Researchers supported by the US National Science Foundation taught a robotic mini-cheetah how to run – fast. The robotic cheetah, which was trained by simulated experience to adapt to changes in terrain, broke the record for the fastest run ever recorded.
The team used a “learn from experience” model to train the robot cheetah. Humans have created robots that can walk, lift and jump, but fast and efficient running is not yet in the repertoire of robotic animals. Running requires the robot to respond to rapid changes in the environment and terrain.
Using learn-by-experience models, artificial intelligence and machine learning, the team taught the robotic cheetah how to adapt to changes in its environment while in motion. Using simulated scenarios, the robot can quickly experience and learn from different terrains.
The researchers say that manually training the robot to adapt is a time-consuming, laborious and tedious endeavor. Scientists believe that teaching robots to themselves may solve the problem of scalability and allow robots to build more diverse skills and tasks. They have now begun to apply their approach to a wider set of robotic systems.