<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[u013250861的博客]]></title><description><![CDATA[]]></description><link>https://blog.csdn.net/u013250861</link><language>zh-cn</language><generator>https://blog.csdn.net/</generator><copyright><![CDATA[Copyright &copy; u013250861]]></copyright><item><title><![CDATA[自动驾驶端到端大模型VLM/VLA论文笔记：ORION]]></title><link>https://blog.csdn.net/u013250861/article/details/160420912</link><guid>https://blog.csdn.net/u013250861/article/details/160420912</guid><author>u013250861</author><pubDate>Thu, 23 Apr 2026 00:40:20 +0800</pubDate><description><![CDATA[自动驾驶端到端大模型VLM/VLA论文笔记：ORION，ReCogDrive。]]></description><category></category></item><item><title><![CDATA[无人驾驶：名词02【Traffic-related contexts：交通相关的其他元素（红绿灯等）】]]></title><link>https://blog.csdn.net/u013250861/article/details/160420631</link><guid>https://blog.csdn.net/u013250861/article/details/160420631</guid><author>u013250861</author><pubDate>Thu, 23 Apr 2026 00:14:26 +0800</pubDate><description><![CDATA[无人驾驶：名词02【Traffic-related contexts：交通相关的其他元素（红绿灯等）】]]></description><category></category></item><item><title><![CDATA[无人驾驶：名词04【Multi-modal Trajectory：多模态输出轨迹（变道、加速、减速等多种可能轨迹）】]]></title><link>https://blog.csdn.net/u013250861/article/details/160420557</link><guid>https://blog.csdn.net/u013250861/article/details/160420557</guid><author>u013250861</author><pubDate>Thu, 23 Apr 2026 00:12:20 +0800</pubDate><description><![CDATA[无人驾驶：名词03【Multi-modal Trajectory：多模态输出轨迹（变道、加速、减速等多种可能轨迹）】]]></description><category></category></item><item><title><![CDATA[无人驾驶：名词03【Planning Trajectory：主车输出轨迹】【Prediction Trajectory：动态障碍物预测轨迹】]]></title><link>https://blog.csdn.net/u013250861/article/details/160420501</link><guid>https://blog.csdn.net/u013250861/article/details/160420501</guid><author>u013250861</author><pubDate>Thu, 23 Apr 2026 00:11:11 +0800</pubDate><description><![CDATA[【Multi-modal Trajectory：多模态输出】]]></description><category></category></item><item><title><![CDATA[无人驾驶：名词01【AV：主车】【Agent：动态障碍物（社会车辆）】【Static Obstacle：静态障碍物（锥桶、水马等）】【Map：地图元素（车道线/道路边界等）】]]></title><link>https://blog.csdn.net/u013250861/article/details/160420321</link><guid>https://blog.csdn.net/u013250861/article/details/160420321</guid><author>u013250861</author><pubDate>Thu, 23 Apr 2026 00:09:54 +0800</pubDate><description><![CDATA[自动驾驶：名词01【AV：主车；Agent：动态障碍物；Planning Trajectory：主车输出轨迹；Prediction Trajectory：动态障碍物预测轨迹；Multi-modal】自动驾驶：名词【AV：主车；Agent：动态障碍物；Static Obstacle：静态障碍物；Map：地图元素（车道线/道路边界）】Traffic-related contexts：交通相关元素。]]></description><category></category></item><item><title><![CDATA[端到端 Planner 中 Trajectory 如何用于计算 Loss？自动驾驶领域的经典 Loss 总结]]></title><link>https://blog.csdn.net/u013250861/article/details/160383314</link><guid>https://blog.csdn.net/u013250861/article/details/160383314</guid><author>u013250861</author><pubDate>Wed, 22 Apr 2026 00:13:00 +0800</pubDate><description><![CDATA[ADE：Average Displacement Error，平均位移误差FDE：Final Displacement Error，终点位移误差AHE：Average Heading Error，平均航向误差FHE：Final Heading Error，终点航向误差LcomfortLdirLcomfort​Ldir​抖动急刹急转逆行不顺着车道走。]]></description><category></category></item><item><title><![CDATA[端到端 Planner 中 Trajectory 如何用于计算 Loss？自动驾驶领域的经典 Loss 总结]]></title><link>https://blog.csdn.net/u013250861/article/details/160383223</link><guid>https://blog.csdn.net/u013250861/article/details/160383223</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 22:59:13 +0800</pubDate><description><![CDATA[ADE：Average Displacement Error，平均位移误差FDE：Final Displacement Error，终点位移误差AHE：Average Heading Error，平均航向误差FHE：Final Heading Error，终点航向误差LcomfortLdirLcomfort​Ldir​抖动急刹急转逆行不顺着车道走。]]></description><category></category></item><item><title><![CDATA[一段式端到端自动驾驶技术方案深度研究报告【2026-04-21】]]></title><link>https://blog.csdn.net/u013250861/article/details/160382955</link><guid>https://blog.csdn.net/u013250861/article/details/160382955</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 22:38:22 +0800</pubDate><description><![CDATA[是连续轨迹、token 化动作、occupancy rollout 还是 diffusion 采样。扩散与 VLA 的兴起，本质都是在解决“单峰回归不足以表达驾驶行为”的问题，只是前者从概率建模入手，后者从语义认知入手。]]></description><category></category></item><item><title><![CDATA[自动驾驶：Trajectory格式【nuPlan(x,y,ψ,t) 、Waymo(x,y,bbox_yaw,vₓ,vᵧ,valid)、AV2、nuScenes】]]></title><link>https://blog.csdn.net/u013250861/article/details/160382361</link><guid>https://blog.csdn.net/u013250861/article/details/160382361</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 21:52:48 +0800</pubDate><description><![CDATA[从公开数据集看，差异主要在 4 件事上：1）是只标注，还是标注；2）坐标系用还是；3）每个轨迹点只存xy，还是还带上ψ（航向）、v（速度）、时间戳、valid mask 等；4）时间采样密度和预测时域不同。nuPlan 更像“规划监督”，Waymo/Argoverse 2 更像“多目标轨迹/运动预测监督”，nuScenes 的 prediction 接口则常被拿来做未来点监督。]]></description><category></category></item><item><title><![CDATA[一段式端到端04：基于VLA的一段式端到端（ORION、OpenDriveVLA、ReCogDrive）]]></title><link>https://blog.csdn.net/u013250861/article/details/160381984</link><guid>https://blog.csdn.net/u013250861/article/details/160381984</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 21:22:54 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[一段式端到端03：基于扩散模型的一段式端到端（DiffusionDrive、Diffusion Planner、DiffE2E）]]></title><link>https://blog.csdn.net/u013250861/article/details/160381959</link><guid>https://blog.csdn.net/u013250861/article/details/160381959</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 21:21:37 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[一段式端到端02：基于世界模型的一段式端到端（Drive-OccWorld、OccLLaMA）]]></title><link>https://blog.csdn.net/u013250861/article/details/160381937</link><guid>https://blog.csdn.net/u013250861/article/details/160381937</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 21:20:26 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[一段式端到端01：基于感知的一段式端到端（UniAD、VAD、PARA-Drive）]]></title><link>https://blog.csdn.net/u013250861/article/details/160381904</link><guid>https://blog.csdn.net/u013250861/article/details/160381904</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 21:19:27 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[二段式端到端：PLUTO、CarPlanner、Plan-R1]]></title><link>https://blog.csdn.net/u013250861/article/details/160381877</link><guid>https://blog.csdn.net/u013250861/article/details/160381877</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 21:17:45 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[端到端规划:PLUTO【Pushing the Limit of Imitation Learning-based Planning for Autonomous Driving】【名词介绍】]]></title><link>https://blog.csdn.net/u013250861/article/details/160381214</link><guid>https://blog.csdn.net/u013250861/article/details/160381214</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 21:03:02 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[π0.7：a Steerable Generalist Robotic Foundation Model with Emergent Capabilities]]></title><link>https://blog.csdn.net/u013250861/article/details/160380261</link><guid>https://blog.csdn.net/u013250861/article/details/160380261</guid><author>u013250861</author><pubDate>Tue, 21 Apr 2026 20:03:01 +0800</pubDate><description><![CDATA[Figure 1: Fig. 1: We introduce π0.7\pi _ { 0 . 7 }π0.7​ , a steerable generalist robot foundation model that can perform dexterous tasks across many tasks, environments, and robots. π0.7\pi _ { 0 . 7 }π0.7​ is trained with diverse prompts that contain not ]]></description><category></category></item><item><title><![CDATA[GoalFlow：端到端自动驾驶中的多模态轨迹生成]]></title><link>https://blog.csdn.net/u013250861/article/details/160348356</link><guid>https://blog.csdn.net/u013250861/article/details/160348356</guid><author>u013250861</author><pubDate>Mon, 20 Apr 2026 23:41:33 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[自动驾驶：数据集采集和标注]]></title><link>https://blog.csdn.net/u013250861/article/details/160348250</link><guid>https://blog.csdn.net/u013250861/article/details/160348250</guid><author>u013250861</author><pubDate>Mon, 20 Apr 2026 23:27:50 +0800</pubDate><description><![CDATA[自动驾驶中数据集如何采集和标注？一条视频告诉你]]></description><category></category></item><item><title><![CDATA[自动驾驶：数据闭环]]></title><link>https://blog.csdn.net/u013250861/article/details/160347966</link><guid>https://blog.csdn.net/u013250861/article/details/160347966</guid><author>u013250861</author><pubDate>Mon, 20 Apr 2026 23:16:10 +0800</pubDate><description><![CDATA[开环or闭环？自动驾驶数据应该这样选]]></description><category></category></item><item><title><![CDATA[202106-nuPlan：面向自动驾驶车辆的基于 ML 的闭环规划基准【多模态的，这意味着对于每个agent，我们会预测最有可能的N条轨迹】]]></title><link>https://blog.csdn.net/u013250861/article/details/160313507</link><guid>https://blog.csdn.net/u013250861/article/details/160313507</guid><author>u013250861</author><pubDate>Sun, 19 Apr 2026 20:32:50 +0800</pubDate><description><![CDATA[在本工作中，我们提出了全球首个用于自动驾驶的闭环机器学习规划基准。尽管基于机器学习的运动规划器日益增多，但由于缺乏公认的数据集和评估指标，该领域的进展受到了限制。现有的自动驾驶汽车运动预测基准主要集中在短期运动预测，而非长期规划。这导致以往的研究采用开环评估和基于 L2 的指标，而这些方法并不适合公平地评估长期规划。我们的基准通过引入大规模驾驶数据集、轻量级闭环仿真器以及运动规划专用指标，克服了这些局限性。]]></description><category></category></item></channel></rss>