁居个体通勤者出行行时间、空间分布、方式结构以及全过程出行方式链在迁居前后的微观变化特征，结果发现迁居前后社区建成环境、公交服务及路网可达性、临近轨道接驳设施服务、居住区位以及家庭基础经济条件的变化，均是影响居住外迁个体通勤多维出行决策变化的重要因素。同时，项目扩展运用神经网络和多智能体强化学习方法，联合构建居住外迁个体通勤链中的多维出行决策机理解析模型，混合仿真和预测居住外迁个体微观通勤行为演变及宏观通勤交通结构与时空分布的动态变化特征。研究成果可为从微观尺度预见居住外迁引发的潜在通勤交通问题提供科学的理论依据和分析方法。