By: Karli Petrovic

Agriculture has a big data problem. With the implementation of autonomous robots and other advancing technologies, farm equipment is gathering more information than ever before. Normally, this would be a good thing. More data often translates to greater knowledge, deeper insights and better decision-making—if you can understand it, that is. This is the challenge for farmers. The data is available, but the ability to interpret it and use it effectively can be tricky.

That’s where NVIDIA comes in. The accelerated computing company is developing solutions that help farmers do more with their ag equipment, the technologies that power the machines and the data that is collected as a result.

“People often wonder what a company that makes chips and hardware is doing at an agriculture event like FIRA, but we’re not a general purpose computing company,” says Amit Goel, NVIDIA’s head of product management and ecosystem. “Accelerated computing requires you to have a deep domain understanding. You really need to know the application and understand the customer’s workflow and figure out how you can map it. We see all the issues that an industry faces, and we try to see where we can add value. We’re thinking about how we can solve agriculture’s problems using our computing platform.”

NVIDIA is engaged with the agriculture industry in a number of important ways. One big project within the company is Earth-2. The technology provides high-resolution climate and weather simulations to help farmers and others predict weather conditions and act based on what they learn. NVIDIA also uses generative artificial intelligence and large language models to help ag professionals aggregate the massive amount of data they collect and consume it in a way that enables them to answer questions and make informed decisions.

Given the increasing interest in autonomy and automation for farming applications, NVIDIA also creates solutions that focus on the entire lifecycle of how robots are designed, developed, deployed and managed. The company’s robust computing platform combines hardware and software, and is equipped to handle the full range of robotics needs.

“There are two main computers that you need to design for autonomous applications,” Goel says. “The first is the AI Factory or the big computer. This is something that can live in the cloud or can live in the data center. Then there's a second computer, NVIDIA Jetson, which is what lives on the machine. This is what goes on a tractor. This is what goes into a drone. This is what’s used in a cart that moves around the field.”

After the computers, the next step was to create workflows that could train AI models using the data the computers collected from the machines. NVIDIA developed powerful platforms to do just that. These platforms enable users to simulate how a robot will behave. For example, an autonomous sprayer has sensors that collect image data on the crop and weeds. The data then needs to be fed into an AI algorithm to determine – almost instantaneously –whether it is looking at a weed it needs to spray or a crop that needs to be left alone. If it is a weed, the machine needs to know where to find it before adjusting the nozzle and delivering a targeted herbicide application. The workflows ensure the sprayer operates as intended.

On the big computer side, there are two platforms that power the workflows: NVIDIA AI Enterprise and NVIDIA Omniverse. On the small computer side, Jetson uses NVIDIA Isaac to deploy intelligence on the actual robots. Together, Goel says, these computers and platforms handle the training, maintaining, developing and deployment of autonomous machines in agriculture. Having a comprehensive solution of this caliber has been a long time coming.

“When we started looking at autonomy in automation, we realized that the computers that were available in the industry today were not sufficient for it,” Goel says. “These machines need to do a lot of things. They need to connect to a lot of sensors, from cameras to radar ultrasonics, and these sensors create a lot of data.”

“There’s no way that you can send this data over to the cloud, so first, you need to have a way to ingest this data,” he continues. “The second thing you need to do is process this data, and data processing means two things. One is converting it into a format that can be used for algorithms. The second is figuring out what algorithm to use on the data. So, when we looked at the entire pipeline, we realized that we needed to build a specific computer that can solve all these tasks from end to end.”

That computer is Jetson. It has certainly been an overwhelming success for the ag industry, but it has been useful elsewhere, too. Goel says that pretty much every company trying to do anything with AI is using Jetson as the brain for the computer because “it is purpose-built for autonomy and automation.” Having a great computer, however, is only one piece. Hardware alone will only take the farming sector so far. That’s why NVIDIA continues to innovate on all fronts.

“Our latest family of Jetson is powered by our Orin system-on-chip (SoC) architecture, and it has seven different processors bundled into that one chip,” Goel says. “It can do camera processing. It can do AI processing. It has a CPU for general purpose computing. It has video processing, video encoding and video decoding. It has a dedicated deep learning accelerator.”

“We packed in all these different things that are needed, but I often tell people that the Ferrari by itself is not going to win the Formula One,” he continues. “You need a driver. So, we built great hardware that has everything needed to solve problems, but we didn't stop there. We also build the software layer on top of it that allows you to extract the value of this hardware.”

Overall, this is just the beginning for NVIDIA. While Jetson and the other solutions are exciting and have had a positive impact, there’s a lot more to be done. The company is already looking at new ways to help its customers access better data, make better decisions and solve their most pressing problems. Much of what NVIDIA is doing now is focused on crunching information from huge databases in order to extract the insights farmers need to minimize chemical inputs, forecast yield and harvest efficiently without additional labor.

“The robots get a lot of attention, but I think there are many hard computing problems in the lifecycle of agriculture,” Goel says. “Combining what we have on the computer side with Jetson with data science and generative AI that’s able to process this large amount of data that gets collected—it’s going to really unlock a tremendous amount of value for the farmers.”

Related Content:

Click here for more Industry News.