Evolving Adaptive Traffic Signal Controllers for a Real Scenario Using Genetic Programming with an Epigenetic Mechanism

2017 16th IEEE International Conference on Machine Learning and Applications (ICMLA) Pub Date : 2017-12-01 DOI:10.1109/ICMLA.2017.00-41

Esteban Ricalde, W. Banzhaf

引用次数: 7

Abstract

An important challenge for traffic signal control is adapting to irregular changes in traffic. In recent years, different heuristics have been developed to address this issue. However, most of them are tested in artificial scenarios under controlled circumstances. In this paper, we present the first implementation of Genetic Programming in the evolution of traffic signal controllers for a real-world scenario. The evolved controllers are compared with a static control and an actuated control. The results indicate a significant improvement over traditional methods. Moreover, additional experiments indicate that the evolved controllers have the ability to adapt to unplanned changes in traffic conditions.

查看原文

微信好友朋友圈 QQ好友复制链接

本刊更多论文

基于遗传规划和表观遗传机制的自适应交通信号控制器

适应交通的不规律变化是交通信号控制面临的一个重要挑战。近年来，人们开发了不同的启发式方法来解决这个问题。然而，它们中的大多数都是在受控环境下的人工场景中进行测试的。在本文中，我们提出了遗传规划在交通信号控制器进化中的第一个实现。将改进的控制器与静态控制和驱动控制进行了比较。结果表明，与传统方法相比，该方法有了显著的改进。另外，实验表明，改进后的控制器具有适应交通状况意外变化的能力。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文去求助

来源期刊

2017 16th IEEE International Conference on Machine Learning and Applications (ICMLA)

自引率

0.00%

发文量

期刊最新文献

Tree-Structured Curriculum Learning Based on Semantic Similarity of Text Direct Multiclass Boosting Using Base Classifiers' Posterior Probabilities Estimates Predicting Psychosis Using the Experience Sampling Method with Mobile Apps Human Action Recognition from Body-Part Directional Velocity Using Hidden Markov Models Realistic Traffic Generation for Web Robots