Inside the XPRIZE Pressure Cooker

AT THE RANGER STATION ON THE EDGE OF SINGAPORE’S CENTRAL CATCHMENT NATURE RESERVE, 21 LUNCH ORDERS HAVE BEEN SEPARATED INTO TIDY STACKS: pad thai, fried rice with pineapple and prawns, mango salad, and Vietnamese iced coffee that is sweating copiously onto the plastic folding table.

Our team trundles into the station and settles onto the rough flagstone floor with their lunches. There are botanists, birders, conservationists and engineers of every stripe with specialties as diverse as their lunch orders. Soon the sounds of plastic cutlery being unwrapped and water bottles uncapped add to the background noise of insects screaming in the forest. A coil of shiny silver ductwork shivers as a portable air conditioning unit struggles to suck the humidity from the rainforest and breathe a bit of cool relief into the packed room.

But 10 minutes’ walk through the forest and down a narrow service road, mechanical engineer Juliet O’Riordan is still at our group’s test site. Both lunch and relief can wait; she is searching for a suitable branch upon which to hang our team’s success in the XPRIZE Rainforest competition. She’s feeling immense pressure to get it right. The team has traveled across the world to be here, her teammates are depending on her, and a documentary film crew is tracking her every move.

XPRIZE Rainforest is a race to discover, understand and preserve the health of the world’s richest, densest ecosystems using autonomous technology to catalogue the biodiversity found there. To win the competition’s $10 million purse, teams try to develop the quickest, most complete and most efficient means of identifying species at every level of the rainforest, canopy to floor.

Tropical forests are among the most biodiverse places on earth, and they’re quickly disappearing.

THE XPRIZE RAINFOREST CHALLENGE IS MORE THAN AN ACADEMIC EXERCISE. Tropical forests are among the most biodiverse places on earth, and they’re quickly disappearing—cleared to make room for agriculture, or mines, or housing developments, or other applications judged to have more value than the standing forest.

But the services that intact forests provide to society are poorly understood, hard to quantify and nearly impossible to assign monetary value to, and that makes it difficult to make the case for conservation.

But we’re trying to change that here at the XPRIZE Rainforest competition semifinals in Singapore. By collecting and analyzing data about the species found here, we hope to offer new insights into the complicated, interdependent relationships that exist in the rainforest and the wealth of benefits they provide.

To advance to the finals, we have to demonstrate our intent and our ability to generate these insights—and that depends wholly on our successful collection of sound recording, photographs, and biological data collected from the rainforest.

We have only 24 hours to collect data, and 48 hours to transform the data into insights.

WE’RE COMPETING AGAINST 12 OTHER TEAMS FROM AROUND THE WORLD. The XPRIZE judges tell us that the approaches are wildly different across the semifinalist pool, but won’t give details; we have however, heard rumors of robotic dogs that fetch images of native foliage and of new techniques for collecting environmental DNA from the forest floor.

Our team, the Blue Devil Forest Divers, is led by Duke University Associate Professor of Electrical and Computer Engineering Martin Brooke and Duke Professor of Conservation Ecology Stuart Pimm and takes a drone-based data collection approach. We’ve descended upon Singapore’s Windsor Nature Park with a fleet of 22 drones ranging in size from an eastern box turtle to a giant tortoise.

Most of them are ANAFIs, small vehicles manufactured by the Parrot company, which are packed into hard-shell suitcases the color of nonpareil mint drops. In two table-sized crates, though, wrapped like mummies in swathes of stiff brown hex paper, are the components of the large drones that we call the Mother and the Relay. These drones are unique, both to the competition and to rainforest conservation. They’ve been designed and built from scratch by Duke University students enrolled in Brooke’s Ocean Engineering and Rainforest Engineering classes, each new wave of students squeezing a bit more utility and performance out of the drones every semester.

Mother is a hybrid drone that uses a gasoline-powered engine to extend the range that her battery affords. She’s essentially a school bus, ferrying the little battery-powered drones deep into the forest, much farther than they could fly on their own. The other large drone—the Relay—is a hybrid-fueled communications way station that hovers, tethered, above the canopy to listen for messages from Mother and communicate them to researchers on the ground.

The scheme allows our team to efficiently canvass more of the forest than if we were depending only upon battery power. In the semifinals, the test plot is within throwing distance of a service road, so the large drones aren’t even necessary— the small drones can reach the center of the plot under their own power. But this round of competition is an excellent opportunity for our team to learn how to ship, reassemble and safely fuel the large drones in addition to flying them.

The data collection period begins tomorrow morning.

SINGAPORE IS A DENSELY POPULATED CITY-STATE, AND ITS AIRSPACE IS STRICTLY CONTROLLED. The large drones are therefore subject to extensive airworthiness checks: the ability to display signal strength and failure warnings for each fuel source, to give flight telemetry, to perform a controlled four-minute flight with changing orientation, to communicate an uninterrupted video screen and to land within a designated area.

The engineers assigned to the large drones soldier through dozens of attempts and recalibrations as the sun rises higher and the open field begins to stream. Mother’s steady chirps are punctuated by the noise of two roosters fighting atop a wall that separates the test field from an adjacent housing block.

Just when it seems the overheated students might give up, Mother suddenly pops into the sky like a kernel of corn. She climbs to around 20 feet and hovers like a dragonfly above the cheering team.

But the sweet victory is short-lived. Mother falls from the sky on her next flight, coming apart neatly when she hits the ground as she was designed to do. Her motor and electronics can be salvaged this way, and her frame rebuilt back home.

The engineers attribute the crash to the PID tuning—the software parameters that affect how the flight controller responds to commands—and all agree it could have been corrected with a bit more practice. The wind, or even the weight of the additional zip ties securing the motor to the frame, might have thrown the drone out of balance.

But a thorough critique will have to wait.

It’s noon, and the clock is now ticking.

THE LARGE DRONE PILOTS, relieved of their vehicles, start registering flights on the small drones. A trio of University of Wisconsin-Madison botanists, led by orchid expert Stephanie Pimm Lyon, hovers over the pilots’ shoulders, directing them to areas of possible interest to photograph. Images of leaves and vines begin pouring in almost immediately, to their gratification. But other groups are waiting on data from two yet-to-be-deployed traps: a recorder that captures birdsong and a sticky insect lure.

We have two Tupperware containers full of insect bait. One contains slumping chunks of mango and banana whose edges are slowly dissolving into soup. Another is packed with balls of raw pork that have taken on a greenish cast and smell like fish guts. Liquid as dark as swamp water pools at the bottom. These slurries will be irresistible to flies and beetles, sending them careening into the sticky sides of the traps.

Duke biomedical engineering lab administrator Maggie Gatongi and J.A. Jones Distinguished Professor of Electrical and Computer Engineering Nan Jokerst have set up a mobile DNA barcoding lab in the ranger station. They wait with small, high-powered microscopes to photograph the stuck insects—the first step of 14 in their protocol, which will end in a foolproof species ID. In their lab coats and sterile gloves, they double-check and triple-check their procedures as they wait for the first batch of captured insects to arrive.

Which brings us back to O’Riordan’s hunt for the perfect branch, and the corresponding challenge: there don’t seem to be any good branches, let alone perfect ones, from which to hang an insect trap.

She and her co-pilot, ECE alum Alex Xu, send numerous drones scouting through the test plot. The vegetation is thick and weedy, and the long, whip-thin branches frequently tangle in the drones’ propellors and send them tumbling to the forest floor. They’ve been lucky so far—if they crashed instead into a treetop or other inaccessible area, the drone couldn’t be retrieved until the end of the competition. In that scenario, the team would lose not only their drone, but also the trap or data it was carrying as well.

By the time darkness falls…

ALMOST AS QUICKLY AS A SWITCH FLIPPING, HERE NEAR THE EARTH’S EQUATOR – THE PILOTS HAVE CAPTURED 238 IMAGES OF THE RAINFOREST’S PLANT SPECIES. One trap, a bioacoustics recorder, is now hanging on a thick vine, where it will remain overnight to eavesdrop on the early-morning conversations of the waking birds.

Despite this success, pressure is still building. “It is essential that in the morning we place an insect trap inside the valid area,” Brooke urges in a message to the group via WhatsApp. “We’ll take it steady and make it happen,” Xu responds in the thread, to a chorus of heart emojis that pop up from his teammates.

A group of pilots return to the test site at 4 AM after a quick nap. They outfit the drones with thermal cameras, hoping to spot the warm auras of macaques or wild boar through the cool blue canopy—a strategy that had worked well both in Durham and in the jungle of La Selva Biological Station in Costa Rica, where the team tested its approach last summer.

The sun rises on a still rainforest, illuminating only plants and foliage.

It’s 9 AM, and only three hours are left in the collection period.

BROOKE URGES O’RIORDAN TO RETRIEVE THE SOUND RECORDER, BUT SHE AND XU ARE STILL SCANNING ENDLESSLY FOR PERFORATIONS IN THE DENSE FOLIAGE, searching for a place to safely drop an insect trap. We only need one to prove that our DNA barcoding technique works.

At this critical moment, a safety officer makes an unwelcome announcement: there’s a Category 1 lightning storm approaching, and all flights are grounded until the threat passes. Only six people may remain beneath the tent, so the rest of us scurry up the hill to the ranger station, through the sweltering jungle.

Singapore’s Central Catchment Reserve is essentially an inner-city woodlot. It’s different from mature rainforests, where old emergent trees compose the forest’s crown. Below, the canopy is so dense that it creates shade and a dark, sparse forest floor. Here in Singapore, emergent trees are few and far between, and the canopy below allows much more sunlight to filter down to the forest floor, creating conditions where small plants and shrubs thrive.

In the mature rainforest, there were innumerable unobstructed branches for O’Riordan to aim for. In Singapore, there are none.

The lightning warning is lifted just before 11 AM.

One hour remains in which to collect and retrieve data from the rainforest.

THE DRONE PILOTS IN THE RANGER STATION SPRING BACK TOWARD THE TEST SITE, where O’Riordan and Xu are already initiating their flight plan. Botanist Lilly Johnson is just behind them, carrying a length of red braid made from strips of a plastic biohazard bag.

No one spares the time to ask questions about this makeshift lanyard; every eye is on O’Riordan as she assembles another insect trap beneath the red tent.

One of the most remarkable things about this team is the latitude that Brooke and Pimm have allowed the students and recent graduates who comprise the team to plan for themselves and make their own decisions. The faculty members provide vision and they provide resources: expert entomologists, large trucks, drone transponders and controllers, portable DNA sequencers, mountains of Malaysian takeout. But apart from assisting with one particularly tricky bit of soldering, Brooke has, until this morning, remained hands-off. It speaks to the pressure the entire team is under that this dynamic has now changed.

“Try to recover the sound recorder,” Brooke urges her.

As the two confer, Johnson brings a small drone to the table. From a sheet of yellow fly paper, she cuts long strips with the speed and precision of a sous chef and affixes them to the drone with the braided lanyard and runs with it to the top of the test site.

With just 36 minutes left in the collection period, our pilots have split into two factions, each frantically trying to engineer a win for the entire team.

At the top of the hill, one group is making a last-ditch attempt to collect an insect sample by dragging a long loop of fly paper through the canopy. At the bottom of the hill, Xu stands in the middle of the narrow service road, lofting a drone above his head, searching for the strongest signal to support O’Riordan’s recovery efforts.

Midflight, O’Riordan and Xu lose their signal and must abandon the attempt. It’s unclear whether the competing flight at the top of the hill is interfering.

Brooke charges up the hill to intervene, and O’Riordan uses the distraction to quietly launch an attempt to retrieve the sound recorder. She knows the spot she wants and aligns herself to the opening she sees; her drone buzzes alongside her. It takes only a minute for the Parrot to reach the sound recorder, and it snaps up the device with a strong magnet and turns back toward the service road, carrying its payload.

Everyone has pivoted their attention back to O’Riordan—our whole team, the safety officers, the XPRIZE staffers and the film crew with their cameras rolling—so everyone hears when the drone starts to go down, making a sound like a weed whacker as it becomes entangled in the undergrowth. The drone and the recorder fall so close to the road that we can see the vehicle’s propellers still fighting valiantly to turn from amidst a nest of vines. Its lights blink orange in a silent SOS.

22 minutes remain on the clock.

WE HAVE NO INSECTS. WE HAVE NO SOUND DATA. O’Riordan sinks briefly to the ground, folding in on herself like a paper crane.

The drone is almost close enough to touch and the urge to simply pluck it from the brush is strong, but the rules of the competition prevent human beings from entering the plot.

One student commandeers the communications team’s tripod and telescopes its legs to their longest. Another finds a fallen branch taller than he is. Working in tandem, they use the two implements like giant chopsticks and reach for the sound trap. Their jostling separates the drone from the recorder and the precious data slips perilously close to the forest floor.

There is absolutely no way this will work.

But they somehow flip the device out of the thicket and onto the dirt between the road and the brush, where it’s fair game.

A doctoral student in Pimm’s group snatches the sound recorder and scampers away to download the data, quick as a macaque with a contraband protein bar.

The flypaper, meanwhile, is dotted with tiny specks of dark matter that appear to be nothing, but an XPRIZE staffer documents the time and location of the collection anyway, just as the clock runs out. Jokerst and Gatongi examine the loop of sticky tape under the compact microscopes they’ve brought from Duke, methodically zooming in on one amorphous blob after another.

Then, there it is, swimming up through the microscope’s light and into sharp focus against the yellow tape: a segmented abdomen and a pair of translucent wings.

The team’s allotted time at the ranger station has ended.

BUT THE NATIONAL UNIVERSITY OF SINGAPORE HAS OFFERED SPACE IN ONE OF ITS WET LABS FOR THE COMPETITORS TO PERFORM DNA AND eDNA ANALYSIS, so Gatongi and Jokerst shift into high gear to prepare our single sample for processing.

The collection period ended at full boil, but Gatongi is an island of cool. In the NUS wet lab, she is perfectly at home; she has taught hundreds of students to perform the polymerase chain reaction, or PCR, protocol that she now initiates.

She and Jokerst turn their results over to a recent biomedical engineering alum at 1 AM. It takes another three hours for him to run the DNA sequence on a base caller and name the insect: Scirtothrips dorsalus, a species of invasive thrip. The DNA ID confirms the visual one made by renowned entomologist Roger Kitching.

To the terabytes of data collected by the team, this small, precious bit of information is added. It’s proof that the DNA barcoding technique works and could be applied on a bigger scale in the finals of the competition.

Brooke emails the team’s final report to XPRIZE. There are 63 unique taxa the team identified in the competition using photo, audio and DNA identification methods. The species richness of the rainforest plot has been described, and heat tree diagrams show how each individual species fits into the whole. From the chaos of the competition, an ordered snapshot of the rainforest’s biodiversity has emerged.

The team disperses to all corners of the globe — to South Africa, to study the fragmentation of elephant populations; to new jobs in the forestry service and existing ones at Microsoft; to internships at Apple; to deserved vacations. But they still steal time to pore over the footage of Mother’s crash, looking for clues to inform her next incarnation and make her more resilient, and to troll iNaturalist after hours to see how the other teams fared at their plant identification task and to work out the odds of our team making it to the finals.

Several weeks later, at the International Congress for Conservation Biology in Kigali, Rwanda, the XPRIZE Foundation announces the finalists.

THE SIX TEAMS WITH THE HIGHEST SPECIES COUNT ADVANCE. The Blue Devil Forest Divers are not among them.

The shock of the news is like being blanched in an ice water bath. But the team rallies to thank one another and prepares to pass the baton. Only four undergraduate students on the team will return to Duke in the fall.

Brooke has competed in five different XPRIZE competitions now and says that managing the constant turnover is the biggest challenge—but Duke’s participation isn’t driven by winning, it’s driven by the university’s educational mission. More than 200 students have contributed their time and effort to this project, and many more will have the opportunity in the coming years. They learn and apply technical skills, but they also learn to manage complex logistical challenges, team dynamics and their own expectations.

Brooke is teaching Ocean Engineering in the fall and students in the class will be rebuilding the large drones and outfitting them for long-range ocean device delivery. Brooke imagines that there might be an ocean-focused XPRIZE competition in the near future and if there is, Duke will take a swing at it.

“We did everything we set out to do,” says Brooke. “We had great opportunities for students in classes to work on authentic projects, and we got to test many of our technologies in rainforest environments. Things went better than I had hoped for, in many ways. And we had an absolute blast.”

“We did everything we set out to do. We had great opportunities for students in classes to work on authentic projects, and we got to test many of our technologies in rainforest environments. Things went better than I had hoped for, in many ways. And we had an absolute blast.”

MARTIN BROOKE
Associate Professor of Electrical & Computer Engineering