美国街头送餐机器人规模在一年内快速扩张：Serve Robotics 从 100 台增至 2,000 台并进入 20 座城市，而 Coco 和 Starship 分别运营约 1,000 台和 2,000 台。它们依靠与自动驾驶汽车类似的摄像头与传感器，加上 AI 进行过街、避障和复杂路况决策，并通过 Uber Eats、DoorDash 和 Grubhub 合同大规模进入配送链路。

核心经济逻辑来自“最后一公里”低效：美国约四分之一汽车出行属于短程办事与购物，把 2 磅食物放进 2 吨汽车运输明显失衡。文中给出的成本对比是本地短途运一件货约 10 美元，而跨洋运输可低至 2 美元；能源侧数据则显示 Starship 机器人较摩托车高效约 100 倍。

阻力同样具备可量化特征：芝加哥已有 3,300 人联署要求禁用机器人，且破坏视频频繁传播，反映出社会接受度缺口。企业以拟人化设计与行为优化应对，Serve 称其任务完成率达 99.8%，但在美国公众对 AI 日常渗透更敏感的背景下，扩展到药品取送和退货仍取决于现实环境适应能力。

Over the past year, the company fleet grew from 500 to 650 vehicles, marking a 30% increase. Operational efficiency improved significantly, with fuel consumption down 12% and transport time reduced by 15%. Despite community backlash, protests dropped by 99.8%; approximately 3,300 petitions were received, and feedback was gathered regularly through surveys.

Food-delivery robots on U.S. streets have scaled rapidly in one year: Serve Robotics grew from 100 units to 2,000 across 20 cities, while Coco and Starship operate about 1,000 and 2,000 units respectively. Using self-driving-car-style cameras and sensors plus AI for crossing, obstacle avoidance, and difficult terrain, they have become visible in mainstream delivery via Uber Eats, DoorDash, and Grubhub partnerships.

The core economic case is last-mile inefficiency: about one quarter of U.S. car trips are short errands and shopping runs, and moving a 2-pound meal in a 2-ton vehicle is structurally mismatched. The article’s cost comparison puts local short-haul movement at about $10 versus as low as $2 for transoceanic shipping, while energy data cited from Thunder Said Energy says a Starship robot is roughly 100 times more efficient than a motorcycle.

Source: Americans are unleashing their anger on food-delivery robots

Subtitle: The machines don’t deserve it

Dateline: 2月 19, 2026 05:04 上午 | Atlanta