Robots, just like humans, require sensors in order for them to “see” like we would with our eyes. However, also just like humans, these sensors don’t always work so well especially in dim lighting conditions or in the dark. In such a scenario, a robot would almost be useless, but that’s not the case with Cassie.
Cassie is a robot developed by researchers from Oregon State University which has successfully completed a stair climb all by itself. What makes this accomplishment so special is the fact that Cassie is a “blind” robot, meaning it is a robot devoid of the typical sensors, like cameras, that you might find in other self-navigating robots.
To help Cassie walk up a flight of stairs without sensors, the researchers used what is known as sim-to-real Reinforcement Learning to teach it about proprioception, also known as body awareness. Basically this involved the robot learning how to handle various situations and terrain, like stairs and flat ground, and it ultimately learned how to do it with about 80% efficiency ascending stairs, and 100% efficiency at descending stairs.
However, it should be noted that during the trials, the robot was made to run at a set speed because the researchers found that if it ran too fast or slow, it increased the chances of failure. Despite those caveats, the researchers are optimistic as the test demonstrated the capabilities of blind locomotion in robots which could open the door to more possibilities in the future.