李雪原
发明人: 李雪原,高鑫,胡铭靖
申请人: 北京理工大学
申请号: 202410196496.8
申请日期: 2024.02.22
摘要: 本发明公开了基于多维加权图分层强化学习的自动驾驶多车决策方法,包括以下步骤:对交通场景进行图表征,对表征图进行时序的表示,获得表征图的节点特征矩阵;通过引入专家知识,根据节点特征矩阵中车辆的种类、横纵向相对位置、横纵向相对速度,获得横纵向维度图的子邻接矩阵;基于横纵向维度图的子邻接矩阵构建横向决策模 ...
发明人: 李雪原,高鑫,刘浩,胡铭靖
申请人: 北京理工大学
申请号: 202410196471.8
申请日期: 2024.02.22
摘要: 本发明公开了一种基于嵌套图强化学习的网联智能车辆编队决策方法,属于车联网和自动驾驶技术领域。方法包括:S1、采集编队内车辆状态信息,并对所述状态信息进行处理,得到编队间嵌套图以及车辆间嵌套图;S2、采用特征提取网络对所述编队间嵌套图以及所述车辆间嵌套图进行特征提取,基于提取的特征得到每辆智能车辆的动 ...
作者: Rahaman, Md. Faishal1; Li, Xueyuan1; Zakaria, Khan Md2; Al, Amin Md3; Buse, Kurt1; Bashar, S.M. Abul4
出处: 4th International Conference on Innovative Research in Applied Science, Engineering and Technology, IRASET 2024 Fez, Morocco 2024
作者: Wang, Kaifeng1; Liu, Qi1; Li, Xueyuan1; Yang, Fan1
出处: 35th IEEE Intelligent Vehicles Symposium, IV 2024 38 Sinhwayeoksa-ro 304 beon-gil, Andeok-myeon Seogwipo-si, Jeju Island, Korea, Republic of 2024
会议录: 337-344
作者: Liu, Qi1; Tang, Yujie2; Li, Xueyuan1; Yang, Fan1; Gao, Xin1; Li, Zirui3
出处: 35th IEEE Intelligent Vehicles Symposium, IV 2024 38 Sinhwayeoksa-ro 304 beon-gil, Andeok-myeon Seogwipo-si, Jeju Island, Korea, Republic of 2024
会议录: 376-383
作者: Yang, Fan1; Li, Xueyuan1; Liu, Qi1; Li, Xiangyu1; Li, Zirui1 (1School of Mechanical Engineering, Beijing Institute of Technology, Zhongguancun South Street, Beijing; 100081, China)
出处: Sensors 2024 Vol.24 No.8
作者: Gao, Xin1; Luan, Tian2; Li, Xueyuan1; Liu, Qi1; Ma, Zhaoyang3; Meng, Xiaoqiang1; Li, Zirui1 (1School of Mechanical Engineering, Beijing Institute of Technology, Beijing, China;2First Automobile Works Group Corp, Changchun, China;3School of Computer and Information Technology, Beijing Jiaotong University, Beijing, China)
出处: IEEE Internet of Things Journal 2024 P1-1
作者: Gao, Xin1; Ma, Zhaoyang2; Li, Xueyuan1; Meng, Xiaoqiang1; Li, Zirui1, 3 (1School of Mechanical Engineering, Beijing Institute of Technology, Beijing, China;2School of Computer and Information Technology, Beijing Jiaotong University, Beijing, China;3Chair of Traffic Process Automation, 'Friedrich-List' Faculty of Transport and Traffic Sciences, TU Dresden, Germany)
出处: arXiv 2024
作者: Gao, Xin1, 3; Li, Xueyuan1; Liu, Hao1; Li, Ao1; Ma, Zhaoyang2; Li, Zirui1, 4 (1The School of Mechanical Engineering, Beijing Institute of Technology, Beijing, China;2The School of Computer and Information Technology, Beijing Jiaotong University, Beijing, China;3The Center for AI Safety and Governance, Institute for AI, Peking University, Beijing; 100871, China;4The Chair of Traffic Process Automation, 'Friedrich-List' Faculty of Transport and Traffic Sciences, TU Dresden, Germany)
出处: arXiv 2024
发明人: 李雪原,高鑫,栾天,孟小强,刘琦
申请人: 北京理工大学
申请号: 202310874379.8
申请日期: 2023.07.17
摘要: 本发明公开了一种基于深度强化学习的伦理驱动多模态决策方法,包括:获取车辆拍摄图像,基于所述车辆拍摄图像和感知模型获取周围环境形态特征和动态特征;构建多模态神经网络,将人类伦理反馈的伦理系数引入至所述多模态神经网络中进行训练后对所述周围环境形态特征和动态特征进行计算,获得相应动作的Q值;基于强化学习算 ...