Carnegie Melon professor develops real-time learning four-legged robot

Meet the new four-legged robot that’s almost human, but not quite.

The artificial intelligence robot developed by Carnegie Melon University Professor Deepak Pathak can adapt in real-time and teach itself how to maneuver through rocks, sand, grass and oil-covered surfaces.

Looking something like the bare bones of a small animal, the robot learns to react to new situations the same way people do.

Most robots aren’t able to adjust to changing conditions so quickly, but Rapid Motor Adaptation seeks to change that. The technology is the first learning-based system that doesn’t rely on hand-coded motions.

The RMA adapts robot joints in real-life to auto-adjust to changing situations. It is a step toward building AI systems that can do the same.

The research team says the robot will eventually be used for search and rescue efforts, and space exploration.