Science & Tech

Scholar makes robots that detect land mines

7 min read

On Oct. 10, 2005 — he remembers the date exactly — Thrishantha Nanayakkara was driving down a country road, headed for a science workshop at Jaffna Central College, a high school in the far north of Sri Lanka. The event was designed to distract potential child soldiers from the allure of war.

His cell phone rang. It was a government official, with a tip-off. “Turn back,” the caller said, in so many words, “or you will be killed.”

Nanayakkara, an artificial intelligence expert, said the phone call saved his life. Later that day, the school principal at Jaffna was shot to death.

The phone call was also the last straw for Thrish (as he likes to be called) — the last of many threats from Tamil rebels. Within days, his wife and two children moved from their home, and moved three times again in 2006.

That year, he said, “was horrible” — months of fear, hiding, and furtive living as Thrish scrambled for a job overseas.

In 2007, after a semester at the University of Texas, San Antonio, Thrish and his family slipped back into Sri Lanka for visas. He resigned his post as commissioner of the Sri Lanka Inventors Commission, took a sabbatical from teaching at the University of Moratuwa, and by July 1 — gratefully — was a visiting scholar at Harvard’s School of Engineering and Applied Sciences.

Just four years earlier, lured by the hope of permanent peace, Thrish had moved back to Sri Lanka after earning a doctoral degree in Japan and doing postdoctoral work in the United States. In 2003, Sri Lanka was enjoying a cease-fire in the two-decade war between the Tamil Tigers and government forces.

In December 2004, drama visited the Indian Ocean nation in a killer wave: the tsunami that killed 225,000 people in 11 countries, including tens of thousands in coastal Sri Lanka. Thrish and hundreds of other scholars volunteered for relief work, bagging bodies, distributing food, and absorbing a nation’s collective shock.

In 2005, came another killer wave: renewed fighting by Tamil rebels. They targeted scholars or politicians who addressed “burning problems,” said Thrish. “Whenever someone proposes a concrete solution, they kill them.”

The fateful cell phone call came, in part, because Thrish was confronting one of those burning problems.

He and a team of researchers were at work on an animal-robot team for hunting land mines. Sri Lanka is one of 50 countries affected worldwide by a legacy of antipersonnel mines. As many as 3 million lie buried in the island’s rich farmland or in forests nearby.

Thrish had already correctly mapped a test mine field outside a Sri Lankan army base, using a robotic device loosely strapped to a mongoose. In the space of one morning, the little mammal — light, agile, and equipped with an acute sense of smell — had sniffed out every buried mine.

But the Tamil guerillas were targeting ideals as well as ideas, said Thrish. Among those they killed was a Sri Lankan legal scholar and a foreign minister, he said. “Their ideals were different. That was their only criminal act.”

In the fall of 2007, three months after arriving at Harvard, Thrish became one of the scientists and writers supported by the Scholars at Risk program, administered by the Harvard University Committee on Human Rights Studies.

A gifted researcher, he continued work on land-mine-hunting robots by starting work with experts at the Massachusetts Institute of Technology (MIT).

A year later, Thrish was named a 2008-09 Radcliffe Fellow. He finished two books: a novel inspired by his tsunami experience and a college textbook on machine intelligence systems.

But in two years at Harvard, the focus of this slight, modest scientist has been an investigation of how animals process odors, and exactly how they walk.

Knowing more about these fundamental mammalian processes, said Thrish, will usher in a new generation of robots — self-learning, nimble, and equipped with sensitive (though artificial) noses.

In looking for land mines, such robots even in their present form reduce risk to humans and protect the environment. Once precisely located, the mines can be dug up without removing layers of fertile topsoil or virgin trees.

Thrish’s working model — a robot the size of a serving platter — has eight legs, one motor, springlike joints, and low clearance (6 inches). It’s small enough to squeeze under barbed-wire fences and light enough to not set off a land mine.

The mongoose is light too — less than 10 pounds. It’s a slender, intelligent mammal with the best developed sense of smell in nature, after the elephant and the pig. A third of its brain function is given over to processing odor signals.

The mongoose also has excellent locomotive skills in a forest environment, and can be trained to detect buried explosives quickly, said Thrish. “100 percent detection takes just three weeks.”

At Harvard, Thrish worked with molecular biologist Naoshige Uchida on rodent olfaction. He also worked with Robert Howe, Harvard’s Gordon McKay Professor of Engineering, and assistant professor of computer science Radhika Nagpal.

They supervised Matthew Valente ’09, who used Thrish’s improved understanding of legged locomotion in rough terrain to build a field model robot that was tested in mud, brush, and sand.

At MIT, Thrish worked with Russ Tedrake’s Robot Locomotion Group to study the walking patterns and to understand what causes uncertainty in leg locomotion, and how animals cope with it.

A major barrier in robot research is this uncertainty. “Most robots,” said Thrish, “get stuck or fall down in uncertain environments.” With a laugh, he called this emerging field of machine motion “careful walking.”

Thrish no longer works on robots coupled with rodents, but he still studies odor-guided behavior and legged robot locomotion.

But the mongoose and its ilk will have a continuing role, said Thrish. They “teach” self-learning robots how to move through cluttered terrain, and continue to provide scientists with models for man-made odor sensors.

In April, Thrish helped organize a two-day conference on humanitarian de-mining — the first ever at Harvard and MIT. It included experts in artificial noses, as well as in field robotics, odor-guided behavior, seismic sensing, and humanitarian action.

In June, Thrish and his family will depart for a new life in the United Kingdom, where, at King’s College, University of London, he’ll continue research on legged locomotion for robots, deformable robotic bodies, and primate bipedal locomotion.

At Harvard, Thrish burrowed into his science, but he also broadened his perspective on the land mine detection problem, connecting up with experts in political science, philanthropy, humanitarian aid, and other humanistic disciplines.

“The engineering we do,” he said, “cannot be disassociated from the humanities, from the environment, or from politics.”

Scholars at Risk

The Scholars at Risk program at Harvard started in 2002, when it sponsored an Ethiopian geographer imprisoned for his work. (He showed that a certain famine had political, not natural, causes.) Since then, the program has found temporary academic sanctuary for 26 scholars who face harassment, imprisonment — or worse — in their native countries. Many choose to remain anonymous.

In 2009-10 the program will host four scholars — from Eritrea, Iran, Iraq, and Pakistan. The 2008-09 class, including Thrishantha Nanayakkara, numbered six — from Iran, Iraq, Sri Lanka, Syria, and Uzbekistan. Organizers have awarded Harvard fellowships to novelists, anthropologists, composers, physicists, historians, human rights advocates, and scholars of law, government, and literature.

The Harvard program is affiliated with the Scholars at Risk Network, an international consortium hosted by New York University.