查看原文
其他

IJCAI 2019 | 4小时244页《AI in Transportation》教程带你系统梳理人工智能在出行领域的应用

李群Tracy 滴滴科技合作 2021-09-05

点击上方“蓝色字体”,选择“置顶公众号”

精彩内容,即刻送达

编者按

2019年8月10-16日,人工智能领域国际顶级会议IJCAI在中国澳门盛大启幕。10号,滴滴AI Labs首席算法工程师王征博士、强化学习团队负责人秦志伟(Tony)博士带来4小时《AI in Transportation》教程,带你系统梳理人工智能在出行领域的应用,深度强化学习在滴滴的探索及进展,并详细整理了可用的数据资源和工具包。



教程内容可在https://outreach.didichuxing.com/IJCAI2019/tutorial/ 下载。


【讲者简介】


01

王征

滴滴AI Labs首席算法工程师



Dr. Zheng Wang is a principal researcher in DiDi AI Labs. He received his Ph.D. degree from Tsinghua University in 2011 and worked as a research fellow in Arizona State University in 2011-2014, then as a research faculty in the University of Michigan at Ann Arbor in 2014-2016. He has received several awards, including best research paper award runner-up in KDD and best paper award in IEEE International Conference in Social Computing (SocialCom). He served as the PC member of leading conferences, such as ICML, NIPS, KDD and IJCAI, and gave tutorial in KDD and ICDM. He is now leading the R&D team, working on designing and developing novel machine learning systems and services for DiDi map, DiDi safe driving and DiDi capacity prediction platform. He designed the novel machine learning and deep learning solutions of DiDi ETA and route planning services.


02

秦志伟(Tony)

滴滴AI Labs强化学习团队负责人



Dr. Zhiwei (Tony) Qin leads the reinforcement learning research at DiDi AI Labs, working on core problems in ride-sharing marketplace optimization. He received his Ph.D. in Operations Research from Columbia University and B.Sc. in Computer Science and Statistics from the University of British Columbia, Vancouver. Tony is broadly interested in research topics at the intersection of optimization and machine learning, and most recently in reinforcement learning and its applications in operational optimization, digital marketing, traffic signals control, and education. He has published in top-tier conferences and journals in machine learning and optimization, including ICML, KDD, IEEE ICDM, WWW, JMLR, and MPC. He has served as Senior PC/PC of NeurIPS, AAAI, IJCAI, KDD, JMLR, TPAMI, and other operations research journals.


【教程摘要】


现代交通,从马车到第一条道路的出现,从自行车到汽车,再到高速公路,以及现今智慧交通的演化经历了数千年的时间。纵观古今,每一次技术革命都会给人类的出行方式带来质的飞跃。人工智能时代,给共享出行提供了一个前所未有的机遇,AI引领交通变革,势不可挡。本教程中,两位讲者系统地从地图、供需预测、派单、数据集及工具等方面梳理了AI在出行领域的应用。


王征正在讲解ETA


其中,王征分享了AI在地图匹配、路径规划、预估到达时间、交通预测方面取得的进展。报告中,王征还展示了滴滴在AI for Social Good(AI赋能社会)上的最新进展——《司机关怀助手》,能够细心聆听每一次对话,智能区分谈话内容领域,提供精准回应,理解用户个行人语言习惯,智能提取关键信息,提供全方位的智能服务。该项工作也被IJCAI 2019 Video Track收录。此前一年,滴滴成立AI for Social Good共创平台,与十多所高校、科研机构和社会组织展开合作,在安全、健康、环境、无障碍等几大核心方向进行合作项目研究。


在介绍相关数据集时,王征还提到,“滴滴盖亚数据开放计划”自2017年10月正式启动以来,已陆续发布了包括轨迹数据、POI检索数据、驾驶场景视频数据、出行数据等6个数据集。通过开放脱敏数据集,滴滴期待着和学术界一起探索,赋能出行!


秦志伟在分享多智能体强化学习在派单上的应用


秦志伟则深入地介绍了强化学习及其在派单等决策系统(如派单)上的探索。IJCAI期间,秦志伟团队的工作「在线订单分配和司机调度的深度强化学习模型」被大会Demo Track收录,通过报告和现场交互展示,得到了与会专家学者的高度评价,深度强化学习在滴滴的应用取得的成果也引起了广泛的关注。在时空维度不断变化的供需情况中的订单分配和司机调度是网约车交易市场中的重要问题。手动的heuristic方法很难考虑这类资源分配问题里的动态性。端到端的机器学习方法则是潜在更好的方法。之前的工作都采用了分别解决派单和调度问题的思路,并且司机是作为独立智能体来最大化各自的奖励。在这篇文章中,我们呈现一个深度强化学习的方法来解决派单和调度的联动问题。在把司机作为独立智能体之上,我们更从系统角度,来提供一个针对所有司机的全局派单调度的方法。该项工作已被ICDM 2019收录。



教程大纲:



扎根出行领域,滴滴也希望能够利用人工智能技术让出行更美好。


滴滴@IJCAI系列解读将陆续推出,欢迎关注!


点击阅读原文获取教程全文!


相关阅读

DiDi@IJCAI 2019 | 相约50周年盛会,澳门唱响AI之歌(预告篇)

DiDi@IJCAI 2019 | 相约50周年盛会,澳门唱响AI之歌(论文篇)


编辑 | 洛羽

视频 小程序 ,轻点两下取消赞 在看 ,轻点两下取消在看

您可能也对以下帖子感兴趣

文章有问题?点此查看未经处理的缓存