Today’s robots easily execute “pick-and-place” tasks: finding an object in one specified location and bringing it to another. But what if there are obstacles in a robot’s way? And what if the machine is on uncertain terrain, like the surface of the Moon?

Carnegie Mellon University’s Home Exploring Robot Butler, or HERB, uses new software to sort clutter more efficiently. (Image Credit: CMU)

Researchers at Carnegie Mellon University (Pittsburgh, PA) have developed an algorithm that enables robots to understand their environment and find ways around clutter. In collaboration with NASA, the CMU team will test its software on the agency’s K-Rex lunar rover.

A Rover Learns Its Moves

With the new Carnegie Mellon technology, a rover is given a rough geometric model of what the world looks like. Then, through simulations, the robot can experiment with the consequences of manipulating nearby objects.

“It’s literally like playing chess,” said Siddhartha Srinivasa, associate professor of robotics, “but playing chess in the real world by pushing, pulling, moving, and sliding objects around.”

The robot is programmed to understand the basic physics of its world: what can be pushed, lifted, or stepped on, for example. Once a solution and pathway configuration is found, closed-loop feedback controllers keep the machine on track.

Using cameras and perception technologies, a robot “sees” what the world looks like. The additional algorithms find a path through the uncertain terrain.

“Just like when you’re trying to go through a maze to find a goal, you’re actively seeking out information, you’re actively trying to probe the world, and you’re actively trying to be very goal-directed,” said Srinivasa.

The rearrangement planner software was developed in Srinivasa’s lab by Jennifer King, a Ph.D. student in robotics, and Marco Cognetti, a Ph.D. student at Sapienza University of Rome, who spent six months in Srinivasa’s lab.

The algorithms were first tested on Carnegie Mellon’s Home Exploring Robot Butler (HERB), a 4-foot, 6-inch-tall, 400-pound machine designed to assist in household tasks like opening doors and retrieving milk from the refrigerator.

The K-Rex rover during an engineering field test at the Basalt Hills quarry, California in October 2012. (Image Credit: NASA)

The same insight that allows a robot to reconfigure the items in a refrigerator, however, can be transferred to rover navigation.

Preparing for the Moon

With wheeled-rover navigation, terrain must be frequently examined in advance so that the vehicle avoids non-traversable areas. The Carnegie Mellon researchers want to equip NASA’s K-Rex robot with the ability to actively landscape its world and clear out efficient traversal routes in real time.

“What if your rover realizes that it has to go past a big hole, and that if it pushed a bunch of rocks, it could actually close up that big hole and create a traversability and go over that?” said Srinivasa.

The Carnegie Mellon team has worked closely with Terry Fong, Director of the NASA Intelligent Robotics Group at NASA Ames Research Center, preparing the K-Rex technology for its lunar mission in 2018. As a NASA Space Technology Research Fellow, CMU’s King has collaborated with NASA researchers on the rover’s interaction with the environment.

In addition to four-wheel-drive and all-wheel-steering, the electric-power K-Rex rover features lidar, stereo cameras, two inertial measurement units (IMUs), a compass and inclinometer, differential GPS, and vehicle odometry.

Designed to move autonomously, the K-Rex travels at slow speeds, between 0.25 and 3.22 miles per hour. The mobile robot supports scouting, mapping, sampling, and site preparation applications.

This summer, the team will perform tests at NASA’s Roverscape, a rover test area at Ames Research Center in Mountain View, CA. The K-Rex rover will attempt to achieve a goal by learning to reconfigure terrain and move objects around the simulated lunar landscape.

In initial tests, the robots will push easier-to-model objects, like cardboard boxes. If successful, the team will then incorporate more complex items, including rocks and other targets that the rover is more likely to encounter.

Srinivasa says the CMU-developed algorithm is a potentially “disruptive” technology for rover navigation, particularly as engineers focus on payload optimization and squeezing as much functionality out of existing hardware as possible.

“We want to be able to enable robots, especially space robots, to be able to use every part, every ounce of their body, their metal, their geometry, to be more effective at manipulating the world,” said Srinivasa.

This article was written by Billy Hurley, Associate Editor, NASA Tech Briefs. For questions and comments, email This email address is being protected from spambots. You need JavaScript enabled to view it.. For more information, visit .