We have already covered a bunch of self balancing robots. Getting them to do front flips and bounce on one wheel is very impressive. That’s exactly what the RAI Institute researchers have accomplished. They used reinforcement learning to pull this off. This robot can essentially move like a world class athlete. It has 2 motors in the lower half that control steering and velocity of the bike. The upper part has 4 motors that handle jumping.
This robot has a roll cage, steering column, and carbon fiber frame. It uses IMUs to measure acceleration and gravity. It can also estimate contact time, magnitude and direction of the ground reaction forces with its sensors. It also has a LiDAR and forward camera for navigation.
[HT]