The desert is a tricky place for robots to navigate - just ask Aaron Johnson. The assistant professor of mechanical engineering at Carnegie Mellon University recently won the Army Research Office’s Young Investigator Award for his work designing intelligent interaction between robots and their environments.
Johnson’s experiences testing robots in the Mojave Desert as a Ph.D. student at the University of Pennsylvania cemented his interest in getting robots to overcome challenging terrain, and he started thinking about jumping and leaping behaviors.
"It was clear that we could handle some terrain but not others-the bots had particular difficulty with areas where the rocks were bigger than their legs," said Johnson, who earned his bachelor’s degree from CMU’s Department of Electrical and Computer Engineering.
He will apply these ideas to his newly funded project as he investigates ways to model uncertainty when faced with rocky hills.
Johnson’s research applies to all types of challenging terrain, but this project will focus specifically on getting robots to climb steep, rocky hills.
"What makes this challenging is the uncertainty that comes with the rough terrain," said Johnson, who also has a courtesy appointment in the Department of Electrical and Computer Engineering and the Robotics Institute , where he also was a postdoctoral fellow. "It’s not necessarily the steepness of the slope but the unevenness and the fact that every step is going to be a little bit different."
Despite continuing advancements with cameras and sensors, uncertainty plays a large role when trying to get a robot to climb a rocky hill, he said. Even the best cameras cannot overcome certain perspective issues, such as seeing the top of a rock that the robot has to jump up onto. There also are uncertainties associated with predicting contact variables, such as friction and the actual shape of the terrain. The slightest error in calibration, execution or when calculating contact can cause the robot to slip or fall.
Johnson’s project also will address robust robot behavior, which means that the robot can recover if it makes a mistake. A robot that is robust to changing conditions means a robot can be more successful in crossing challenging terrain. To design robust behavior, Johnson will work on developing robots’ feedback controls systems in conjunction with trying to model contact condition uncertainty.
For his project, Johnson will work with both wheeled and legged robots because robot morphology, or design, affects applications.
"There’s only been a few places that I’ve seen a legged robot do something that a wheel really couldn’t do with all-wheel drive on," Johnson said. He added that wheeled robots require much more skilled "drivers" than legged robots to cross rough terrain. Johnson said he suspects that research is moving toward a machine design that requires legs rather than wheels, but his overall project will help him compare each type of robots’ best performance and determine when each type of robot should be used.
The United States Army’s Young Investigator Award funds research proposals that address the nation’s security needs, and robots present a mostly untapped security source. Johnson’s main objective is to develop robots that can go anywhere in the world over any type of terrain, which would allow the Army to use robots in lieu of soldiers for certain tasks or operations.
Robots that can go anywhere in the world also opens up other applications such as environmental monitoring and exploration. As the Earth’s climate changes, robots that can reach remote areas would substantially help scientists predict these changes and potentially help them design countermeasures. Advancing robots’ ability to cross rough terrain also will result in robots that can better navigate other unstructured environments such as cluttered homes and offices.