RLChina 前沿讲习班第1期——自动驾驶专题
## 导读 前沿讲习班是RLChina举办的学术活动,每期就某一研究话题邀请若干位专家作线上报告,并组织感兴趣的同学交流研讨。第1期前沿讲习班的话题是自动驾驶,将由清华大学李升波老师和华为诺亚罗军老师为大家分享最新的研究进展,欢迎大家参与。互动方式:**在本帖留言,可与报告嘉宾互动**。 ## 简介 ### 主题 RLChina 前沿讲习班第1期——自动驾驶专题 ### 时间 2021年10月12日 19:00至20:30 ### 网址 B站RLChina直播间 [http://live.bilibili.com/22386217](http://live.bilibili.com/22386217) ### 报告人 李升波 清华大学 ([课件](https://gitee.com/rlchina/rlchina-workshop/attach_files/855581/download),[回放](https://www.bilibili.com/video/BV1pf4y1g7jx?spm_id_from=333.999.0.0)) 罗军 华为诺亚方舟实验室 ([课件](https://gitee.com/rlchina/rlchina-workshop/attach_files/850546/download),[回放](https://www.bilibili.com/video/BV1fT4y1d7ie?spm_id_from=333.999.0.0)) ### 主持人 张海峰 中科院自动化所 ------------------------------------ ## 报告信息 ### 第一场 19:00-19:45 #### 报告人:李升波([课件](https://gitee.com/rlchina/rlchina-workshop/attach_files/855581/download),[回放](https://www.bilibili.com/video/BV1pf4y1g7jx?spm_id_from=333.999.0.0)) ![ ](https://rlchian-bbs.oss-cn-beijing.aliyuncs.com/images/2021/10/11/0285692470f70761907622a82c9b5d7a.jfif) #### 报告人简介 清华大学车辆学院副院长,长聘教授。留学于斯坦福大学,密歇根大学和加州伯克利大学。从事智能网联汽车、强化学习、最优控制与估计等研究。发表SCI论文>80篇,引用数超过9500次,入选ESI高引10篇(学科前1%),国内外学术会议论文奖11次。入选国家高层次科技创新人(2021)、科技部中青年科技创新领军人才(2020)、首届北京市基金委杰青(2018)、青年长江学者(2016)、国家基金委优青(2016)等。获中国汽车工业科技进步特等奖(2020)、国家科技进步二等奖(2018)、国家技术发明二等奖(2013)等。担任IEEE ITS学会的全球理事会委员、中国汽车工程学会青工委主任(首任)、IEEE Trans on ITS副主编、IEEE ITS Mag副主编、Automotive Innovation副主编等。 #### 报告标题 Apply Reinforcement Learning in Autonomous Vehicle Design #### 报告摘要 Unlike general intelligence for computer games, self-driving vehicles are faced with several problems such as high complexity of road structure, strong randomness of traffic conditions and participants, and hard safety constraints. Current mainstream decision and control methods either suffer high computing complexity or poor interpretability on real-world autonomous driving tasks. This report will focus on an interpretable and computationally efficient autonomous driving method on the basis of newly proposed integrated decision and control (IDC) framework, which decomposes a driving task into static path planning and dynamic optimal tracking that are structured hierarchically. The IDC framework can utilize an actor-critic RL algorithm to solve the constrained optimal control problem, in which its parametrized value and policy functions become path selector and path tracker, respectively. It also explores the demand of high-level intelligent vehicles for the next generation of artificial intelligence, and provide suggestions for core technology development and industrial implementation. #### 报告回放 <iframe src="//player.bilibili.com/player.html?aid=378608266&bvid=BV1pf4y1g7jx&cid=424513775&page=1" scrolling="no" border="0" frameborder="no" framespacing="0" allowfullscreen="true" width="600" height="450"></iframe> ------------------------------------ ### 第二场 19:45-20:30 #### 报告人:罗军 ([课件](https://gitee.com/rlchina/rlchina-workshop/attach_files/850546/download),[回放](https://www.bilibili.com/video/BV1fT4y1d7ie?spm_id_from=333.999.0.0)) ![ ](https://rlchian-bbs.oss-cn-beijing.aliyuncs.com/images/2021/10/11/016f7b65b91b9594887d981f102afe46.jfif?x-oss-process=image/resize,l_200) #### 报告人简介 Jun Luo studied computer science at Peking University and has a PhD in Computer Science and Cognitive Science from Indiana University Bloomington. He previously taught Cognitive Science at the University of Toronto and worked for several small and large companies. He joined Huawei Technologies Canada in 2016, where he currently serves as a Distinguished Researcher as part of Huawei Noah’s Ark Lab for Artificial Intelligence #### 报告标题 Why Autonomous Driving Needs RL and How to Use RL in Autonomous Driving #### 报告摘要 While reinforcement learning (RL) is yet to be taken up vigorously in today's real-world autonomous driving (AD) R&D, it is inevitable that RL will come to play a central role in AD. Without leveraging RL, we may never achieve AD solution that is generally usable for public urban environments. In this talk, we will explore the reasons why RL may be indispensable for AD and examine ways in which RL may be used in next-generation AD. #### 报告回放 <iframe src="//player.bilibili.com/player.html?aid=933500040&bvid=BV1fT4y1d7ie&cid=424514046&page=1" scrolling="no" border="0" frameborder="no" framespacing="0" allowfullscreen="true" width="600" height="450"> </iframe> ------------------------------------ ## 联系我们 Email: <rlchinacamp@163.com> ![Description](https://jidi-images.oss-cn-beijing.aliyuncs.com/rlchina2021/rlcn.jpeg?x-oss-process=image%2Fresize%2Cl_200) 来源:[https://mp.weixin.qq.com/s/fa7LDOo4Cz76I4qZjMLJWQ](https://mp.weixin.qq.com/s/fa7LDOo4Cz76I4qZjMLJWQ)
导读
前沿讲习班是RLChina举办的学术活动,每期就某一研究话题邀请若干位专家作线上报告,并组织感兴趣的同学交流研讨。第1期前沿讲习班的话题是自动驾驶,将由清华大学李升波老师和华为诺亚罗军老师为大家分享最新的研究进展,欢迎大家参与。互动方式:在本帖留言,可与报告嘉宾互动。
简介
主题
RLChina 前沿讲习班第1期——自动驾驶专题
时间
2021年10月12日 19:00至20:30
网址
B站RLChina直播间 http://live.bilibili.com/22386217
报告人
李升波 清华大学 (课件,回放)
罗军 华为诺亚方舟实验室 (课件,回放)
主持人
张海峰 中科院自动化所
报告信息
第一场 19:00-19:45
报告人:李升波(课件,回放)
报告人简介
清华大学车辆学院副院长,长聘教授。留学于斯坦福大学,密歇根大学和加州伯克利大学。从事智能网联汽车、强化学习、最优控制与估计等研究。发表SCI论文>80篇,引用数超过9500次,入选ESI高引10篇(学科前1%),国内外学术会议论文奖11次。入选国家高层次科技创新人(2021)、科技部中青年科技创新领军人才(2020)、首届北京市基金委杰青(2018)、青年长江学者(2016)、国家基金委优青(2016)等。获中国汽车工业科技进步特等奖(2020)、国家科技进步二等奖(2018)、国家技术发明二等奖(2013)等。担任IEEE ITS学会的全球理事会委员、中国汽车工程学会青工委主任(首任)、IEEE Trans on ITS副主编、IEEE ITS Mag副主编、Automotive Innovation副主编等。
报告标题
Apply Reinforcement Learning in Autonomous Vehicle Design
报告摘要
Unlike general intelligence for computer games, self-driving vehicles are faced with several problems such as high complexity of road structure, strong randomness of traffic conditions and participants, and hard safety constraints. Current mainstream decision and control methods either suffer high computing complexity or poor interpretability on real-world autonomous driving tasks. This report will focus on an interpretable and computationally efficient autonomous driving method on the basis of newly proposed integrated decision and control (IDC) framework, which decomposes a driving task into static path planning and dynamic optimal tracking that are structured hierarchically. The IDC framework can utilize an actor-critic RL algorithm to solve the constrained optimal control problem, in which its parametrized value and policy functions become path selector and path tracker, respectively. It also explores the demand of high-level intelligent vehicles for the next generation of artificial intelligence, and provide suggestions for core technology development and industrial implementation.
报告回放
第二场 19:45-20:30
报告人:罗军 (课件,回放)
报告人简介
Jun Luo studied computer science at Peking University and has a PhD in Computer Science and Cognitive Science from Indiana University Bloomington. He previously taught Cognitive Science at the University of Toronto and worked for several small and large companies. He joined Huawei Technologies Canada in 2016, where he currently serves as a Distinguished Researcher as part of Huawei Noah’s Ark Lab for Artificial Intelligence
报告标题
Why Autonomous Driving Needs RL and How to Use RL in Autonomous Driving
报告摘要
While reinforcement learning (RL) is yet to be taken up vigorously in today’s real-world autonomous driving (AD) R&D, it is inevitable that RL will come to play a central role in AD. Without leveraging RL, we may never achieve AD solution that is generally usable for public urban environments. In this talk, we will explore the reasons why RL may be indispensable for AD and examine ways in which RL may be used in next-generation AD.
报告回放
联系我们
Email: rlchinacamp@163.com
导读
前沿讲习班是RLChina举办的学术活动,每期就某一研究话题邀请若干位专家作线上报告,并组织感兴趣的同学交流研讨。第1期前沿讲习班的话题是自动驾驶,将由清华大学李升波老师和华为诺亚罗军老师为大家分享最新的研究进展,欢迎大家参与。互动方式:在本帖留言,可与报告嘉宾互动。
简介
主题
RLChina 前沿讲习班第1期——自动驾驶专题
时间
2021年10月12日 19:00至20:30
网址
B站RLChina直播间 http://live.bilibili.com/22386217
报告人
李升波 清华大学 (课件,回放)
罗军 华为诺亚方舟实验室 (课件,回放)
主持人
张海峰 中科院自动化所
报告信息
第一场 19:00-19:45
报告人:李升波(课件,回放)
报告人简介
清华大学车辆学院副院长,长聘教授。留学于斯坦福大学,密歇根大学和加州伯克利大学。从事智能网联汽车、强化学习、最优控制与估计等研究。发表SCI论文>80篇,引用数超过9500次,入选ESI高引10篇(学科前1%),国内外学术会议论文奖11次。入选国家高层次科技创新人(2021)、科技部中青年科技创新领军人才(2020)、首届北京市基金委杰青(2018)、青年长江学者(2016)、国家基金委优青(2016)等。获中国汽车工业科技进步特等奖(2020)、国家科技进步二等奖(2018)、国家技术发明二等奖(2013)等。担任IEEE ITS学会的全球理事会委员、中国汽车工程学会青工委主任(首任)、IEEE Trans on ITS副主编、IEEE ITS Mag副主编、Automotive Innovation副主编等。
报告标题
Apply Reinforcement Learning in Autonomous Vehicle Design
报告摘要
Unlike general intelligence for computer games, self-driving vehicles are faced with several problems such as high complexity of road structure, strong randomness of traffic conditions and participants, and hard safety constraints. Current mainstream decision and control methods either suffer high computing complexity or poor interpretability on real-world autonomous driving tasks. This report will focus on an interpretable and computationally efficient autonomous driving method on the basis of newly proposed integrated decision and control (IDC) framework, which decomposes a driving task into static path planning and dynamic optimal tracking that are structured hierarchically. The IDC framework can utilize an actor-critic RL algorithm to solve the constrained optimal control problem, in which its parametrized value and policy functions become path selector and path tracker, respectively. It also explores the demand of high-level intelligent vehicles for the next generation of artificial intelligence, and provide suggestions for core technology development and industrial implementation.
报告回放
第二场 19:45-20:30
报告人:罗军 (课件,回放)
报告人简介
Jun Luo studied computer science at Peking University and has a PhD in Computer Science and Cognitive Science from Indiana University Bloomington. He previously taught Cognitive Science at the University of Toronto and worked for several small and large companies. He joined Huawei Technologies Canada in 2016, where he currently serves as a Distinguished Researcher as part of Huawei Noah’s Ark Lab for Artificial Intelligence
报告标题
Why Autonomous Driving Needs RL and How to Use RL in Autonomous Driving
报告摘要
While reinforcement learning (RL) is yet to be taken up vigorously in today’s real-world autonomous driving (AD) R&D, it is inevitable that RL will come to play a central role in AD. Without leveraging RL, we may never achieve AD solution that is generally usable for public urban environments. In this talk, we will explore the reasons why RL may be indispensable for AD and examine ways in which RL may be used in next-generation AD.
报告回放
联系我们
Email: rlchinacamp@163.com