Carbon Robotics Launches Plant Foundation Model and Autonomous Tractor Kit
The company is pivoting toward 'Physical AI' with a large-scale plant model and a retrofit kit for John Deere tractors.
Carbon Robotics has launched a new AI foundation model and an autonomy retrofit kit to streamline precision weeding. The move signals a shift toward "Physical AI," allowing farmers to customize robotic weeding in real-time without traditional software retraining.
At the center of this update is a "large plant model" pre-trained on 150 million labeled plants. This foundation model replaces the company's previous reliance on crop-specific AI. To achieve this scale, Carbon Robotics partnered with data-annotation firm iMerit, which helped the company expand its labeled dataset from a few thousand images to over a million.
This technical shift enables a new level of field flexibility. Using an iPad app, farmers can now customize their weeders by tagging a small number of thumbnails as either "crop" or "weed." According to Carbon Robotics CTO Alex Sergeev, the goal was to build a process where farmers provide examples and the model works instantaneously. This allows for immediate adjustments in the field without the need for software downloads.
The Shift to Physical AI
Traditionally, agricultural vision systems have been rigid, requiring time-consuming retraining every time a farmer introduced a new crop, encountered a different weed species, or moved to a new geographic region. By utilizing a foundation model that understands general plant structures, Carbon Robotics is removing these barriers to adoption. This approach makes autonomous weeding more scalable across diverse farming environments, as the AI no longer needs to be rebuilt for every specific scenario.
Expanding Autonomy with Carbon ATK
Alongside the AI model, the company released the Carbon ATK (Autonomous Tractor Kit). This kit is designed to retrofit John Deere 6R, 8R, 8RX, and 8RT series tractors from 2019 and newer for fully autonomous operation. The ATK provides capabilities for autonomous obstacle avoidance, path planning, and mission definition without requiring permanent modifications to the machinery.
Industry Implications
By combining a flexible AI foundation with a retrofit kit, Carbon Robotics is enabling farmers to convert existing fleet machinery into precision-weeding systems that can be adapted to different crops in minutes. This reduces the capital expenditure required for autonomy and eliminates the technical bottleneck of specialized AI training for individual farms.
What's Next
As Carbon Robotics pushes further into Physical AI, the industry will be watching how these foundation models perform across a wider variety of unpredictable outdoor environments. The success of the Carbon ATK will likely depend on the adoption rate among John Deere owners and the seamlessness of the retrofit process in diverse soil and weather conditions.