Autonomous driving has received a lot of attention in the automotive industry and is often seen as the future of transportation. Passenger vehicles equipped with a wide array of sensors (e.g., cameras, front-facing radars, LiDARs, and IMUs) capable of continuous perception of the environment are becoming increasingly prevalent. These sensors provide a stream of high-dimensional, temporally correlated data that is essential for reliable autonomous driving. An autonomous driving system should effectively use the information collected from the various sensors in order to form an abstract description of the world and maintain situational awareness. Deep learning models, such as autoencoders, can be used for that purpose, as they can learn compact latent representations from a stream of incoming data. However, most autoencoder models process the data independently, without assuming any temporal interdependencies. Thus, there is a need for deep learning models that explicitly consider the temporal dependence of the data in their architecture. This work proposes CARNet, a Combined dynAmic autoencodeR NETwork architecture that utilizes an autoencoder combined with a recurrent neural network to learn the current latent representation and, in addition, also predict future latent representations in the context of autonomous driving. We demonstrate the efficacy of the proposed model in both imitation and reinforcement learning settings using both simulated and real datasets. Our results show that the proposed model outperforms the baseline state-of-the-art model, while having significantly fewer trainable parameters.
From Imitation to Exploration: End-to-end Autonomous Driving based on World Model
Department of Automation, Shanghai Jiao Tong University(自动化系,上海交通大学);Key Laboratory of System Control and Information Processing, Ministry of Education of China(系统控制与信息处理重点实验室,中华人民共和国教育部);Global Institute of Future Technology, Shanghai Jiao Tong University(未来技术全球研究院,上海交通大学)
WALT: Learning World-Model-Aligned Latent Trajectories for Autonomous Driving
The Hong Kong University of Science and Technology(香港科技大学);Horizon Robotics(地平线机器人);The Chinese University of Hong Kong(香港中文大学);Nanjing University of Posts and Telecommunications(南京邮电大学);Nankai University(南开大学)
Munich University of Applied Sciences, Intelligent Vehicles Lab (IVL)(慕尼黑应用技术大学,智能车辆实验室);FZI Research Center for Information Technology(弗劳恩霍夫信息技术研究中心);Continental Automotive Technologies GmbH(大陆汽车技术有限公司);Technical University of Berlin, Institute of Mathematics(柏林技术大学,数学系);German Aerospace Center (DLR), Institute of Transportation Systems(德国航空航天中心(DLR),交通运输系统研究所);Aptiv Services Deutschland GmbH(Aptiv 德国服务有限公司);AI Lab, ZF Friedrichshafen AG(ZF弗劳恩霍夫汽车有限公司人工智能实验室);DENSO AUTOMOTIVE Deutschland GmbH(DENSO 汽车德国有限公司);Karlsruhe Institute of Technology, Scientific Computing Center(卡尔斯鲁厄理工学院,科学计算中心)
Safety-Critical Traffic Simulation with Guided Latent Diffusion Model
The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州));School of Computer Science and Engineering(计算机科学与工程学院);Sun Yat-sen University(中山大学)
DECAMP: Towards Scene-Consistent Multi-Agent Motion Prediction with Disentangled Context-Aware Pre-Training
School of Computer Science and Engineering, Beihang University(北京航空航天大学计算机科学与工程学院);Robotics and Autonomous Systems Thrust, The Hong Kong University of Science and Technology (GZ)(香港科技大学(广州)机器人与自主系统方向);Division of Emerging Interdisciplinary Areas, The Hong Kong University of Science and Technology(香港科技大学新兴交叉领域研究所)
Invisible Servoing: a Visual Servoing Approach with Return-Conditioned Latent Diffusion
Robotics and Mechatronics group, Faculty of Electrical Engineering, Mathematics and Computer Science, University of Twente(特文特大学机器人与机电组,电气工程、数学与计算机科学学院);Department of Applied Mathematics, University of Twente(特文特大学应用数学系);Department of Computer, Control and Management Engineering, Sapienza University of Rome(罗马萨皮恩扎大学计算机、控制与管理工程系)
Georgia Institute of Technology(佐治亚理工学院);Honda Research Institute(本田研究所);University of California Berkeley(加州大学伯克利分校)
高度交互场景下的运动预测是自动驾驶中的一个具有挑战性的问题。在这种情况下,我们需要准确地预测交互代理的联合行为,以确保自动驾驶车辆的安全和高效导航。近年来,目标条件方法由于其在性能上的优势以及能够捕捉轨迹分布中的多峰性而受到越来越多的关注。在这项工作中,我们研究了目标条件框架下的联合轨迹预测问题。特别是,我们引入了一个条件变分自动编码器为基础的(CVAE)模型显式编码不同的相互作用模式到潜在的空间。然而,我们发现香草模型遭受后验崩溃,不能诱导出所需的信息潜在空间。为了解决这些问题,我们提出了一种新的方法,以避免KL消失,并诱导一个可解释的交互式潜在空间与伪标签。建议的伪标签使我们能够以灵活的方式将领域知识纳入交互。我们激励所提出的方法使用一个说明性的玩具的例子。此外,我们还通过定量和定性评估在Waymo Open Motion数据集上验证了我们的框架。
LetsMap: Unsupervised Representation Learning for Semantic BEV Mapping
University of Freiburg(弗莱堡大学);Qualcomm SARL France(法国高通公司);QT Technologies Ireland Limited(爱尔兰QT技术有限公司);Federal University of Rio Grande(里奥格兰德联邦大学);University of Technology Nuremberg(纽伦堡工业大学)
VAE-MS: An Asymmetric Variational Autoencoder for Mutational Signature Extraction
Center for Clinical Data Science(临床数据科学中心);Aalborg University(奥胡斯大学);Aalborg University Hospital(奥胡斯大学医院);Nuffield Department of Surgical Sciences(外科科学努尔菲尔德部门);University of Oxford(牛津大学);Nuffield Department of Women’s and Reproductive Health(妇女和生殖健康努尔菲尔德部门)
A Critical Assessment of Pattern Comparisons Between POD and Autoencoders in Intraventricular Flows
Center for Computational Simulation (CCS)(计算模拟中心(CCS));John A. Paulson School of Engineering and Applied Sciences(约翰·A·保罗森工程与应用科学学院);Harvard University(哈佛大学)
Learning from Risk: LLM-Guided Generation of Safety-Critical Scenarios with Prior Knowledge
Chinese Academy of Sciences, China(中国科学院);Department of Urban Studies and Planning, Massachusetts Institute of Technology, USA(麻省理工学院城市研究与规划系);School of Vehicle and Mobility, Tsinghua University, China(清华大学车辆与移动系统学院)