Bonsai Robotics has introduced Bonsai World, a simulation system that recreates real-world fields and operating conditions in 3D. By combining field data with AI-generated environments, the US company aims to train and test autonomous machines before they enter a new field, crop or working environment. This could reduce the need for extensive on-site testing and adjustments.
Bonsai World is a new component of Bonsai Intelligence, the autonomy software used on the company’s Amiga robotic platform and retrofitted machines from other manufacturers. The technology builds on more than 50 million real-world data samples collected across over 1 million acres (405,000 hectares), covering different crops, terrain, weather conditions, lighting, machines and operations.
From satellite images to 3D fields
The system starts with satellite imagery of a farm or other working environment. Using Google’s Gemini vision-language model, Bonsai World converts the two-dimensional imagery into a structured 3D environment in which autonomous machines can simulate routes and interactions with their surroundings.
The software can also generate realistic ground-level images and introduce situations that robots may encounter in practice, including dust, debris, animals, other vehicles and changing terrain. This allows Bonsai to expose its autonomy software to conditions that its machines have not yet encountered in the field. The simulations generate additional training data, supplementing information continuously collected by machines already operating in practice.
Less field testing needed
According to Bonsai Robotics, the main advantage is that autonomy software can be prepared for a specific location or application before a machine arrives. Normally, introducing an autonomous machine to a new crop, terrain or task may require additional data collection, field testing and adjustments to the software. Bonsai aims to reduce these steps by recreating likely operating conditions in advance. “Every environment we operate on expands what the system understands,” says Tyler Niday, CEO and co-founder of Bonsai Robotics. He explains that the company initially focused on difficult conditions, including unreliable GPS, limited connectivity and extremely poor visibility caused by dust. Bonsai World is intended to extend that experience to new environments without having to physically encounter every possible situation first.
Google and NVIDIA technology
Bonsai developed the system using Google Cloud and NVIDIA computing technology.
Google’s Gemini model interprets satellite imagery, while Bonsai’s own world model generates realistic simulations based on the company’s field data. NVIDIA GPUs are used to train the models and generate simulated environments. The technology is intended to support the development of autonomy across different crops, machines and tasks. Bonsai currently operates in specialty-crop agriculture in the United States and Australia and is expanding its technology into other sectors, including mining and defence.
Field performance remains to be proven
For growers, the potential benefit is faster deployment of autonomous machines in unfamiliar fields and conditions, with less time needed for initial setup and software adjustments. However, Bonsai has not disclosed independently verified results showing how much field-testing time the new system saves or how accurately performance in simulation translates to actual field operations.
To discover the latest innovations and trends in zero-emissions off-highway machines, meet with solution providers and hear talks from industry leaders, book your place to attend the 9th Design And Development Of Zero-Emissions Off-Highway Machinery Summit, taking place February 24-25, 2027 in Chicago, Illinois.
For more information, visit our website or email us at info@innovatrix.eu for the event agenda. Visit our LinkedIn to stay up to date on our latest speaker announcements and event news.
Source:
