Programming robots to walk on flat, even ground is difficult enough, but Florida’s Institute for Human and Machine Cognition (IHMC) is talking the grander challenge of making sure bipedal robots can successfully navigate rough terrain. The research organization has been demonstrating its work in this area since 2016, but its latest video (via Engadget) shows the progress its made.

In the new video, IHMC’s autonomous footstep planning programming is at work on both Boston Dynamics’ Atlas robot, and the NASA-developed Valkyrie robot (humanoid robots have the coolest names). This video shows off navigation of a heaping pile of cinder blocks, as well as narrower paths which are trickier to navigate because of limited navigation options.

Basically, IHMC manages these complex navigation operations by specifying a beginning and end point for the robot, and then mapping all possible paths on a footstep-by-footstep basis, evaluating the cost of each and ultimately arriving at a best possible path – all of which can occur relatively quickly on modern hardware.

These robots can also quickly adapt to changes in the environment and path blockage thanks to IHMC’s work, and it can even manage single-path tightrope style walking (albeit on a narrow row of cinder books, not on an actual rope).

There’s still work to be done – the team at IHMC says that it’s having about a 50 percent success rate on narrow paths, but its ability to navigate rough terrain with these robots and its software is at a much higher 90 percent, and it’s pretty near a perfect track record on flat ground.