Baidu Discloses Apollo World World Model; Apollo Go Has Surpassed One Million Kilometers of Fully Unmanned Operations
English text is being prepared.
At its AI Day special event, Baidu Apollo disclosed for the first time Apollo World, an agent-driven omni-modal world model for physical AI, and proposed a "2+3" criteria for evaluating world models. The model models traffic participants such as vehicles and pedestrians in virtual scenarios as AI agents with decision-making and interaction capabilities, for use in training and validating unmanned vehicles in virtual cities. Apollo World is currently used in the global testing and validation of Apollo Go, which has surpassed one million kilometers of end-to-end fully unmanned operations; Apollo Go has launched in 28 cities worldwide, completed more than 23 million cumulative ride services, and its total autonomous driving mileage has exceeded 350 million kilometers.
Why read this
AI score 56
Baidu discloses the Apollo World world model for the first time, modeling traffic participants as AI agents, and presents the deployment data showing that Apollo Go has surpassed one million kilometers of fully unmanned operations.
Topics
Body
相关专题top start 相关专题top end 期货推广begin 期货推广end 秒拍begin 秒拍end 视频播放器start 视频播放器end 行情图begin 行情图end
新浪科技讯 11月4日下午消息,在百度AI Day Apollo专场上,百度Apollo首次披露面向物理AI的智能体驱动全模态世界模型Apollo World,并提出世界模型“2+3”判断标准。目前,该模型已用于萝卜快跑全球化测试与验证,端到端全无人运营超百万公里,助力无人驾驶规模化落地。
据悉,Apollo World可将虚拟场景内所有交通参与者建模为具备决策与交互能力的AI智能体,支持无人车在虚拟城市任意起终点通行,打造提前验证“无限未来”的虚拟场景订单。基于Apollo World提供的虚拟训练环境,AI Driver可学习、验证和迭代。
在传统仿真里,其他交通参与者更像是按脚本运行的道具。但在Apollo World构建的虚拟城市中,车辆、行人等交通参与者均被建模为具备决策和交互能力的AI智能体,能够根据周围交通变化调整行动。无人车减速、变道或让行,都可能引发其他参与者的不同反应,使无人驾驶系统在持续变化的交通环境中接受训练与检验。
有行业人士分析称,Apollo World可“日行十万公里”,其一天完成的虚拟路测里程相当于一辆配备安全员的真实路测车连续测试约三年。依托世界模型带来的能力跃升,叠加数据飞轮效应,行业将呈现强者恒强的格局,这一趋势也正把行业推向等待已久的“AGI时刻”。
无人驾驶要走向规模化,必须同时兼顾安全、智能与规模,而这三者之间存在着一个“不可能三角”。长尾场景难靠实车里程充分覆盖,新区域部署存在测试成本与试错风险,跨路段、跨场景连续驾驶能力又需规模化验证。Apollo World把真实运营中的复杂交互与风险场景转化为可重复、可控制、可规模化运行的虚拟测试环境,并生成现实尚未覆盖的新场景,让无人车在实车部署前先在虚拟世界完成驾驶推演,把测试成本和部署风险留在虚拟世界先行消化,让AI教AI开车。
据悉,目前萝卜快跑已落地全球28座城市,累计完成超2300万次出行服务,自动驾驶总里程突破3.5亿公里,其中全无人驾驶里程超过2.4亿公里,并已在左右舵市场开展无人驾驶实践。依托真实驾驶数据,Apollo World提取高价值场景,虚拟测试结果再反哺车端迭代,形成真实运营与虚拟验证相互促进的闭环。(文猛) news_keyword_pub,stock, 正文下iframe 正文下iframe 总声明 总声明 文末二维码 start 文末二维码 start
文末二维码 start 文末二维码 start 编辑姓名及工作代码
编辑姓名及工作代码