Autonomous Driving on Xu'Bloghttps://xuquant.com/zh/categories/autonomous-driving/Recent content in Autonomous Driving on Xu'BlogXu'Bloghttps://xuquant.com/images/profile.jpghttps://xuquant.com/images/profile.jpgHugo -- 0.152.2zhSun, 11 Jan 2026 10:00:00 +0800Vision-Language-Action Models for Autonomous Driving: The Cosmos-Reason Approachhttps://xuquant.com/zh/posts/autodrive/nvidia_vla/Sun, 11 Jan 2026 10:00:00 +0800https://xuquant.com/zh/posts/autodrive/nvidia_vla/深入技术解析 Nvidia Cosmos-Reason (Alpamayo) VLA 自动驾驶系统，涵盖三平面视觉编码、自车捷径规避、变化因数据集范式，以及通过强化学习实现的推理-动作对齐。End-to-End Autonomous Driving: From Modular Decoders to VLA Architectureshttps://xuquant.com/zh/posts/autodrive/e2e-autonomous-driving-evolution/Thu, 01 May 2025 10:00:00 +0800https://xuquant.com/zh/posts/autodrive/e2e-autonomous-driving-evolution/端到端自动驾驶架构演化的技术综述，涵盖规划器解码器选择（AR vs Diffusion vs Flow Matching）、VLA 集成策略，以及数据基础设施、训练优化和评估系统的工程最佳实践。Why Generative Planning? The Non-Convexity Argument Against Regression in Autonomous Drivinghttps://xuquant.com/zh/posts/autodrive/generative-planning-nonconvex/Sat, 15 Mar 2025 10:00:00 +0800https://xuquant.com/zh/posts/autodrive/generative-planning-nonconvex/从第一性原理分析回归式规划器在自动驾驶中失败的原因：可行域是非凸的，MSE 将模式平均到障碍物上，GMM 是补丁而非解决方案，生成式方法是必要的。