基于AlphaZero的地铁列车大量速度曲线自动生成算法

doi:10.11959/j.issn.2096-6652.202118

智能科学与技术学报 ›› 2021, Vol. 3 ›› Issue (2): 179-184.doi: 10.11959/j.issn.2096-6652.202118

• 专题：智能交通系统与应用 • 上一篇下一篇

基于AlphaZero的地铁列车大量速度曲线自动生成算法

芦玉琦¹, 陈德旺²^,³, 赵钊林²

¹ 福建工程学院计算机科学与数学学院，福建福州 350118
² 福建工程学院交通运输学院，福建福州 350118
³ 福州大学数学与计算机科学学院，福建福州 350108

修回日期:2021-05-13 出版日期:2021-06-15 发布日期:2021-06-01
作者简介:芦玉琦（1996- ），女，福建工程学院计算机科学与数学学院硕士生，主要研究方向为机器学习、列车智能驾驶
陈德旺（1976- ），男，福建省“闽江学者”特聘教授，福州大学数学与计算机科学学院教授、博士生导师, 福建工程学院交通运输学院特聘教授，主要研究方向为计算智能、机器学习、列车运行控制和智能交通系统
赵钊林（1973- ），女，福建工程学院交通运输学院副教授，主要研究方向为企业信息化、大数据分析及应用、数据建模技术等
基金资助:
国家自然科学基金资助项目(61976055);福建省财政厅教育科研专项资金资助项目(GY-Z21001)

Algorithm for automatically generating a large number of speed curves of subway trains based on AlphaZero

Yuqi LU¹, Dewang CHEN²^,³, Zhaolin ZHAO²

¹ School of Computer Science and Mathematics, Fujian University of Technology, Fuzhou 350118, China
² School of Transportation, Fujian University of Technology, Fuzhou 350118, China
³ College of Mathematics and Computer Science, Fuzhou University, Fuzhou 350108, China

Revised:2021-05-13 Online:2021-06-15 Published:2021-06-01
Supported by:
The National Natural Science Foundation of China(61976055);Special Fund of Education and Scientific Research of Fujian Provincial Finance Department(GY-Z21001)

摘要/Abstract

摘要：

在以往的地铁列车自动驾驶研究中，驾驶数据通常通过仿真生成单条运行曲线、对人工驾驶数据进行采样得到，不但实现方式比较复杂，而且效率较低、通用性不强。受AlphaZero系统启发，创新性地提出了人工生成虚拟地铁运行数据的思想。首先，根据一种五段式地铁列车速度曲线的运行方法，实现了虚拟数据的计算；然后，结合人类专家的经验设置牵引制动的区间分级、实际运行速度的分级、车站间距及变速距离分级等实际参数，缩小曲线数据范围，使其合理化；最后，通过Python编程得到大量数据，保存为数据集，绘制地铁列车运行时间频次分布图。通过观测发现，该虚拟数据覆盖各种运行时间，比传统数据更有利于地铁列车智能驾驶算法的研究。

关键词: 人机混合增强智能, 地铁列车速度曲线, AlphaZero思想, 虚拟数据生成

Abstract:

In the previous research on subway trains automatic driving, subway driving data is usually obtained through simulation to generate a single operating curve and manual driving data sampling.Not only is the implementation method more complicated, but also the efficiency is low and the versatility is not strong.Inspired by AlphaZero system, the idea of artificial generation of virtual metro operation data was put forward.Firstly, according to a running method of five section subway trains speed curve, the calculation of virtual data was realized.Then, combined with the experience of human experts, the actual parameters such as the classification of traction braking section, the classification of actual running speed, the classification of station spacing and variable speed distance were set to narrow the range of curve data and rationalize it.Finally, a large amount of data was obtained by Python programming, saved as a data set, and the frequency distribution map of subway trains operation time was drawn.It can be observed that the virtual data covers all kinds of operation time, which is more conducive to the research of subway trains intelligent driving algorithm than traditional data.

Key words: human-machine hybrid enhanced intelligence, subway trains speed curve, AlphaZero idea, virtual data gen-eration

中图分类号:

TP520.20，TP580.99

芦玉琦,陈德旺,赵钊林. 基于AlphaZero的地铁列车大量速度曲线自动生成算法[J]. 智能科学与技术学报, 2021, 3(2): 179-184.

Yuqi LU,Dewang CHEN,Zhaolin ZHAO. Algorithm for automatically generating a large number of speed curves of subway trains based on AlphaZero[J]. Chinese Journal of Intelligent Science and Technology, 2021, 3(2): 179-184.

图/表 4

参考文献 17

[1]	王勇博 . 采用遗传算法优化地铁多区间速度曲线和停站时间实现牵引节能的仿真研究[D]. 南京:南京理工大学, 2013.
	WANG Y B . Simulation research on traction energy saving by using genetic algorithm to optimize multi interval speed curve and stop time of Metro[D]. Nanjing:Nanjing University of Science and Technology, 2013.
[2]	WANG G Y , XIAO S , CHEN X ,et al. Application of genetic algorithm in automatic train operation[J]. Wireless Personal Communications, 2018,102(2): 1695-1704.
[3]	何之煜, 徐宁 . 非参数化迭代学习控制的列车自动驾驶控制算法[J]. 铁道学报, 2020,42(12): 90-96.
	HE Z Y , XU N . Automatic train operation control algorithm based on nonparametric iterative learning control[J]. Journal of the China Railway Society, 2020,42(12): 90-96.
[4]	GONG D D , LI G L . Research on multi-objective optimized target speed curve of subway operation based on ATO system[J]. International Core Journal of Engineering, 2020,6(2): 133-137.
[5]	杨杰, 吴佳焱, 王彪 ,等. 基于启发式遗传算法的列车节能运行目标速度曲线优化算法研究[J]. 铁道学报, 2019,41(8): 1-8.
	YANG J , WU J Y , WANG B ,et al. Optimization algorithm of target speed curve of train energy saving operation based on heuristic genetic algorithm[J]. Journal of the China Railway Society, 2019,41(8): 1-8.
[6]	唐振韬, 邵坤, 赵冬斌 ,等. 深度强化学习进展:从 AlphaGo 到AlphaGo Zero[J]. 控制理论与应用, 2017,34(12): 1529-1546.
	TANG Z T , SHAO K , ZHAO D B ,et al. Progress of deep reinforcement learning:from AlphagGo to AlphaGo Zero[J]. Control Theory and Application, 2017,34(12): 1529-1546.
[7]	程瑞军, 陈德旺, 田丽君 . 基于人机混合智能的地铁列车增强智能驾驶框架研究[C]// 2020 中国自动化大会(CAC2020)论文集. 北京:中国自动化学会, 2020.
	CHENG R J , CHEN D W , TIAN L J . Research on the framework of metro train enhanced intelligent driving based on man machine hybrid intelligence[C]// Proceedings of the China Society of Automation. Beijing:China Society of Automation, 2020.
[8]	唐川, 陶业荣, 麻曰亮 . AlphaZero原理与启示[J]. 航空兵器, 2020,27(3): 27-36.
	TANG C , TAO Y R , MA Y L . Principle and enlightenment of AlphaZero[J]. Aero Weaponry, 2020,27(3): 27-36.
[9]	郑南宁 . 人工智能新时代[J]. 智能科学与技术学报, 2019,1(1): 1-3.
	ZHENG N N . The new era of artificial intelligence[J]. Chinese Journal of Intelligent Science and Technology, 2019,1(1): 1-3.
[10]	冷勇林, 陈德旺, 阴佳腾 . 基于专家系统及在线调整的列车智能驾驶算法[J]. 铁道学报, 2014(2): 62-68.
	LENG Y L , CHEN D W , YIN J T . An intelligent train operation (ITO) algorithm based on expert system and online adjustment[J]. Journal of the China Railway Society, 2014(2): 62-68.
[11]	YIN J T , CHEN D W , LI L X . Intelligent train operation algorithms for subway by expert system and rein for cement learning[J]. IEEE Transaction son Intelligent Transportation Systems, 2014,15(6): 2561-2571.
[12]	ZHANG H R , JIA L M , WANG L . Energy consumption optimization of train operation for railway systems:algorithm development and real-world case Study[J]. Journal of Cleaner Production, 2019,214: 1024-1037.
[13]	关于福州市轨道交通 1 号线工程（一期）竣工环保验收调查报告的公示[Z]. 2019.
	Publicity of investigation report on environmental protection acceptance of Fuzhou rail transit line 1 project (phase I)[Z]. 2019.
[14]	关于福州市轨道交通2号线工程竣工环保验收报告的公示[Z]. 2021.
	Publicity of environmental protection acceptance report of Fuzhou rail transit line 2 project[Z]. 2021.
[15]	李卓玥 . 群智能算法在列车运行速度曲线节能优化中的研究[D]. 北京:北京交通大学, 2016.
	LI Z Y . Research on energy-saving optimization of train speed trajectories based on swarm intelligence algorithms[D]. Beijing:Beijing Jiaotong University, 2016.
[16]	ZHAO X H , KE B R , LIAN K L . Optimization of train speed curve for energy saving using efficient and accurate electric traction models on the mass rapid transit system[J]. IEEE Transactions on Transportation Electrification, 2018,4(4): 922-935.
[17]	KIM K M , SUK-MUM O ,, HAN M . A mathematical approach for reducing the maximum traction energy; the case of Korean MRT trains[C]// Proceedings of the International Multi Conference of Engineers and Computer Scientists.[S.l.:s.n.], 2010: 2169-2173.

基于AlphaZero的地铁列车大量速度曲线自动生成算法

Algorithm for automatically generating a large number of speed curves of subway trains based on AlphaZero

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 4

参考文献 17

相关文章 2

Metrics

推荐阅读 0

[1]	张俊, 许沛东, 陈思远, 高天露, 戴宇欣, 张科, 赵杭, 高杰迈, 白昱阳, 李金星, 张浩然, 李湘, 陈玖香. 物理-数据-知识混合驱动的人机混合增强智能系统管控方法[J]. 智能科学与技术学报, 2022, 4(4): 571-583.
[2]	吴宇震, 张俊, 高天露, 孙玉健, 刘金旭. 平行港口：智慧绿色时代下港口工业智联网新形态与体系结构[J]. 智能科学与技术学报, 2021, 3(2): 218-227.