Rainbow Dash can walk on many surfaces such as a foam mattress or a doormat with many random folds.
Robotics keeps advancing, and one example is a robot named Rainbow Dash that can teach itself to walk. This four-legged robot model needs only a few hours to learn to walk backward and forward and to turn right and left.
Researchers from Google, UC Berkeley and the Georgia Institute of Technology published a paper on ArXiv describing a statistical AI technique called deep reinforcement learning, which they used to create a new generation of intelligent robot achievements.
Most earlier self-learning techniques took place in computer simulations. Rainbow Dash, however, used this technology to learn to walk in a real physical environment. Moreover, it can do so without a dedicated teaching mechanism, such as an instructor or pre-programmed data. Rainbow Dash succeeded in walking on many surfaces, including soft foam mattresses and doormats with many random folds.
The deep learning techniques the robot uses include a type of machine learning that continually tries and errs through repeated interaction with the environment. This is similar to computer games in which digital methods learn how to play in order to win. This form of machine learning differs sharply from traditional supervised or unsupervised learning, in which machine learning models require clearly labeled training data. Deep reinforcement learning combines reinforcement learning methods with deep learning, in which the scale of traditional machine learning is greatly expanded by the power of massive computation.