点击下方 卡片 ，关注" 自动驾驶之心 "公众号 戳我-> 领取 自动驾驶近30个 方向 学习 路线 >>自动驾驶前沿信息获取 → 自动驾驶之心知识星球 国庆后高质量的工作不少，分享给大家。涉及VLA、世界模型、强化学习、场景重建3DGS、轨迹规划等方向！ DriveVLA-W0 世界模型增强自动驾驶VLA中的数据缩放定律 中国科学院自动化研究所等单位 提出了 DriveVLA-W0 训练范式，通过世界建模预测未来图像提供密集自监督信号，显著增强了 VLA 模型的泛化能力与数据扩展 性，在 NAVSIM v1/v2 基准上分别达到 93.0 PDMS 与 86.1 EPDMS，并将推理延迟降至基线 VLA 的 63.1%。 主要贡献： 算法框架： 论文标题：DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving 论文链接：https://arxiv.org/pdf/2510.12796 1. 识别并解决 VLA 模型的 "监督不足" 瓶颈 ：提出 DriveVLA-W0 训练范式，以未来图像预测为核心的世界建模（Worl ...

Core Insights - Stellantis is partnering with Pony AI to introduce self-driving electric vehicles in Europe, which is seen as a necessary catalyst for Stellantis amid its recent struggles [1][3] - The stock performance of Stellantis has been poor, with a decline from approximately $27 in March 2024, reflecting challenges in its underlying business [2][8] - Pony AI, a Chinese autonomous mobility technology company, is looking to enhance its technology scale, particularly in Europe, where it has existing deals for testing its technologies [5][6] Company Performance - Stellantis reported trailing sales of $170 billion last year, down from $208 billion two years ago, indicating significant challenges in the automotive market [5][8] - The company faces various challenges, including manufacturing issues, tariff impacts, and pricing pressures due to changing consumer affordability [9] - Pony AI's revenue was reported at $85 million last year, highlighting its need for financial strengthening despite the positive news regarding the partnership [4][10] Market Context - The autonomous driving space is competitive, with major players like Tesla and Google leading the way, making it difficult for other companies to gain traction [6][7] - The overall auto industry is experiencing headwinds, with declining sales reported by major companies, including Tesla, which is perceived differently due to its technology and software focus [7][9] - The partnership may provide Stellantis with an opportunity to integrate technology into its manufacturing and core automobile markets, potentially benefiting both companies [6][7]

2024年11月，小马智行以"PONY"为代码在纳斯达克完成上市，成为全球Robotaxi第一股。10月16日，其收盘价报20.415美元，较其13 美元的美股IPO发行价涨超57%。 港股研究社讯，据港交所10月17日披露，中国头部自动驾驶企业小马智行已正式通过港交所聆讯，即将踏入港股市场。 小马智行PHIP版招股书显示，公司2022年至2024年营收规模持续扩大，2025年上半年增速尤为明显，同比增长43.3%，达3543万美元 （约合人民币2.54亿元）。Robotaxi业务收入在2025年上半年涨势强劲，达325.6万美元（约合人民币2332万元），同比激增178.8%， 其中乘客车费收入在2025年第一季度和第二季度分别实现约800%与超300%同比飙升。 ...

点击下方 卡片 ，关注" 自动驾驶之心 "公众号 戳我-> 领取 自动驾驶近30个 方向 学习 路线 背景介绍 自动驾驶规划 的核心挑战之一在于：如何在多车密集、行为多样的交通场景中，实现安全、可靠、类人的决策。传统基于规则的规划方法依赖人工定义的行为准 则，虽然可解释性强，但在动态、复杂的交通博弈中缺乏泛化能力；而基于学习的规划方法试图直接从海量驾驶数据中学习策略，提升模型的适应性与可扩展性。 然而，当前学习型规划方法仍面临两大核心挑战： 本文希望解决的核心问题是： 能否在不引入额外先验知识的情况下，通过模型自身的学习能力，从有限的数据中提升博弈能力？ 自动驾驶决策任务通常被建模为 一个轨迹生成任务，因此，要想生成高质量的轨迹，需要 更加灵活的轨迹建模策略 。同时，博弈场景涉及自车与周围环境的复杂时空交互关系。例如，当车辆需 要在密集车流中变道时，模型必须同时关注当前车道及目标车道中前后方车辆的行为。这要求模型具备高效且自适应的特征整合与处理能力，能够快速识别并聚焦 关键交通参与者的动作与意图。最后，为了能够在与周围交通参与者的博弈过程中作出准确的反应，模型在 轨迹生成阶段 需要充分考虑周围交通参与者的行为 ...

Waymo has a reputation for moving deliberately in its autonomous-driving rollout. Now it's picking up speed as the market looks to get more competitive. ...

Autonomous-driving firms Pony AI and WeRide have received approval from China’s securities regulator to list shares in Hong Kong, as renewed delisting risks emerge in the US https://t.co/ONhNS0FF71 ...

Core Viewpoint - Chinese autonomous driving firm WeRide is preparing for a secondary listing in Hong Kong, having engaged Morgan Stanley and China International Capital Corp (CICC) for this process [1] Company Summary - WeRide has taken steps to expand its market presence by planning a secondary listing, indicating confidence in its growth potential and the attractiveness of the Hong Kong market for capital raising [1] Industry Summary - The move reflects a broader trend in the autonomous driving sector, where companies are seeking additional funding avenues to support innovation and expansion in a competitive landscape [1]

British autonomous driving technology group Wayve is in talks with investors, including Microsoft and SoftBank , to raise as much as $2 billion in funding, which could value the start-up at about $8 b... ...

Group 1 - Aurora Innovation, Inc. (NASDAQ:AUR) has been recognized as one of the best-performing midcap tech stocks over the last three years [1] - Analysts from Morgan Stanley and Cantor Fitzgerald have reiterated a Buy rating on Aurora, with a price target of $12, while Canaccord Genuity's analyst set a more optimistic target of $15, indicating a potential upside of 191% from current levels [1][2] - The company specializes in developing autonomous driving systems for long-haul trucking and freight logistics, utilizing its proprietary Aurora Driver platform that integrates hardware, software, and machine learning for fully driverless operations [3]

Aeva Technologies, Inc. (NASDAQ:AEVA) is one of the Best Small-cap EV Stocks to Buy According to Analysts. On September 30, Aeva Technologies, Inc. (NASDAQ:AEVA) released a multi-sensor FMCW 4D LiDAR and camera dataset for autonomous vehicle research. The dataset is recognized as the industry’s first open dataset, which features key elements including semantic segmentations, tracking, lane annotations, and object velocity measurements. Management noted that this open dataset provides researchers with high ...