Covariant is an artificial intelligence company that develops AI-powered robotic systems for warehouse automation, logistics, and supply chain operations. Founded in 2017 by Pieter Abbeel, Peter Chen, Rocky Duan, and Tianhao Zhang, the company is headquartered in Emeryville, California. Covariant emerged from cutting-edge research at the University of California, Berkeley, with the mission of building a universal AI capable of enabling robots to see, reason, and act in the real world. The company's core technology is the Covariant Brain, a foundation model for robotic intelligence that enables industrial robots to handle a virtually unlimited range of items in warehouse and fulfillment environments. Unlike traditional robotic systems that require explicit programming for each object type, the Covariant Brain uses deep learning, reinforcement learning, and computer vision to generalize across novel objects, enabling robots to pick, sort, and place items they have never encountered before. This capability is particularly valuable in e-commerce fulfillment, where the diversity of products makes traditional automation impractical. Covariant's AI integrates with existing robotic hardware from leading manufacturers, transforming standard industrial arms into intelligent picking systems. The platform processes visual data in real time to identify items, plan grasps, and execute picks with high accuracy and throughput, even for challenging items such as deformable goods, transparent packaging, and irregularly shaped products. The system continuously learns from its operational experience, improving performance over time as it encounters new scenarios. In 2024, Covariant announced a significant partnership with Amazon, which acquired a license to Covariant's AI technology and hired a portion of the company's team. Covariant operates on an enterprise model, providing its AI platform as a service to logistics and supply chain companies with pricing based on deployment scale and throughput requirements.

AI自动化工具

Covariant 通过将其 AI 与标准工业机械臂集成来自动化仓库拾取和分拣操作。该平台将手工履行流程转变为自主工作流，可处理多样化的产品目录，包括可变形商品和不规则形状物品，这些都是传统自动化系统的难点。

AI电子商务工具

Covariant 的机器人拾取技术特别适用于电子商务履行运营，其中巨大的产品类型多样性使传统自动化变得不切实际。该 AI 使机器人能够处理在线零售目录中的全部物品范围，提高订单履行中心的吞吐量和准确性。

AI 研究工具

Covariant 由来自加州大学伯克利分校的顶尖 AI 研究人员创立，在强化学习、仿真转实物迁移和机器人操控基础模型方面做出了贡献。该公司的 Covariant Brain 代表了构建通用机器人智能的新方法，可从真实运营数据中学习。

AI机器人

Covariant 开发了 Covariant Brain，一个机器人智能基础模型，能使工业机器人在仓库环境中对几乎任何物品进行拾取、分拣和放置。该 AI 使用深度学习和计算机视觉在不需明确编程的情况下对新颖物体进行泛化，使机器人能够以高精度和高速度处理之前从未见过的物品。