농산업에서도 AI, 디지털 트윈, 로봇 등 자동화와 디지털화 전환 추세가 예고되고 있다. 첨단 기술 스타트업 중심으로 AI 기술 개발과 도입이 확대되고 있는 가운데 엔비디아 기술이 농산업 곳곳에도 침투하고 있다.
▲A robot monitoring indoor strawberry cultivation
GPU-based crop health model, fruit harvesting, and quality model training
NVIDIA Jetson, low-cost, high-performance camera for data collection
Isaac Sim Simulation and Digital Twin Indoor Cultivation Utilization
The agricultural industry is also expected to see a shift toward automation and digitalization, with AI, digital twins, and robots. With AI technology development and adoption expanding, centered around cutting-edge technology startups, NVIDIA technology is also penetrating into every corner of the agricultural industry.
NVIDIA announced on the 28th that it is providing AI and robotics solutions to robot-based agricultural startup Zordi to support sustainable agricultural innovation.
Jordi is using NVIDIA software and GPU products to develop AI robots and automate indoor cultivation. Jordi's founder, Gilwoo Lee, had already received investment support for his autonomous farming business before completing his PhD in AI and Robotics Engineering at the University of Washington, and founded Jordi in 2020 with Casey Call, former CEO of vertical farming startup Plenty.
Last year, Jordi signed a three-party joint research and development agreement with Hongseong-gun and Hetero, a private strawberry breeding company, to expand the cultivation of 'Honghee strawberries', a local specialty crop of Hongseong-gun, and is sharing AI and robotics technology to help automate cultivation and expand exports to the U.S.
Jordi uses two types of robots in his hydroponic operations: a scout robot that uses a basic model to gather information about the health of the plants; and a harvest robot that carefully harvests and arranges the fruit and handles other tasks.
Using NVIDIA GPUs, they train everything from crop health models to fruit harvesting and quality assessment models. The company aims to deploy autonomous greenhouse systems around the world to support local markets, reducing the carbon footprint of transportation and growing fresher, tastier fruit sustainably.
It has supplied fruit to Wegmans, a U.S. supermarket chain, with a robotic strawberry grower and is now expanding its business with partner greenhouse farms in New Jersey and California. It is bringing NVIDIA-accelerated AI automation to indoor growing, which is in many ways similar to the advancements in manufacturing and logistics operations.
■ Jetson AGX Orin, Agricultural AI Automation Jordi is developing AI models and robots using the NVIDIA Jetson AGX Orin module. Through this, sensor data is collected and models that recognize the following are tested: △Health status of plants, flowers, and fruits; △Initial symptoms of pests and diseases; △Need for water and nutrition; △Light and temperature management.
“Jetson’s energy efficiency and its ability to use low-cost, high-performance cameras from NVIDIA partners were attractive factors to Jordi,” the representative said. Jordi installs multiple cameras on each robot to collect data.
The representative also said, “Jetson allows us to use a lot of low-cost cameras. With Jetson, we can use a variety of cameras and also manage batteries better.”
We also have a vision to contribute to solving the labor shortage in agricultural fields by developing monitoring robots and harvesting robots. “We hope that harvesting robots can perform more sophisticated tasks, such as pruning leaves, with the help of simulation,” said CEO Lee.
■ Omnibus Isaac Sim, Strengthening the Agricultural Digital Twin Like industrial manufacturers, Jordi is exploring ways to improve its indoor growing operations with AI, optimizing operations through simulations and digital twins using Omniverse’s Isaac Sim.
The software platform visualizes all data collected from the robot sensors and provides a real-time dashboard with a spatial map, allowing real-time visibility of all plants in the facility, making it easy to monitor production remotely.
It supports farm management for inexperienced farm operators by analyzing plant health using basic models and providing selective crop management recommendations. “It’s literally just one more step for us to apply all of this to Isaac Sim and Omniverse,” said Lee, whose PhD thesis explored reinforcement learning and the transfer from simulation to reality.
Jordi is working on grip simulations for cucumbers and tomatoes, as well as strawberries, and is working on expanding into other categories. “For strawberries and other crops, if you handle them carefully, it can help extend their shelf life,” he said.
The goal is to make farming easier and more profitable by creating fully autonomous agricultural systems where AI recommends sustainable crop management decisions and robots do the heavy lifting.
“What’s really important to us is how do we automate all of this and how do we implement AI that actually makes decisions and thinks on a farm where there’s a lot of automation,” the representative said.