This High Schooler Invented an A.I.-Powered Trap That Zaps Invasive Lanternflies

A New Jersey native, Selina Zhang is no stranger to the spotted lanternfly, an invasive species that has ravaged the Garden State’s local agricultural industry for years. The now 18-year-old first noticed the pest when visiting a market near her hometown of Annandale in grade school, intrigued by a cluster of the colorful, moth-like bugs. But the spotted lanternfly’s alluring looks, with bright red underwings peeking out from black polka-dotted forewings, can be deceiving.  

Classified as a plant-hopper, the insect jumps from organism to organism to feed on their sap, affecting over 70 species. In its wake, the plants become stressed, which increases their susceptibility to disease and attack. As the spotted lanternfly ingests high amounts of sugar, it excretes honeydew, a sticky liquid waste that accumulates on trees, impeding photosynthesis and encouraging black sooty mold to sprout.  

The spotted lanternfly, native to China, likely hitchhiked its way to the United States on a stone shipment in 2012 as mere egg masses. In 2014, a group of the plant-hoppers was found covering a wooden area in Berks County, Pennsylvania, marking the first in a long line of spotted lanternfly infestations. The invasive species has now plagued 17 states.

“As I got older, I wanted to take concrete action,” says Zhang. “I wanted to build an innovative solution that took into account my personal perspective and existing research to target this bug in ways we haven’t before.” 

The most common forms of management are insecticides and sticky bands, a strip of tacky material wrapped around a tree trunk to trap pests. Both have ecologically destructive side effects. Insecticides, including neonicotinoids, leave behind harmful residues that are toxic to crucial pollinators, such as honeybees, and sticky bands can inadvertently trap birds, small animals and other insects in their adhesive grip.

To sidestep these negative consequences, Zhang drew inspiration from chess boards and “Dance Dance Revolution.” Combined with weeks of extensive field observation, deep algorithmic programming and an umbrella seized from her family’s patio, the teenager built ArTreeficial, a solar-powered, self-cleaning, artificial-intelligence-driven “tree” that entices the spotted lanternfly and eliminates the bug using an electronic mesh.  

“In essence, ArTreeficial is a supersized mousetrap for combating the invasive spotted lanternfly,” says Zhang. “The specific ‘cheese’ is the main component of tree of heaven [Ailanthus altissima], which the spotted lanternfly is very drawn to.”  

On top of conducting an extensive literature review of existing research, Zhang carefully observed the spotted lanternfly in the wild for weeks, tracking how the bugs grew over time, where they tended to cluster and how they navigated unfamiliar structures. She took over 500 photos of the insects, which she logged in a personal database.  

“The main component of conducting the field observation was to get a better idea of who the enemy is. If I don’t understand how they behave, I can’t make an effective enough trap,” says Zhang.  

Selina Zhang

Watch on

Zhang knew she wanted to create a tree-like structure that mimicked the tree of heaven, a known host plant of the spotted lanternfly, as the primary lure for her trap. So, she uprooted the umbrella from her family’s patio and got to work.

First, Zhang used ultrasound to emit an incense made from the essence of tree of heaven from ArTreeficial’s trunk to entice the pests. The next step was to design an electric net for the synthetic tree, which would use machine learning to shock spotted lanternflies once they landed on the tree. In her first iteration of ArTreeficial, Zhang wrapped a charged net around the trunk, but she ran into two problems. First, the net was a safety hazard; no protective barrier shielded the electricity. And, secondly, hopping spotted lanternflies could potentially evade the single net. Zhang turned back to her field observations: The spotted lanternflies, she recalled, took long, delicate strides, often extending their spindly legs and sometimes wiggling them in the air before placing them down.  

With this behavior in mind, Zhang upgraded to a double-layered net, draping a large diamond-shaped outer net with a tighter inner mesh onto ArTreeficial. This would allow for the topmost layer to act as a safety guard—and for the spotted lanternfly to step through the small holes of the upper mesh and make contact with the inner mesh. 

“For field observations, you have to accumulate a lot of observations over time, and it can sometimes be uneventful,” says Zhang. “But it was also an important test of patience, because you get rewarded by these really interesting things.”  

The double-net structure needed to conform to ArTreeficial’s trunk. But striking the perfect distance between the two nets was tricky. If they’re too close together when electricity is activated, the whole thing will short-circuit. If they’re too far apart, a spotted lanternfly’s leg won’t reach the inner net, failing to complete the circuit. Zhang used rubber tubing to weave the nets together at a fixed distance.

Then came the artificial intelligence. Zhang used her database of photos to inform and program her A.I. model. 

“I didn’t want to just rely on internet pictures, which are often designed to focus on specific aspects of the spotted lanternfly, such as its vibrant red wings, or clusters of them together,” says Zhang. “Part of making a good A.I. model is having good data to train it with. My field observations and pictures give the model a more comprehensive understanding of what it might encounter.” 

Similar to a chess board, Zhang’s nets are divided into multiple squares going up and down the structure. Essentially, when a spotted lanternfly steps on the inner net, the A.I. model is triggered. Electricity courses through that particular section, while the rest remain inactive, and the insect is zapped—like a reverse, lethal “Dance Dance Revolution.” In the music video game “Dance Dance Revolution,” players stand on a tile grid with four directional arrows that correspond to dance steps on the screen. To signal a move in the larger routine, the coordinating arrow tile lights up, awaiting a resolute stomp from the dancer. Zhang’s project is the same idea in principle, just in a different order, with the electronic signal activating after the step.

“It was amazing to see it working,” says Zhang. “It was such a lightbulb moment when I pushed the internet on, tested [the net] with a screwdriver and it gave a spark. Seeing it come to life was an incredible feeling.” 

Zhang is one of 40 finalists in the Regeneron Science Talent Search, the country’s oldest and most prestigious science and math competition for high school seniors. Finalist projects range from revolutionary engineering feats, such as low-cost, universal 3D-printed prosthetics for amputees, to space science breakthroughs, such as analyses of microorganisms on comets and asteroids to hypothesize the establishment of life on Earth. Zhang’s invention falls in the contest’s environmental science category, but it clearly crosses fields.  

“The project uses A.I., it uses chemistry, it’s dealing with climate change and solar power. It’s a whole amalgam of the interdisciplinary nature of science and engineering in this project,” says Maya Ajmera, the president and CEO of Society for Science, which hosts the talent search. “That’s what makes it stand out for me.”  

Zhang hopes to take even more photos of spotted lanternflies in the spring, when the insects begin to hatch, to better train her A.I. model to recognize the pest at all stages of development. Ajmera hopes that Zhang will be able to patent her product. 

“I feel like there’s A.I. for good. And what Selina’s project does is A.I. for good, for humanity, for making the world a better place,” says Ajmera. 

Julie Urban, an entomologist and spotted lanternfly expert at Pennsylvania State University, says that Zhang’s A.I. approach is “incredibly innovative” and could be useful in a multitude of environments.

“If she’s going through the trouble of using A.I. to identify the lanternfly, that in and of itself could be useful, separate from the structure of the tree. She could test the netting to see if it works on other structures or crops—for example, grapes—where you don’t want lanternfly to be,” adds Urban.  

However, the spotted lanternfly is notoriously evasive, with few known predators and even fewer confirmed lures, according to Urban. The tree of heaven, for instance, is a preferred home from the fly’s native range, but it’s not always guaranteed to land on that plant. 

“The thing we’ve been having a really hard time with is that there really are no lures that work for spotted lanternfly. There’s not going to be one silver bullet that works across the board. There’s no panacea,” Urban says. 

From here, Zhang plans to analyze how the essence from dead spotted lanternflies can improve her lure, and to test out her invention in the field.  

Currently, ArTreeficial costs around $200 to manufacture. Zhang, already in communication with local farmers, is hoping that with mass manufacturing and partnerships with local companies, she’ll be able to lower costs over time. That way, she can deploy multiple synthetic trees at once and integrate them into a network, or “ArTreeForest,” as she calls it. Once she proves her system is effective with spotted lanternflies, Zhang aims to extend it to other invasive species. 

“I would love to develop ArTreeficial as a general paradigm, not just for spotted lanternflies, but as a general approach that we can use toward technologies in pest control,” says Zhang. 

As an award-winning violin player who has performed at Carnegie Hall and a member of the Science Bowl and USA Biology Olympiad at North Hunterdon High School, Zhang has no shortage of talent and ideas. A passionate artist and writer, she even turned the spotted lanternfly into a menacing comic book character, hoping to raise awareness of the pest through a creative medium.

“Everywhere it goes, it’s spreading disaster,” says Zhang. “With my comics, I wanted people to better understand this local invasive species and its behavior.”  

As a high school senior, Zhang now has her sights set on college, specifically a university with strong biology and computer science programs so that she can continue to integrate artificial intelligence into biological research.  

“I want to demonstrate that we can use artificial resources as an essential tool for how we help protect our nature resources, especially in agriculture,” says Zhang. “As humans, it’s our responsibility to protect what we have and cultivate these resources responsibly to make sure that they’re here for us to enjoy, as well as generations after us.” 

Catherine Duncan | READ MORE

Catherine Duncan is an intern with Smithsonian magazine.