Researchers at the Robotics and AI Institute and collaborating organizations have developed a new robotic platform called the Ultra Mobility Vehicle (UMV), a bicycle-like robot capable of moving at high speed while performing agile maneuvers such as hopping over obstacles and balancing on a single wheel. The project aims to demonstrate that a relatively simple robotic structure can achieve athletic movement comparable to far more complex legged robots, tells Tech Xplore.
Human cyclists routinely perform impressive feats of balance and agility, including wheelies, hops, and jumps across uneven terrain. Inspired by these abilities, the research team designed UMV to mimic similar movements. The robot uses two in-line wheels, a steering frame, and a rear-wheel drive system in which a motor powers the rear wheel while the front wheel controls direction.
A key innovation lies in the robot’s mass-shifting mechanism. Most of the robot’s weight is concentrated in a “head” unit that contains the batteries, processors, and actuators. This head connects to the bicycle frame through a mechanical linkage resembling a neck and tie rods. By rapidly moving this mass, the robot can generate momentum that allows it to balance dynamically, crouch, hop, or jump onto elevated surfaces.
Despite its capabilities, the system remains mechanically simple. The UMV robot has only five actuated degrees of freedom, far fewer than the dozen or more joints typical of many legged robots. Fewer joints reduce weight, cost, and mechanical complexity while improving reliability.
The robot’s movements are controlled using reinforcement learning, a machine-learning method that allows the system to learn behaviors through simulated training. After training in simulation, the learned policies transfer directly to the real robot without additional tuning. This approach enables the robot to perform diverse actions such as pivoting in place, hopping on its rear wheel, and executing dynamic jumps.
In testing, the robot achieved speeds of about 8 meters per second and successfully jumped onto platforms roughly one meter high, exceeding its own height.
Researchers believe the platform could serve as a testbed for future mobility systems capable of navigating complex environments efficiently. By combining the efficiency of wheels with the agility typically associated with legged robots, the design suggests new possibilities for robots that must move quickly across unpredictable terrain.