Getting a robot to walk is one thing, getting it to walk without tripping on the first obstacle it encounters is quite another. Engineers at the University of Michigan are developing a set of algorithms that allow an unsupported bipedal robot named MARLO to negotiate steep slopes, thin layers of snow, and uneven, unstable ground without toppling over. Designed as a general purpose robotic system, the algorithms may also have applications in advanced prosthetics.
According to the Michigan team led by professor of electrical engineering and computer science, Jessy Grizzle, MARLO has the best walking ability of any robot not equipped with powered ankles. It builds on the work of their previous robot, MABEL, which was a bipedal running robot capable of reaching speeds of up to 10.9 km/h (6.8 mph). However, the older robot could only walk back and forth and had no lateral stability. MARLO, on the other hand, is designed to move and balance in any direction.
MARLO is operated by means of a conventional Xbox controller, which allows the operator to give commands to move and in a given direction. Once it is under way, the robot makes its own decisions as to how to handle rough terrain, such as a test course of inclines, uneven plywood squares covered in astroturf, and bits of foam rubber.
This is not the first robot capable of handling such conditions, but other successful designs, such as the latest Boston Dynamics Atlas robot, use powered joints and a battery of sensors to maintain balance, while MARLO is much simpler.
"The robot has no feeling in her tiny feet, but she senses the angles of her joints — for instance, her knee angles, hip angles and the rotation angle of her torso," says Grizzle. "It's like walking blindfolded and on stilts."
MARLO's capabilities are due to advances made in its navigation algorithms by doctoral student Xingye (Dennis) Da, which combine two 2D algorithms – one for forward and back stability and another for lateral stability. According to Da, this greatly speeded up development. He then wrote a library of 15 gaits for different walking speeds and ground heights, so that MARLO could, without special sensors, recognize changes in the ground height and alter its gait accordingly.
The team says that while MARLO is a great advance over MABEL, it still has a lot of room for improvement, such as in making quick turns or stepping sideways. To achieve this end and to make its responses faster, doctoral student Brent Griffin is working on a fully-integrated 3D controller algorithm that keeps the robot running at an optimal speed as it handles terrain. In addition, the team is also developing a "super-algorithm" to make MARLO more agile.
"We are able to design full 3D walking gaits using a mathematical model of the robot and then apply them directly to MARLO," says Griffin. "Because the implementation works without any robot-specific modifications, it is generalizable to other walking robots."
One important aspect of the team's work is that the algorithms must have general applications, so they can be used in other walking robot projects. One early result of this was at the University of Texas-Dallas, where assistant professor of mechanical engineering and bioengineering Robert Gregg used the algorithm to control a robotic lower-leg prosthesis, which enabled an amputee patient to walk naturally in treadmill tests.
"The ability of MARLO to gracefully navigate uneven terrains is very exciting for my work in prosthetics," says Gregg. "We hope to encode similar abilities into our robotic prosthetic leg so that lower-limb amputees can just as easily walk about the community without having to think about the terrain."
See more of MARLO in action below.
Source: University of Michigan