Group 1: Smart Driving - The smart driving sector maintains a robust upward trend, with the autonomous vehicle market continuously catalyzing growth [1] - Pony.ai has launched its first road tests for Robotaxi in Luxembourg, collaborating with local mobility company Emile Weber, with initial tests starting in the city of Reninge [1] - Pony.ai's global autonomous driving testing mileage has surpassed 45 million kilometers, establishing an operational network covering over 2,000 square kilometers in major cities like Beijing, Shanghai, Guangzhou, and Shenzhen [1] - JD Logistics has introduced its self-developed VAN unmanned light truck, which features a cargo space of 24 cubic meters, the largest in the logistics industry, and can replace traditional 4.2-meter trucks [1] - The VAN unmanned light truck has a maximum range of 400 kilometers when fully loaded and is capable of L4 level autonomous driving on public roads, potentially saving about 60% in costs compared to traditional transportation [1] Group 2: Robotics - The robotics sector also shows a steady upward trend, with domestic robots expected to enter small-batch mass production in the second half of the year despite supply chain fluctuations [2] - Weichuang's humanoid robot Atom has achieved mass delivery and can be remotely controlled from 1,800 kilometers away [2] - ByteDance has produced over 1,000 units of its developed robots, primarily wheeled logistics robots for transporting packages and parts in warehouses and production lines [2] - Horizon Robotics, in collaboration with various institutions, has proposed RoboTransfer, which enhances the training data for robot strategy models, leading to a 251% improvement in model generalization [2] Group 3: Investment Recommendations - ROBO+ is identified as the strongest industrial trend in the automotive sector, with smart driving and humanoid robots being the two most important directions of embodied intelligence [3] - The penetration rate of advanced smart driving is expected to enter a phase of explosive growth by 2025, driving demand for high-performance chips, lidar, optical components, and sensor cleaning systems [3] - Companies with solid fundamentals and high certainty in their robotics business are recommended, especially those with low valuations compared to their main business [3] - The long-term outlook for the robotics industry remains positive, with a complete domestic supply chain and world-leading reserves [3]

近年来，随着人工智能从感知智能向决策智能演进， 世界模型 （World Models） 逐渐成为机器人领域的重要研究方向。世界模型旨在让智能体对环境进行建模并 预测未来状态，从而实现更高效的规划与决策。 与此同时，具身数据也迎来了爆发式关注。因为目前具身算法高度依赖于大规模的真实机器人演示数据，而这些数据的采集过程往往成本高昂、耗时费力，严重 限制了其可扩展性和泛化能力。尽管仿真平台提供了一种相对低成本的数据生成方式，但由于仿真环境与真实世界之间存在显著的视觉和动力学差异（即 sim-to- real gap），导致在仿真中训练的策略难以直接迁移到真实机器人上，从而限制了其实际应用效果。 因此如何高效获取、生成和利用高质量的具身数据，已成为当 前机器人学习领域的核心挑战之一 。 项目主页： https://horizonrobotics.github.io/robot_lab/robotransfer/ 模仿学习（Imitation Learning）已成为机器人操作领域的重要方法之一。通过让机器人 "模仿" 专家示教的行为，可以在复杂任务中快速构建有效的策略模型。然 而，这类方法通常依赖大量高质量的真实机器 ...