Researchers taught a robot dog to skateboard, using an approach that mirrors how humans learn complex physical skills.
The University of Michigan team's quadrupedal robot can mount a skateboard, push off to gain speed, glide smoothly, and navigate challenging terrain — from carpets to slopes to steps.
Their system, called Discrete-time Hybrid Automata Learning (DHAL), helps the robot naturally transition between different motion states without needing pre-programmed movements. Like a human skater, it smoothly switches from pushing to gliding and back again. The researchers added LED indicators that change color to show which "mode" the robot is operating in — green for pushing, red for gliding, blue for transitions.
The system achieving 100% success rates on basic surfaces and maintaining 80% success even on slopes.
Previously:
• Simone Giertz makes an electric skateboard for Scraps the dog
• Time lapse: making a cruiser skateboard
• Athletic kitten sees a skateboard and immediately figures out how to ride it (video)
