U of Michigan robot learns to walk on rugged, uneven terrains

The University of Michigan has a bipedal robot, and it is able to walk without support. The researchers working with the robot recently showcased their creation walking on unstable and uneven ground, marking an update in its movement capabilities. The types of difficult terrain include ones covered with a "thin layer" of snow and steep slopes. The algorithms powering the robot's walking abilities could one day aid other bipedal robots in tough environments.

The robot is called MARLO, and it is capable of "3D walking," a type of walking that means the robot is able to walk in any direction rather than in just two dimensions. The previous robot, MABEL, was limited to a boom and 2D walking. MARLO was built by Oregon State University's Jonathan Hurst.

Interestingly enough, the team figured out how to get MARLO to walk in 3D space using 2D algorithms; there are two controllers instead of one, with one controller handling the side-to-side movements, and the other controlling forward-and-backward movements. One of the team members, Xingye Da, built 15 different gaits into a library that pertain to various ground heights and walking rates.

Equipped with this mixture of hardware and software, MARLO takes steps that can be adjusted to an environment by adjusting as necessary to different ground heights, blending library gaits as needed to achieve her movements. In the future, MARLO may be equipped with a fully 3D controller rather than the aforementioned pair, enabling her to better adjust her speeds in difficult terrains.