DIVER: Reinforced Diffusion Breaks Imitation Bottlenecks in End-to-End Autonomous Driving
DIVER:强化扩散突破端到端自动驾驶的模仿瓶颈
机构 * School of Artificial Intelligence (School of Software), Yanshan University(燕山大学人工智能学院(软件学院)) ; Beijing Key Laboratory of Traffic Data Mining and Embodied Intelligence, School of Computer Science and Technology, Beijing Jiaotong University(北京交通大数据挖掘与具身智能关键实验室,北京交通大学计算机科学与技术学院) ; Horizon Robotics(地平线机器人) ; School of Mechanical and Aerospace Engineering, Nanyang Technological University(南洋理工大学机械与航空航天工程学院) ; University of Macau(澳门大学) ; School of Electrical Engineering and Computer Science, The University of Queensland(昆士兰大学电子工程与计算机科学学院)
专题命中 端到端驾驶 :autonomous driving(title,abstract);end-to-end driving(abstract);分类 cs.RO、cs.CV
AI总结 DIVER通过融合强化学习与扩散生成,生成多样化可行轨迹,提升自动驾驶的泛化能力,解决模仿学习中的模式崩溃问题。
Comments 17 pages, 10 figures