Introducing the UMV Robot
The Robotics and Artificial Intelligence Institute (RAI Institute) has unveiled an innovative robot named UMV, which uniquely combines the mechanics of a bicycle with the agility of legged robots. This cutting-edge machine can reach top speeds of 28.8 km/h and execute impressive stunts, including jumps, wheelies, and front flips.
Design Features and Training Process
Weighing 23.5 kg, UMV was developed using reinforcement learning techniques. Its training took place in a simulated environment that accounted for multiple factors such as:
- mass distribution
- surface traction
- motor power
- control delays
- sensor inaccuracies
The team utilized the NVIDIA Isaac Simulator alongside the Proximal Policy Optimization algorithm to train specialized control models tailored for different movement modes—regular riding, wheelies with jumps, flips, and hops. These models transitioned seamlessly to the physical robot without requiring further on-device training.
During testing, UMV demonstrated remarkable capabilities, such as climbing onto platforms as high as one meter, traversing them, and jumping down safely. The robot successfully repeated a climbing sequence 15 times with consistent stability. When performing front flips, UMV compresses its frame mid-air to reduce rotational drag, achieving an angular velocity near 17 radians per second. The entire flip, from takeoff to landing, lasts under half a second, highlighting the robot’s precision and control.
Trials were conducted inside a laboratory equipped with motion capture technology to accurately track UMV’s position and movements. Researchers plan to enhance the robot’s design further by improving battery life, motor drivers, and onboard sensors to reduce reliance on external systems. Potential applications for UMV include urban cargo delivery and exploration across complex terrain over long distances.
The development of UMV marks a significant milestone in robotics by merging the strengths of wheeled and legged locomotion. Its speed and maneuverability make it a promising tool for logistics and field research. Advances in battery and sensor technology are expected to greatly enhance the performance and autonomy of such robots in the near future.
In addition to the impressive capabilities of the UMV robot, the field of robotics continues to evolve with innovations like a miniature robot that excels at jumping, climbing, and swimming, showcasing the diverse applications of advanced engineering. These developments highlight the growing potential of agile robots in various environments.