Robot Bicycle Mimics Batman's Batpod with Stunts
Engineers develop an AI-powered two-wheeled robot capable of wheelies, jumps, and front flips.
Researchers have developed an AI-powered robot bicycle, dubbed the Ultra Mobility Vehicle (UMV), that can perform dynamic stunts similar to those seen in Hollywood films. Inspired by the agility of mountain bikers, the UMV balances on two wheels and can execute maneuvers like wheelies, track stands, and jumps, drawing parallels to the iconic Batpod from the Batman franchise.
At the core of the UMV's design is a powerful internal moving mass that allows the robot to rapidly shift its weight and control its momentum. This mechanism enables it to maintain balance and execute complex aerial feats. The robot, weighing 23.5 kg, can travel at eight meters per second and jump over obstacles up to 130% of its own height. Researchers at the Robotics and AI Institute in Massachusetts created the UMV, publishing their findings in the journal Science Robotics.
To achieve its capabilities, the team utilized computer simulations to test various designs before optimizing for jumping height and single-wheel balance. Unlike conventional robots that might use multiple wheels or legs, the UMV relies on its dynamic weight shifting and momentum for stability. A significant portion of the robot's mass is located in a movable 'head' unit above the frame, allowing it to mimic the movements of a human rider performing tricks.
The AI system was trained using reinforcement learning, a trial-and-error artificial intelligence method. The robot practiced thousands of simulated maneuvers before successfully transferring these skills to its physical prototype. Through this training, the UMV learned to ride forward and backward, balance stationary, perform sideways hops on a single wheel, and turn in place. One emergent behavior was a "shimmy turn," which the AI developed autonomously.
Further demonstrations showcased the UMV's ability to autonomously approach obstacles, jump onto a one-meter-high platform, traverse it, and dismount. It also demonstrated a full front flip in under half a second, landing and continuing to ride. The researchers suggest that future iterations of this technology could have practical applications in areas such as urban deliveries, reconnaissance missions, or navigating difficult terrain where conventional vehicles would struggle.
