地理学报 ›› 2011, Vol. 66 ›› Issue (6): 796-804.doi: 10.11821/xb201106008

• 商业地理 • 上一篇    下一篇

基于多代理人的零售业空间结构模拟

朱玮, 王德   

  1. 同济大学城市规划系, 上海 200092
  • 收稿日期:2010-10-26 修回日期:2011-05-03 出版日期:2011-06-20 发布日期:2011-06-20
  • 作者简介:朱玮(1978-), 通讯作者, 男, 博士, 讲师。主要研究方向为城市商业空间和消费者行为、城市规划模型方法。E-mail: weizhu@tongji.edu.cn
  • 基金资助:

    国家自然科学基金项目(40871080); 同济大学青年优秀人才培养行动计划(0100219087)

Multi-agent Based Simulation of Retail Spatial Structure

ZHUWei, WANG De   

  1. Department of Urban Planning, Tongji University, Shanghai 200092, China
  • Received:2010-10-26 Revised:2011-05-03 Online:2011-06-20 Published:2011-06-20
  • Supported by:

    National Natural Science Foundation of China, No.40871080; Tongji University Outstanding Young Talent Cultivation Action Plan

摘要: 中心地理论的提出由来已久,但对现实中多种因素复杂作用下的中心地体系进行实证研究并指导规划实践一直难以开展。传统模拟研究方法采用数学线性规划的思路,通过优化目标函数来实现对中心地体系的还原与考察,但复杂程度有限,难以用于实证研究和实践。多代理人技术的出现提供了新的研究思路。本文介绍笔者开发的一个零售业空间结构模拟系统。该系统以多代理人技术为基础,通过模拟商业中心和消费者两类代理人的行为,得出商业中心结构分布。商业中心被分为3 个等级,可以根据经营收入调整空间位置和自身服务等级;消费者以一定的频率产生不同等级的购物需求,并以距离为指标选择商业中心作为出行目的地。模拟过程从产生一定数量的一级商业中心开始,经过循环模拟消费者选择行为和商业中心的位置等级调整,最终达到商业中心等级和空间分布的平衡状态。作为对这个模拟系统的检验,分别对5 种典型的消费者空间分布场景进行模拟,包括:均匀分布、单中心、卫星城、分散组团、带形组团。结果表明这个系统的模拟性能稳定,生成的商业中心结构合理。

关键词: 零售业空间结构, 多代理人模拟, 商业中心, 消费者行为, 选择模型, 中心地理论

Abstract: Although central place theory has been proposed for a long time, empirical research and planning practices on central place systems in the real world under the complex multi-factor interactions are difficult to carry out. Conventional simulation-based research methods use linear programming technique to derive and investigate central place systems through optimizing certain objective functions. But their limitations on model complexity restrict their further application in empirical research and practice. Multi-agent technology provides a new perspective. This paper proposes a multi-agent retail spatial structure simulation system, which can derive structural distributions of shopping centers through simulating the behavior of shopping centers and consumers. The shopping centers in the simulation system have three levels and can adjust locations and service levels according to the revenues. Consumers generate different levels of shopping needs periodically and choose shopping centers based on travel distance. The simulation process starts with a number of shopping centers. After recursively simulating the behavior and interactions of centers and consumers, equilibrium is reached with the distributions of shopping center levels and locations emerged. As a validation of the system, simulations are carried out on five typical scenarios of consumer spatial distribution, including even distribution, single-center distribution, satellite distribution, grouped distribution, and belt distribution. Results show the system is stable with reasonable shopping center systems generated.

Key words: retail spatial structure, multi-agent simulation, shopping center, consumer behavior, choice model, central place theory