基于多Agent技术的个性化随机慢化机制研究

Personalized Mechanism of Randomization Based on Multi-Agent Technology

摘要: 元胞个体的同质化与单一性，以及模型演化算法的简单性，导致传统元胞自动机(CA)模型忽略参与者个体差异，使用共性化随机慢化概率统一表述随机慢化行为，难以完整反映随机慢化行为的影响因素。本文通过多层次Agent模型中的理想模型和参与者特征模型对随机慢化行为影响因素进行分解，进而通过多Agent协同模型进行随机慢化行为影响因素集成，形成个性化随机慢化机制。引入该机制后，对于个体交通参与者，其随机慢化行为不再由简单、抽象的随机慢化概率涵盖，而是由其独有的个性化的随机慢化机制决定；而对于整个交通系统，由于参与者个体之间的差异性，导致不同参与者随机慢化的机制体现出个性化差异，系统对于随机慢化行为模拟的完整性得到有效增强。

Abstract: Due to the homogeneity of cellular and the simplicity of algorithm, traditional CA modeling use common randomization probability to describe behaviors of velocity randomization, which is difficult to reflect the factors on velocity randomization completely. In this paper, we decompose factors of velocity randomization in “ideal model” and “participant characteristic model”, integrate those factors and create the mechanism of personalized randomization in “Multi-Agent interaction model”. Integrity of the simulation on behaviors of velocity randomization is enhanced effectively by introducing the mechanism we create. For individual traffic participants, behaviors of velocity randomization are no longer explained by a simple, abstract randomization probability, but decided by its unique personalized velocity randomization mechanism. While for the entire transportation system, differences between the individual participants lead to different personalized velocity randomization mechanism.