Short Wave - Grape Growers' Next Collaborators? Robots

Episode Date: December 10, 2024

If you crossed WALL-E with a floor lamp, it might look a little like the PhytoPatholoBot. These robots aren't roving through space or decorating a living room — they're monitoring the stems, leaves and fruit of Cornell AgriTech's vineyards, rolling down each row and scanning for mildew.In this episode, host Emily Kwong and producer Hannah Chinn take a trip to Cornell to check out these new robots. How do they work? How effective are they? And what do local grape farmers – and neighbors – think about them? Interested in more robotics stories? Email us at [e-mail]. We'd love to hear from you!See pcm.adswizz.com for information about our collection and use of personal data for sponsorship and to manage your podcast sponsorship preferences.NPR Privacy Policy

You're listening to Shortwave from NPR. Hey, shortwavers, Emily Kwong here with Shortwave producer Hannah Chin. And I see you've brought a friend. I have. This is the sound of a robot rolling down a row of grapes and diagnosing the problems in the grape prop. Drives by itself. And while it kind of traverses through the vineyard, it takes the images of the... grapevine canopy and start to use the input image to identify the disease infection.

So this is Yu Jiang. He's an assistant professor of systems engineering and data analytics for specialty crops at Cornell Agritech. Very long title. Super long title. But what it means is that he's the nationwide expert on grape robotics. And he and his colleagues at Cornell Agritechac, which is in upstate New York, where I went to interview him, built this grape scanning robot. I'm going to send you a picture of that robot now, Emily. and we can make it the cover image for this episode so our listeners can look at it too. Okay. Oh, look at her go.

Yeah, it's like if Wally, like from Pixar had a baby with a floor lamp. It's how I feel that this would look. Yeah, exactly. It's around five, maybe a little over five feet tall. It's got wheels on the bottom and a camera on the side and then this kind of computer set up on the back.

For data crunching, I imagine. Well, I want to learn more about this robot. Like, does it have a name? This one specifically does not have a name, but as a whole, all of these robots are called phytopatholobots, or PPPB, for short. PPPB, phytopatholabotabot. But also I think from the engineering perspective, I love the acronym, because it's short for parts per billing, means super accurate. But what are they doing? Why are they roving these rows with cameras of all things? Are we just setting ourselves up to be replaced by

robo farmers? I mean, Emily, I can't give it all away. Let's just wait until after the break. Fair. Okay. So today on the show, the infections imperiling our grapes and the robots that could help farmers fight back. You're listening to Shortwave, the science podcast from NPR. Han, I am obsessed with robots. I actually dressed up as Eve for Halloween with my partner Duncan as Wally. That's so cute. And I really want to know all about Wally's cousins.

This robot is so cute. Let's first talk about the problem the robot is solving for. it has to do with crops and crop disease. Yes. So there's actually two types of diseases that growers are trying to tackle here. There's powdery mildew and downy mildew. These are two of the most devastating, I guess, diseases found in grapes. This is Philip Lujan.

He's a plant pathologist and assistant professor at New Mexico State University. And he's responsible for disease diagnostics for crops really throughout the whole state. He says these two diseases reduce the shift. sugar content in these grapes. They make the grapes bitter and lower quality. Even as little as 3 to 5% of mildew infection in a vineyard can make the entire grape crop unusable for wine production. An infection of this extent, it's been described by researchers as giving the wine an oily, viscous mouth feel. Oh, terrible. Okay. What is causing this mildew? Isn't mildew usually related to dampness? Yeah. So these two diseases, Downey Mildo and Poutary Mildo,

They're caused by two totally different organisms. Oh, like these infections are not related to each other. No, not at all. Philip told me one is caused by a fungus, and the other one is caused by a fungal organism. Kind of like a water mold. Okay. But the effect is loki similar.

The name is very descriptive of powdery mildew. It just looks like some white powder on the top of the leaves. And if you look at Downy Mildew infected leaves, on the top of the leaves, you're going to get sort of a yellowing color, sort of like an oil spot. You can tell which leaves are affected because they're like flaky or dirty looking. Powdery mildew creates cracks or lesions in the grapes, and downy mildew makes them more shrively and discolored. That's a sad state for these poor plants. Okay.

And these types of mildew, they aren't just gross looking. They're also really bad for the plants and their grapes. It reduces photosynthesis. It also reduces, like, crop yields as well. It can also affect grape quality. And, Emily, once that mildew sets in, it's really hard to shake. What do you mean? So Philip lives in New Mexico, where Downey Mildo isn't as much of an issue, right?

Because it's so dry. And that's true also in California, which is where the majority of U.S. grapes are produced. But he says that powdery mildew, the fungus spread disease, is capable of spreading anywhere. It can even overwinter, which means it survives through the cold season and then infects new leaves when it rains in the spring. When you get that rainfall, that rain splash, they sort of start the growing process of the fungus on the leaf. It develops spores, and then those spores splash onto the new leaf in the spring. So that's sort of that disease cycle. That is so pernicious.

I mean, it shows the tenacity of this fungus, respect. But also, these points are. Or grapes. Yeah. But grape growers are constantly working to combat this threat. Growers often spray 10 to 15 times fungicides per year. And so that's very concerning because, you know, these pathogens are always evolving. And so they're becoming more and more resistant to these sprays because they're spraying so often. So it's like when humans take antibiotics, the pathogens in our body can become more resistant in response to antibiotics. And then we have to take strong. stronger ones more often and it can become this vicious cycle. That's exactly the analogy that Philip used to explain this to me. He said these grape diseases are developing resistance faster than we're developing treatments.

And so for that reason, we need to diversify the methods that we're using. So it takes a really, really long time for a new fungicide to come out into the market. And so, you know, it could take 10, 15, 20 years to develop that. And so, you know, these fungi are becoming resistant at a much quicker rate. So if we don't want to use so many fungicides, the best thing to do is detect this mildew as early as possible, way before it becomes a problem. Right. Treat it at the source. How do people go about doing that? So it's a really hard job, and it's usually done by skilled workers, right? people who are trained on how to spot these early stages of mildew.

But they also have to be out in the field where there's no shade. They're walking down every row and they're looking at each plant. And Yu Jiang from Cornell, he told me it can be really hard on those workers. I know many people are going to see, oh yeah, I'm going to enjoy the sunlight. But imagine you're going to work eight hours per day for the whole year of your job, just being in the field. Walking up and down the grape road. Yes. And you need to check all these diseases we just look at around.

You change our mind and you lose your passion. Okay, so is this the job for these uniquely equipped grape robots? This is a job for the robots. They go up and down the rows and they scan all the grape vines, like I mentioned earlier, and they take hundreds of thousands of pictures to detect the presence of mildew. But Emily, even among all these things, hundreds of thousands of pictures, there's a problem. And it has to do with lighting. Hmm. Sometimes, you know, the image you're taking in the morning would look very different than

in the evening, right? And that's actually caused a lot of problems later on for the AI models. Oh, they're using AI. Okay. So this is like a data problem. Like the training data used for these robots says like disease looks one way, but out in the world, you're faced with all the variables of lighting of any photographer where it might not look the same? Yeah, exactly. And especially many of these disease symptoms are based on the color. So that illumination is going to be the biggest challenge for the color consistency or the image quality consistency throughout the day.

So to solve this problem of color consistency, they use flash for every single photo. It's like the grape robot version of a ringlight. Nice. Well, they want that lighting to be consistent. So these robots, they roll down the row, they have their flash on. Yep. And they're taking a lot of photos. Two photos per second.

They just sound like my parents. You see current the robot kind of taking the picture. As I mentioned, once the picture is taken, it will transfer directly to the computer inside the small box. The computer then analyzes those photos and compares them to existing photos of disease plants to make a diagnosis. And then the AI model will do the... inference, basically calculate how many pixels representing the infections versus how many pixels representing the canopy. And then after the scanning, it will generate the map to showing the infection severity

at each geographic location. Okay. So basically what he's saying is that it creates this huge map of the grape field, right? Indicating both where the mildew is and how bad it is. This is very cool. Combined so much new technology from robotics, AI. How accurate, though, is this method? It is super accurate.

So we just kind of finished a study to further verify its accuracy in terms of a disease detection for various kind of disease in different states like California, New York, West Virginia. And we found overall we achieve around, you know, over 90% of the accuracy for Donny Moldew, pottery mildew and associated virus. Okay, 90% accuracy. That's pretty good. And what happens when mildew is detected? When they find it, they still have to treat it, right?

So you told me that eventually they're hoping to equip these robots with sprays so that they can treat the infections directly out in the field when they find them. That's cool. Yeah. And he showed me the next generation of these robots. They use similar technology to a Tesla, so they're actually self-driving. and they have cameras on both sides so they can scan twice as fast as the original robot. You know, I said these robots were the cousin of Wally, but they're really the cousin of Eve.

They're out here scanning plants to, like, protect nature. They're little eco-warriers or something. Absolutely. But he's working on other robots, too. Most of them are stored in a huge barn when they're not working. Some of them are small, and they roll through the field and pull weeds. Some of them are huge, like eight or nine feet tall, and they zap mildew in the middle of the night with UV light. But all of them are meant to make human grape farmers jobs easier.

And Emily, you know that not everyone is as excited about the robots as he is, right? These big robots that treat mildew in the middle of the night, he told me that one time the neighbors called the police because they didn't know what was going on. People just want to make sure everything is right. The operators can't have explained all the situation. But I think now people really understand, especially with years of this, you know, extension and all the region. activities people understand. Yeah, I'd be concerned if I saw a nine-foot-tall robot shining a UV light in the field next to me in midnight.

I would think I was getting abducted. Yeah, like, is it aliens? Is it what's happening? And the other concern that people have that you mentioned to me is like, what if these robots take human jobs, right? Are they going to replace us? Many people think, oh yeah, robots going to replace human, no. In my opinion, going to just a compliment.

give us more capacity, more power that we can do we more than Earth as an individual in the future. So he's saying these robots, if they were to scale in grape farms, in places that need them, they would have the potential to help the farmers with their work. Yeah, exactly. And he's working with pilot groups, with farmers around Cornell, asking how they can make this technology work for them. Like, the ultimate goal of these robots isn't to take our job. jobs. It's just to help us do them differently. Hannah Chin, thank you for telling us all about these amazing robots.

Yeah, thanks for having me. This episode was produced by Rachel Carlson, and it was edited by our showrunner, Rebecca Ramirez. Hannah Chin obviously did the fabulous reporting. Tyler Jones, check the facts. Zote Vangenhofen with the audio engineer. Beth Donovan is our senior director, and Colin Campbell is our senior vice president of podcasting strategy. I'm Emily Kwong. And I'm Hannah Chin.

Thanks for listening to Shortwave. from NPR.