以下文章来源于BigBite思维随笔 ，作者BigBite BigBite思维随笔 . 为什么需要纵向端到端 为什么我们需要纵向端到端？如果大家还能记起横向端到端前的城市辅助驾驶功能是什么样子，就会回忆起那时候的城市辅助驾驶在绕行超车时候非常犹 豫，挣扎，经常会被莫名其妙的障碍物卡死导致接管。正如单纯依赖规则，依赖动态规划等方法无法完全实现拟人高效的超车，绕行等动作一样，单纯依 赖规则或者优化方法也难以保证纵向控制的拟人丝滑。 Big Bite Small Talk, 杂谈随笔，聊科技，AI，成长，理财，经验杂谈。Stay Hungry 而有驾驶经验的人应该都知道，日常驾驶中区分新手司机和老司机的关键其实在于老司机对速度，对刹车的控制力，踩油门，甚至打方向新手都能一学就 会，真正的驾驶智慧体现在对减速的控制。 作者 | EatElephant 来源 | 知乎 原文链接： https://zhuanlan.zhihu.com/p/1950344892866883662 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 ...

Core Viewpoint - The launch of DeepRoute IO 2.0 by Yuanrong Qixing marks a significant advancement in the autonomous driving sector, transitioning from "end-to-end" models to the more sophisticated Vision-Language-Action (VLA) model, which is better suited for complex driving scenarios [1][2]. Group 1: Technology and Development - DeepRoute IO 2.0 integrates a self-developed VLA model that enhances the ability to handle intricate road conditions compared to traditional models [1]. - The VLA model addresses the "black box" issue of conventional end-to-end models by providing effective information linkage, analysis, and causal reasoning [1][2]. - The VLA technology is expected to be adaptable for vehicles priced above 150,000 yuan, with potential to extend down to 100,000 yuan models in the future [5]. Group 2: Market Position and Competition - The competition in the autonomous driving market is intensifying, with various automakers like Xiaopeng, Li Auto, Geely, and Chery entering the VLA model space [1]. - Yuanrong Qixing's CEO, Zhou Guang, emphasizes the company's early investments in defensive driving, which provide a competitive edge in the evolving landscape of autonomous driving technology [1][2]. Group 3: Safety and Commercialization - The company has achieved significant commercial milestones, securing over 10 model partnerships and delivering nearly 100,000 vehicles equipped with urban navigation assistance systems [6]. - Zhou Guang acknowledges the heightened consumer expectations for safety in autonomous driving, indicating a need for continuous improvement despite the company's leading position in evaluations [6][9]. - The focus on "defensive driving" is central to the system's training, aiming to instill a sense of risk awareness in AI, akin to biological evolution [9]. Group 4: Future Outlook - The market for high-level assisted driving is anticipated to expand, providing Yuanrong Qixing with greater opportunities as they continue to innovate and adapt their technology [9].

防御性驾驶分两种：1️⃣ 开 cybertruck —— 物理防御。2️⃣ 开 劳斯莱斯、宾利、法拉利 —— 魔法防御。不管哪一种防御都颇为破费。 ...