Martian Rocks, Robot Retrieves

Extreme Explorers' Hall of Fame
The autonomous Antarctic meteor finder, Nomad, uses artificial intelligence to recognize and classify promising rocks
Credit: Carnegie Mellon,

Carnegie Mellon University’s Nomad robot conducts autonomous searches for meteorites in Antarctica. The meteor finder has examined more than 100 indigenous rocks, studied about 50 in detail and classified seven specimens as meteorites.

"This is a small step for a robot and a big step for robot kind," said William L. "Red" Whittaker, principal investigator for the Robotic Search for Antarctic Meteorites initiative.

The robot enters the Extreme Explorers Hall of Fame, as the first such ‘smart’ sample finder. Nomad’s expedition and the discoveries it has made are significant because it marks the first time a robot, relying on sensors and artificial intelligence, has been able to find a meteorite lying on the ice and distinguish it from ordinary rocks in the area.

"This marks the first discovery in the natural world by robotic machine intelligence and sets a precedent for a new class of robotic science on Earth and in space," said Whittaker.

Far From Home Alone

The robot made its discoveries at Elephant Moraine in eastern Antarctica, 160 miles northwest of the United States base at McMurdo Station. The area is an important site for meteorite discovery, with nearly 2,000 specimens recovered during seven previous visits, including the first meteorite identified as definitely being from Mars.

This expedition took place in an area that hadn’t been searched in nearly a quarter century-last examined in 1979 by scientists from the National Science Foundation’s Antarctic Search for Meteorites (ANSMET).

"The robot correctly classified three other indigenous meteorites and misclassified one as terrestrial rock", said Dimitrios Apostolopolous, a systems scientist at Carnegie Mellon’s Robotics Institute and project manager of the Robotic Antarctic Meteorite Search initiative. "Nomad achieved these results autonomously and without any prior knowledge about the samples."

Most of the rock fragments – called chondrites – that Nomad found are relatively common types, composed mainly of rock with small metallic infusions that probably originated from asteroids.

One achondrite meteorite which Nomad classified as interesting is so rare that the robot didn’t have the data in its base to make a determination.

Typically, the robot travels over an area the size of a football field in patterns similar to those a person would make when mowing a lawn. The search pattern is sometimes referred to as a linear search, a method commonly also applied by law enforcement cadets when they look for forensic evidence at a crime scene. The robot does not try to look-ahead or focus its search in an area that it has previously made discoveries. Thus the goal is comprehensive coverage in a first pass through a region of interest.

Antarctic meteor map
Major investigated regions of Antarctica where meteors have been successfully identified
Credit: JSC/NASA Meteor Program

To fool the ‘smart robot’, the search site also contained false meteorite look-alikes, sometimes called "meteorwrongs," as well as typical Antarctic rocks.


When Nomad encounters a promising rock, it deploys its manipulator arm containing a high-resolution camera and a spectrometer to gather visual images and spectroscopic data upon which to determine a specimen’s composition.

When Nomad found its first meteorite, it had already completed 350 meters of linear searches and had examined seven other rocks.

"With these finds, Nomad takes robotics beyond the typical concerns about nuts and bolts and into the universe of scientific inquiry," said Ralph Harvey, principal investigator for ANSMET.

With the help of machine learning and statistical techniques, Nomad puts a numerical value on its confidence that a rock is or is not of extraterrestrial origin. It classified its first meteorite with a confidence rating 2.5 times higher than any other rock it examined. The more rocks it studied, the higher its confidence rating went in making its determinations.

Martian Paradox: Why Do the Alien Rocks Keep Visiting Earth?

The November 4 issue of Science Magazine has further piqued meteor-hunter’s interest in the large interplanetary flux of such rocks, particularly as samples that might be identified by robotic searchers as originating on Mars.

Computer simulations published by James Head and Jay Melosh of the University of Arizona, with Boris Ivanov of the Russian Academy of Sciences, have tried to field a question that has long plagued meteor hunters: Why does the Earth continue to be visited so frequently by Martian rocks?

The question involves a paradox. It takes a big impact to fling rocky remnants into Martian orbit or beyond. Five kilometers per second [11,180 miles per hour] is the escape velocity of Mars, the speed needed to leave the planet without going into orbit around it. The number of such big events is expected to be too rare to account for what has been discovered so far. Previous guesses showed that it might take up to a 8 mile-wide rock to launch fragments towards Earth.

That’s Funny….

As writer, Isaac Asimov pointed out: ‘The most exciting phrase to hear in science, the one that heralds new discoveries, is not "Eureka!" (I found it!) but "That’s funny …"’. Indeed, the Martian paradox is a case of some funny mathematics conflicting with what is observed.

The problem is that despite the need for a big impact or explosion to sample a Martian meteorite arriving eventually on Earth, there are relatively many of the samples. So far, meteorite hunters have found about 26 rocks on Earth that have been identified as having come from Mars (some of these broke apart upon entering the atmosphere, so the 26 rocks were found as about 40 separate pieces).

Scientists had thought it took a serious wallop to instigate these interplanetary exchanges. Impacts of this size and larger occur every 200,000 years or so on Mars. Yet the new research finds that craters as small as 1.9 miles (3 kilometers) wide on Mars could have been the starting points for rocky odysseys.

This minimum crater diameter is at least four times smaller than previous estimates. The simulations show such 2 mile wide impacts from the past may include more than 10 million rock fragments average 2 inches [5 cm] or larger.

What’s Next

Most remarkably, at any given moment, the interplanetary sample transit works out to about one Martian meteorite landing on Earth each month.

In addition to the two dozen Mars samples, more than 20,000 space rocks have been collected for study.

So, the Nomad robot has both a good-sized database to learn from and much travelling to do. "We all dream of a future where robots can act and perhaps even think independently," he added. "Nomad has now taken the first steps along that path," concluded Harvey.

Nomad’s expedition to Elephant Moraine is a collaborative effort between the Robotics Institute at Carnegie Mellon University and the National Science Foundation’s (NSF) Antarctic Search for Meteorites program. It is being performed under the auspices of NSF’s Office of Polar Programs. The Nomad robot has been developed through research at Carnegie Mellon’s Robotics Institute and is funded by grants from NASA’s Surface Systems Thrust of the Cross Enterprise Technology Development and Space Telerobotics programs.

Related Web Pages:

National Science Foundation
Robotics Institute
Antarctic Search for Antarctic Meteorites Program (ANSMET)
Big Signal Project
Atacama Desert Expedition
Patriot Hills Expedition
Nomad Engineering
The Planetary Materials Curation
Meteorite Repository (JSC) Rock Descriptions
NIPR Meteorite Collection (Japan)